Our clients are industry-focused digital marketing companies or large organizations that manage their digital marketing in-house

Fluency

fluency

It’s always fun when extremely talented engineers bring up the topic of continuing down an individual contributor path versus entering formally into people leadership. I recently had the chance to talk about the usual trade-offs in conversations about people leadership vs individual contributions. Topics like, do you think you can effectively lead through others, or are you a brilliant engineer? Are you sure you want to manage people instead of doubling down on your expertise? Naturally, The conversation generally shifts to an equally fascinating topic: quantifying effectiveness. For individual contributors, effectiveness can be measured in various ways, such as algorithmic efficiency or the robustness of systems and platforms they develop. In contrast, measuring performance and effectiveness in leadership introduces a different kind of subjectivity. It tends to lead into an opportunity where I quote from a concept from my favorite leadership book (at this point, I should use an Amazon federal link, hah) that followership is a useful measure of leadership effectiveness because it's both simple and quantifiable. Consider this: how many people you lead would choose to follow you if you switched teams, organizations, or even companies? This question underpins the argument in , where the author challenges the conventional wisdom of universally definable and measurable leadership. Instead, the book suggests that "followership" is the real measure of leadership effectiveness, proposing that the essence of leadership is not a singular trait or capability. Nine Lies About Work Coincidentally, my friends at Quotient a that offers complementary insights, identifying key attributes of effective engineering managers from both managers' and engineers' perspectives. These attributes include fostering a positive work environment, enabling autonomy, and nurturing talent. Notably, the study suggests de-emphasizing technical expertise in favor of qualities that promote team cohesion, psychological safety, and personal growth, reflecting a shift in how leadership effectiveness is perceived in the technical domain. recently analyzed Microsoft study However, I have reservations about the interpretations presented in the Microsoft research paper, especially the assertion that effective managers place less emphasis on technical leadership. This discrepancy might stem from the definition of 'technical' prowess, particularly within software and machine learning engineering. From my experience, there's a correlation between effective leadership and technical acumen. However, I'd reinterpret 'technical' abilities to include the capacity to engage in multiple "languages of abstraction." Exceptional leadership, especially in technical disciplines, involves the versatility to engage scientifically with data and empirical evidence, mathematically to argue points with precision, and logically to construct cogent arguments by effortlessly forming valid arguments (any casual reference to using modus ponens or hypothetical syllogism is a win). This abstract multi-lingual fluency enables leaders to connect with their teams on multiple intellectual levels, fostering a deeper appreciation for the leader’s technical capabilities. I think effective leadership, within most engineering domains, hinges as much on technical fluency as it does on vision, empathy, and adaptability. It's this synthesis of skills and the ability to create an environment where folk would happily follow — that distinguishes truly effective leaders.

Embracing Followership and Technical Fluency in Modern Leadership

Authors:
(1) Prerak Gandhi, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, prerakgandhi@cse.iitb.ac.in, and these authors contributed equally to this work;
(2) Vishal Pramanik, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, vishalpramanik,pb@cse.iitb.ac.in, and these authors contributed equally to this work;
(3) Pushpak Bhattacharyya, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai. Table of Links Abstract and Intro
Motivation
Related Work
Dataset
Experiments and Evaluation
Results and Analysis
Conclusion and Future Work
Limitations and References
A. Appendix A. Appendix A.1. Ethics Consideration We have taken all the scripts from IMDB and IMSDb databases. The website has a disclaimer regarding using its scripts for research, which can be found at this link https://imsdb.com/ disclaimer.html. We have used the scripts fairly and without copyright violation. A.2. Annotator Profiles We required the help of external annotators in two cases: (i) Manually Annotating the Scripts and (ii) Creating scenes and their descriptions from the scripts. For the first task, we took the help of 10 annotators. Their ages ranged from 21-28, and all were Asian. They were given detailed guidelines with examples for annotating. There were also periodic sessions to confirm their understanding and solve their doubts and mistakes. For the second task, we took the help of two annotators. Both of them are Asian females aged between 21-23. Both of them were given detailed guidelines for the scene-writing task. A few data points were picked randomly and checked to find out and correct conceptual mistakes. The annotators had bachelors and masters degree in STEM and Arts. A.3. Evaluation Metrics The evaluation metrics are described below: • Perplexity (PPL): Perplexity is one of the most common metrics for evaluating language models. They are computed as exponential of entropy. The smaller the value of the PPL, the greater the fluency of the generated text. • BLEU: BiLingual Evaluation Understudy is a common metric in many NLP tasks, especially in the field of Machine Translation. It measures the overlap between the generated output and gold standard data. Although this metric does not consider the model’s creativity, we can deduce the difference between the candidate text and the reference text using BLEU. The higher the BLEU measure, the better it is. • ROUGE: Recall-Oriented Understudy for Gisting Evaluation is typically used for evaluating automatic summarization. In our case, it measures the longest overlapping sequence between the generated and original plots. The higher the ROUGE measure, the better it is. • N-grams: We measure the redundancy and diversity of the movie plots by computing the repetition and distinction n-gram scores. A.4. Screenplay Structure A movie script or a screenplay has a different format than a story. A script is a group of scenes. Each of these scenes consists of a few major components, which are discussed below: Scene Headings/Sluglines- This component describes the when and where of the scene. It can be thought of as the first shot that a camera takes of a new scene. For example, INT. - RESTAURANT - NIGHT indicates that the scene starts inside a restaurant at night. Sluglines are normally written in capital letters and are left-aligned. Character Names- they are mentioned every time a character is going to utter a dialogue. The name of each character is mentioned in uppercase and is centre aligned. Dialogues- dialogues are the lines that the characters say. They appear right after the character name in a script and are centrally aligned. Action Lines- action lines describe almost everything about a scene. They can be described as the narration of each script. Action lines can be present after either dialogues or sluglines and are left-aligned. Transitions- a transition marks the change from one scene to the next. They also depict how a scene is ended. For example, DISSOLVE, FADE, and CUT are different keywords used to indicate a transition. They are usually in upper case and are right-aligned. Figure 8 shows an example of the screenplay elements. A.5. Story Templates Over time various templates have been developed that help to create stories. One of the most famous templates is the 3-act structure (Field, 1979). This structure divides a story into a setup, confrontation, and resolution. In this work, we have used the 4-act structure which we now describe in detail. Act 1- This is the opening/introduction act. It describes the protagonist’s character and briefly introduces the movie’s theme. The act ends with the start of a new journey for the protagonist. Act 2A- Due to the vast span of Act 2, it can be divided into two acts. This act usually contains the start of a love story. It also entertains the audience as the protagonist tries to adapt to their new journey. The act ends as the movie’s midpoint, one of the film’s critical moments, with either a very positive or negative scene. Act 2B- This act usually contains the protagonist’s downfall. The villain or antagonist starts to gain an advantage, and the protagonist loses something or someone significant. The act ends with the protagonist realizing their new mission after reaching rock bottom. Act 3— The protagonist has realized the change required in them and sets out to defeat the antagonist in a thrilling finale. The movie then ends by displaying a welcome change in the protagonist that was lacking in the beginning. A.6. Fine-Tuning GPT-3 GPT-3 was deemed publicly available last year by OpenAI (Brown et al., 2020). Its best model has 175B parameters, which is much more than GPT2’s 2.9B parameters. We have fine-tuned multiple plot generation models with GPT-3 along with a scene generation model. The multiple combinations of plot generation models are short or long prompts and with or without genres. The GPT-3 model and hyperparameters remain the same for all the above combinations. We have fine-tuned the GPT-3 Curie model for four epochs. For generating text, GPT-3 offers various hyperparameters to tune and get closer to our desired results. For testing, we set other hyperparameters as follows: the temperature as 0.7, top-p as 1, frequency penalty as 0.1, presence penalty as 0.1, and max tokens as 900. This paper is available on arxiv under CC 4.0 DEED license. Authors: (1) Prerak Gandhi, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, prerakgandhi@cse.iitb.ac.in, and these authors contributed equally to this work; (2) Vishal Pramanik, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, vishalpramanik,pb@cse.iitb.ac.in, and these authors contributed equally to this work; (3) Pushpak Bhattacharyya, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai. Authors: Authors: (1) Prerak Gandhi, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, prerakgandhi@cse.iitb.ac.in, and these authors contributed equally to this work; (2) Vishal Pramanik, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, vishalpramanik,pb@cse.iitb.ac.in, and these authors contributed equally to this work; (3) Pushpak Bhattacharyya, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai. Table of Links Abstract and Intro Motivation Related Work Dataset Experiments and Evaluation Results and Analysis Conclusion and Future Work Limitations and References A. Appendix Abstract and Intro Abstract and Intro Motivation Motivation Related Work Related Work Dataset Dataset Experiments and Evaluation Experiments and Evaluation Results and Analysis Results and Analysis Conclusion and Future Work Conclusion and Future Work Limitations and References Limitations and References A. Appendix A. Appendix A. Appendix A.1. Ethics Consideration We have taken all the scripts from IMDB and IMSDb databases. The website has a disclaimer regarding using its scripts for research, which can be found at this link https://imsdb.com/ disclaimer.html. We have used the scripts fairly and without copyright violation. A.2. Annotator Profiles We required the help of external annotators in two cases: (i) Manually Annotating the Scripts and (ii) Creating scenes and their descriptions from the scripts. For the first task, we took the help of 10 annotators. Their ages ranged from 21-28, and all were Asian. They were given detailed guidelines with examples for annotating. There were also periodic sessions to confirm their understanding and solve their doubts and mistakes. For the second task, we took the help of two annotators. Both of them are Asian females aged between 21-23. Both of them were given detailed guidelines for the scene-writing task. A few data points were picked randomly and checked to find out and correct conceptual mistakes. The annotators had bachelors and masters degree in STEM and Arts. A.3. Evaluation Metrics The evaluation metrics are described below: • Perplexity (PPL): Perplexity is one of the most common metrics for evaluating language models. They are computed as exponential of entropy. The smaller the value of the PPL, the greater the fluency of the generated text. • Perplexity (PPL): • BLEU: B i L ingual E valuation U nderstudy is a common metric in many NLP tasks, especially in the field of Machine Translation. It measures the overlap between the generated output and gold standard data. Although this metric does not consider the model’s creativity, we can deduce the difference between the candidate text and the reference text using BLEU. The higher the BLEU measure, the better it is. • BLEU: B L E U • ROUGE: R ecall- O riented U nderstudy for G isting E valuation is typically used for evaluating automatic summarization. In our case, it measures the longest overlapping sequence between the generated and original plots. The higher the ROUGE measure, the better it is. • ROUGE: R O U G E • N-grams: We measure the redundancy and diversity of the movie plots by computing the repetition and distinction n-gram scores. • N-grams: A.4. Screenplay Structure A movie script or a screenplay has a different format than a story. A script is a group of scenes. Each of these scenes consists of a few major components, which are discussed below: Scene Headings/Sluglines- This component describes the when and where of the scene. It can be thought of as the first shot that a camera takes of a new scene. For example, INT. - RESTAURANT - NIGHT indicates that the scene starts inside a restaurant at night. Sluglines are normally written in capital letters and are left-aligned. Scene Headings/Sluglines- Character Names - they are mentioned every time a character is going to utter a dialogue. The name of each character is mentioned in uppercase and is centre aligned. Character Names Dialogues - dialogues are the lines that the characters say. They appear right after the character name in a script and are centrally aligned. Dialogues Action Lines - action lines describe almost everything about a scene. They can be described as the narration of each script. Action lines can be present after either dialogues or sluglines and are left-aligned. Action Lines Transitions - a transition marks the change from one scene to the next. They also depict how a scene is ended. For example, DISSOLVE, FADE, and CUT are different keywords used to indicate a transition. They are usually in upper case and are right-aligned. Transitions Figure 8 shows an example of the screenplay elements. A.5. Story Templates Over time various templates have been developed that help to create stories. One of the most famous templates is the 3-act structure (Field, 1979). This structure divides a story into a setup, confrontation, and resolution. In this work, we have used the 4-act structure which we now describe in detail. Act 1 - This is the opening/introduction act. It describes the protagonist’s character and briefly introduces the movie’s theme. The act ends with the start of a new journey for the protagonist. Act 1 Act 2A - Due to the vast span of Act 2, it can be divided into two acts. This act usually contains the start of a love story. It also entertains the audience as the protagonist tries to adapt to their new journey. The act ends as the movie’s midpoint, one of the film’s critical moments, with either a very positive or negative scene. Act 2A Act 2B - This act usually contains the protagonist’s downfall. The villain or antagonist starts to gain an advantage, and the protagonist loses something or someone significant. The act ends with the protagonist realizing their new mission after reaching rock bottom. Act 2B Act 3 — The protagonist has realized the change required in them and sets out to defeat the antagonist in a thrilling finale. The movie then ends by displaying a welcome change in the protagonist that was lacking in the beginning. Act 3 A.6. Fine-Tuning GPT-3 GPT-3 was deemed publicly available last year by OpenAI (Brown et al., 2020). Its best model has 175B parameters, which is much more than GPT2’s 2.9B parameters. We have fine-tuned multiple plot generation models with GPT-3 along with a scene generation model. The multiple combinations of plot generation models are short or long prompts and with or without genres. The GPT-3 model and hyperparameters remain the same for all the above combinations. We have fine-tuned the GPT-3 Curie model for four epochs. For generating text, GPT-3 offers various hyperparameters to tune and get closer to our desired results. For testing, we set other hyperparameters as follows: the temperature as 0.7, top-p as 1, frequency penalty as 0.1, presence penalty as 0.1, and max tokens as 900. This paper is available on arxiv under CC 4.0 DEED license. This paper is available on arxiv under CC 4.0 DEED license. available on arxiv

"Kurosawa": A Script Writer's Assistant: Appendix

"কুরোসাওয়া": একজন স্ক্রিপ্ট রাইটার সহকারী: পরিশিষ্ট

„Kurosawa“: Der Assistent eines Drehbuchautors: Anhang

"Kurosawa": asistente del guionista: Apéndice

"Kurosawa" : l'assistant d'un scénariste : annexe

"कुरोसावा": एक पटकथा लेखक का सहायक: परिशिष्ट

「黒澤」脚本家アシスタント 付録

"쿠로사와": 각본가의 어시스턴트: 부록

"Kurosawa": Assistente de Roteiro: Apêndice

«Куросава»: Помощник сценариста: Приложение

"Kurosawa": Bir Senaryo Yazarının Asistanı: Ek

"Kurosawa": Trợ lý biên kịch: Phụ lục

“黑泽明”：编剧助理：附录

Authors:
(1) Prerak Gandhi, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, prerakgandhi@cse.iitb.ac.in, and these authors contributed equally to this work;
(2) Vishal Pramanik, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, vishalpramanik,pb@cse.iitb.ac.in, and these authors contributed equally to this work;
(3) Pushpak Bhattacharyya, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai. Table of Links Abstract and Intro
Motivation
Related Work
Dataset
Experiments and Evaluation
Results and Analysis
Conclusion and Future Work
Limitations and References
A. Appendix 6. Results and Analysis We present our observations and evaluations. The nature of our task makes human evaluation take precedence over automatic evaluation (it is for automatic movie script generation, after all!). The qualitative analysis of our generated plots and scenes is based on feedback from 5 professional scriptwriters of our industry partner, the well-known media platform. 6.1. Plot Generation 6.1.1. Automatic Evaluation Table 1 shows auto-evaluation scores for the multiple GPT-3 plot generation models. 6.1.2. Human Rating We conducted human evaluation on Hollywood annotated short input model. The evaluation was done by five groups of 3 people, with each group having been assigned 10 unique plots. The ratings given for the 5 features are in Figure 5. The average scores for fluency, creativity, likability, coherence and relevance are 3.98, 3.29, 2.97, 2.65 and 2.55, respectively. Fluency of almost 4 is an indicator of the power of GPT-3 as a language model. Creativity and likeability are respectable at a value of around 3.0. The low BLEU scores support the average creativity score (Table 1). Figure 5 indicates that coherence and relevance still have major room for improvement. The MAUVE (Pillutla et al., 2021) value measures the gap between neural text and human text. We have separately calculated the MAUVE scores for 20 plots and 50 plots. The weighted average of the MAUVE scores for the two experiments is 0.48 which is reasonably good. 6.1.3. Qualitative Observations Professional scriptwriters from our industry partner have given the following observations: Non-annotated Hollywood Plots • The build-up is creative and interesting, but the ending becomes incoherent. • Some characters which are introduced in the beginning are never mentioned again. • The output is not portraying the key points or the theme mentioned in the input. Annotated Hollywood Plots • The plots are much more coherent, and the endings are logical. • There is still hallucination present (a common feature of all models). • The longer inputs made the plots more attentive to the key points. Annotated Hollywood Plots with Genres included • Along with the above points, now the plots generated are more tilted towards the genre or genres of the movie the writer wants to create. • Addition of genre gives some control over the kind of plot generated by the model. Annotated Bollywood plots • The outputs show incoherence in the last two paragraphs and repetition of the same characters throughout the plot. • The flow of the plot is not fast enough, i.e., the plot does not move ahead much. • Many of the outputs have a 1990s theme around them, where the characters are separated and then find each other later. This is due to a skewed dataset with lesser modern plots. 6.2. Scene Generation We fine-tuned GPT-3 for scene generation with our dataset. We generated ten scenes using the models mentioned in 5.1. Figure 7 in the appendix. shows an example of a completely generated scene. 6.2.1. Human Ratings We conducted a human evaluation on 10 scenes generated by the above model. 5 people evaluated the scenes using the Likert Scale. The ratings for the five features can be seen in Figure 5. The average scores for fluency, creativity, likability, coherence, and relevance are 4.48, 3.9, 3.48, 3.46 and 3.86, respectively. All of the values are above the neutral mark and imply that the generated scenes are close to human-written scenes. 6.2.2. Qualitative Observations In this section, we analyze the quality of the scenes generated by the GPT-3 model. This analysis has been done by professional scriptwriters from the previously mentioned media company. • The model produces a well-structured scene. • It can create new characters and fabricate dialogues even when they are unimportant. • The key points from the input can be found in the output. • There are some lines that are repetitive. • The output is not completely coherent. This paper is available on arxiv under CC 4.0 DEED license. Authors: (1) Prerak Gandhi, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, prerakgandhi@cse.iitb.ac.in, and these authors contributed equally to this work; (2) Vishal Pramanik, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, vishalpramanik,pb@cse.iitb.ac.in, and these authors contributed equally to this work; (3) Pushpak Bhattacharyya, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai. Authors: Authors: (1) Prerak Gandhi, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, prerakgandhi@cse.iitb.ac.in, and these authors contributed equally to this work; (2) Vishal Pramanik, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, vishalpramanik,pb@cse.iitb.ac.in, and these authors contributed equally to this work; (3) Pushpak Bhattacharyya, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai. Table of Links Abstract and Intro Motivation Related Work Dataset Experiments and Evaluation Results and Analysis Conclusion and Future Work Limitations and References A. Appendix Abstract and Intro Abstract and Intro Motivation Motivation Related Work Related Work Dataset Dataset Experiments and Evaluation Experiments and Evaluation Results and Analysis Results and Analysis Conclusion and Future Work Conclusion and Future Work Limitations and References Limitations and References A. Appendix A. Appendix 6. Results and Analysis We present our observations and evaluations. The nature of our task makes human evaluation take precedence over automatic evaluation (it is for automatic movie script generation, after all!). The qualitative analysis of our generated plots and scenes is based on feedback from 5 professional scriptwriters of our industry partner, the well-known media platform. 6.1. Plot Generation 6.1.1. Automatic Evaluation Table 1 shows auto-evaluation scores for the multiple GPT-3 plot generation models. 6.1.2. Human Rating We conducted human evaluation on Hollywood annotated short input model. The evaluation was done by five groups of 3 people, with each group having been assigned 10 unique plots. The ratings given for the 5 features are in Figure 5. The average scores for fluency, creativity, likability, coherence and relevance are 3.98, 3.29, 2.97, 2.65 and 2.55 , respectively. Fluency of almost 4 is an indicator of the power of GPT-3 as a language model. Creativity and likeability are respectable at a value of around 3.0. The low BLEU scores support the average creativity score (Table 1). Figure 5 indicates that coherence and relevance still have major room for improvement. 3.98, 3.29, 2.97, 2.65 2.55 The MAUVE (Pillutla et al., 2021) value measures the gap between neural text and human text. We have separately calculated the MAUVE scores for 20 plots and 50 plots. The weighted average of the MAUVE scores for the two experiments is 0.48 which is reasonably good. 0.48 6.1.3. Qualitative Observations Professional scriptwriters from our industry partner have given the following observations: Non-annotated Hollywood Plots Non-annotated Hollywood Plots • The build-up is creative and interesting, but the ending becomes incoherent. • Some characters which are introduced in the beginning are never mentioned again. • The output is not portraying the key points or the theme mentioned in the input. Annotated Hollywood Plots Annotated Hollywood Plots • The plots are much more coherent, and the endings are logical. • There is still hallucination present (a common feature of all models). • The longer inputs made the plots more attentive to the key points. Annotated Hollywood Plots with Genres included Annotated Hollywood Plots with Genres included • Along with the above points, now the plots generated are more tilted towards the genre or genres of the movie the writer wants to create. • Addition of genre gives some control over the kind of plot generated by the model. Annotated Bollywood plots Annotated Bollywood plots • The outputs show incoherence in the last two paragraphs and repetition of the same characters throughout the plot. • The flow of the plot is not fast enough, i.e., the plot does not move ahead much. • Many of the outputs have a 1990s theme around them, where the characters are separated and then find each other later. This is due to a skewed dataset with lesser modern plots. 6.2. Scene Generation We fine-tuned GPT-3 for scene generation with our dataset. We generated ten scenes using the models mentioned in 5.1. Figure 7 in the appendix. shows an example of a completely generated scene. 6.2.1. Human Ratings We conducted a human evaluation on 10 scenes generated by the above model. 5 people evaluated the scenes using the Likert Scale. The ratings for the five features can be seen in Figure 5. The average scores for fluency, creativity, likability, coherence, and relevance are 4.48, 3.9, 3.48, 3.46 and 3.86 , respectively. All of the values are above the neutral mark and imply that the generated scenes are close to human-written scenes. 4.48, 3.9, 3.48, 3.46 3.86 6.2.2. Qualitative Observations In this section, we analyze the quality of the scenes generated by the GPT-3 model. This analysis has been done by professional scriptwriters from the previously mentioned media company. • The model produces a well-structured scene. • It can create new characters and fabricate dialogues even when they are unimportant. • The key points from the input can be found in the output. • There are some lines that are repetitive. • The output is not completely coherent. This paper is available on arxiv under CC 4.0 DEED license. This paper is available on arxiv under CC 4.0 DEED license. available on arxiv

"Kurosawa": A Script Writer's Assistant: Results and Analysis

"কুরোসাওয়া": একজন স্ক্রিপ্ট লেখকের সহকারী: ফলাফল এবং বিশ্লেষণ

„Kurosawa“: Assistent eines Drehbuchautors: Ergebnisse und Analyse

"Kurosawa": un asistente de guionista: resultados y análisis

"Kurosawa" : un assistant de scénariste : résultats et analyse

"कुरोसावा": एक पटकथा लेखक का सहायक: परिणाम और विश्लेषण

「黒澤」脚本家アシスタント：結果と分析

"쿠로사와": 각본가의 조수: 결과 및 분석

"Kurosawa": um assistente de roteirista: resultados e análises

«Куросава»: помощник сценариста: итоги и анализ

"Kurosawa": Bir Senaryo Yazarının Asistanı: Sonuçlar ve Analiz

"Kurosawa": Trợ lý biên kịch: Kết quả và phân tích

“黑泽明”：编剧助理：结果与分析

Authors:
(1) Prerak Gandhi, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, prerakgandhi@cse.iitb.ac.in, and these authors contributed equally to this work;
(2) Vishal Pramanik, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, vishalpramanik,pb@cse.iitb.ac.in, and these authors contributed equally to this work;
(3) Pushpak Bhattacharyya, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai. Table of Links Abstract and Intro
Motivation
Related Work
Dataset
Experiments and Evaluation
Results and Analysis
Conclusion and Future Work
Limitations and References
A. Appendix 5. Experiments and Evaluation We fine-tune GPT3 with our datasets (refer appendix A.6). 5.1. Plot Generation We have created 5 models by fine-tuning GPT-3 with our movie plot dataset in the following manner, (i) original (without annotation) (O): input- short storylines, output- plots without any annotations, (ii) annotation and short input (AS): input- short storylines, output- plots annotated with 4-act structure, (iii) annotation and long input (AL): input long, more descriptive storylines, output- plots annotated with 4-act structure, (iv) annotation and short input with genres included (ASG): input short storylines and genre, output- plots annotated with 4-act structure, (v) annotation and long input with genres included (ALG): input- long and more descriptive storylines along with the genre, output- plots annotated with 4-act structure. For automatic evaluation we use BLEU (Papineni et al., 2002), Perplexity (Jelinek et al., 1977), ROUGE (Lin, 2004). We also use human evaluation in the form of a five-point Likert Scale (Likert, 1932). The rating system has 1-> Strongly Disagree, 2-> Disagree, 3-> Neutral, 4-> Agree, 5-> Strongly Agree. Human-written stories are assumed to have a rating of 5 for each of the following 5 features: (1) Fluency: grammatical correctness; (2) Coherence: logical ordering of sentences and paragraphs; (3) Relevance: Whether the key points from the prompt have been highlighted in the output; (4) Likability: The measure of how much the story is enjoyable; (5) Creativity: If the output introduced any new events, character profiles, or relationships. For plot generation, we generate 50 plots from 50 test prompts. We divide the stories into five groups of 10 and assign three evaluators to each group. For scene generation, we generate ten scenes from 10 test prompts. We assign five evaluators to rate these ten stories. This paper is available on arxiv under CC 4.0 DEED license. Authors: (1) Prerak Gandhi, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, prerakgandhi@cse.iitb.ac.in, and these authors contributed equally to this work; (2) Vishal Pramanik, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, vishalpramanik,pb@cse.iitb.ac.in, and these authors contributed equally to this work; (3) Pushpak Bhattacharyya, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai. Authors: Authors: (1) Prerak Gandhi, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, prerakgandhi@cse.iitb.ac.in, and these authors contributed equally to this work; (2) Vishal Pramanik, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai, vishalpramanik,pb@cse.iitb.ac.in, and these authors contributed equally to this work; (3) Pushpak Bhattacharyya, Department of Computer Science and Engineering, Indian Institute of Technology Bombay, Mumbai. Table of Links Abstract and Intro Motivation Related Work Dataset Experiments and Evaluation Results and Analysis Conclusion and Future Work Limitations and References A. Appendix Abstract and Intro Abstract and Intro Motivation Motivation Related Work Related Work Dataset Dataset Experiments and Evaluation Experiments and Evaluation Results and Analysis Results and Analysis Conclusion and Future Work Conclusion and Future Work Limitations and References Limitations and References A. Appendix A. Appendix 5. Experiments and Evaluation We fine-tune GPT3 with our datasets (refer appendix A.6 ). appendix A.6 5.1. Plot Generation We have created 5 models by fine-tuning GPT-3 with our movie plot dataset in the following manner, (i) original (without annotation) ( O ): input- short storylines, output- plots without any annotations, (ii) annotation and short input (AS) : input- short storylines, output- plots annotated with 4-act structure, (iii) annotation and long input ( AL ): input long, more descriptive storylines, output- plots annotated with 4-act structure, (iv) annotation and short input with genres included ( ASG ): input short storylines and genre, output- plots annotated with 4-act structure, (v) annotation and long input with genres included ( ALG ): input- long and more descriptive storylines along with the genre, output- plots annotated with 4-act structure. original O annotation and short input (AS) annotation and long input AL annotation and short input with genres included ASG annotation and long input with genres included ALG For automatic evaluation we use BLEU (Papineni et al., 2002), Perplexity (Jelinek et al., 1977), ROUGE (Lin, 2004). We also use human evaluation in the form of a five-point Likert Scale (Likert, 1932). The rating system has 1-> Strongly Disagree, 2-> Disagree, 3-> Neutral, 4-> Agree, 5-> Strongly Agree. Human-written stories are assumed to have a rating of 5 for each of the following 5 features: (1) Fluency : grammatical correctness; (2) Coherence : logical ordering of sentences and paragraphs; (3) Relevance : Whether the key points from the prompt have been highlighted in the output; (4) Likability : The measure of how much the story is enjoyable; (5) Creativity : If the output introduced any new events, character profiles, or relationships. BLEU Perplexity ROUGE Fluency Coherence Relevance Likability Creativity For plot generation, we generate 50 plots from 50 test prompts. We divide the stories into five groups of 10 and assign three evaluators to each group. For scene generation, we generate ten scenes from 10 test prompts. We assign five evaluators to rate these ten stories. This paper is available on arxiv under CC 4.0 DEED license. This paper is available on arxiv under CC 4.0 DEED license. available on arxiv

"Kurosawa": A Script Writer's Assistant: Experiments and Evaluation

"কুরোসাওয়া": একজন স্ক্রিপ্ট রাইটারের সহকারী: পরীক্ষা এবং মূল্যায়ন

„Kurosawa“: Der Assistent eines Drehbuchautors: Experimente und Evaluation

"Kurosawa": asistente de guionista: experimentos y evaluación

"Kurosawa" : un assistant de scénariste : expérimentations et évaluation

"कुरोसावा": एक पटकथा लेखक का सहायक: प्रयोग और मूल्यांकन

「黒澤」脚本家アシスタントの実験と評価

"구로사와": 각본가의 조수: 실험과 평가

"Kurosawa": um assistente de roteirista: experimentos e avaliação

«Куросава»: помощник сценариста: эксперименты и оценки

"Kurosawa": Bir Senaryo Yazarının Asistanı: Deneyler ve Değerlendirme

"Kurosawa": Trợ lý biên kịch: Thử nghiệm và đánh giá

“黑泽明”：编剧助理：实验与评估

I entered the programming world without a formal degree, relying on a Udemy course and some personal projects. Four months later, I landed an internship as an FE developer at BMW. For almost two years, I worked as a Front-End (FE) developer. Deep down, I’ve always wanted to dive into Back-End (BE) development, but my team’s leadership believed it wasn’t the right time yet. The beginning was tough. I had no real understanding of programming; I just knew I was passionate about it. My initial interest wasn’t in creating websites or dealing with the visual aspects users see. Instead, I was drawn to the underlying technology and processes. Starting with FE was a practical choice, as friends in the field advised me it was a more accessible entry point for beginners. In retrospect, they were right, at least for that time. Don’t get me wrong — FE development isn’t easier than BE. They each come with their own complexities. It took me a long time to get my bearings, to program with fluency, and to become self-sufficient. Once I had established myself in my FE role, I began to express my interest in BE more assertively. I started attending BE refinements and sought out study advice relevant to our project’s BE context. My initiative was welcomed, and within a couple of sprints, I had my first BE task. First Impressions Have you ever walked into a supermarket for the first time and felt completely lost, trying to take everything in but seeing nothing clearly? That’s exactly how I felt when I first saw the BE repository. I couldn’t make sense of how everything interacted or what communicated with what. I tried reading the documentation, but it only added to my confusion. It felt like I was back to day one of programming. This time, however, I had a general understanding of the business context, which made my frustration even greater. As an intern, I didn’t know what I was supposed to do or how to do it. This time, I knew what needed to be done but not how to do it, and I was no longer a novice. I had high expectations for myself. I had to step back, ask for help, and study over the weekends to wrap my head around things. Slowly but surely, with a lot of assistance, I began to understand how things worked (at least within the context of my story) and managed to deliver everything on time. The Complexity Difference It’s not that one role is more or less complex than the other, but my experience suggests that BE developers need to be more aware of all the relationships and interconnections within an application. While FE developers focus on how to present data, BE developers are concerned with ensuring all the necessary data is available, which requires a broader understanding of the business. Important Points to Consider If you’re thinking about making this career shift, here are a few things to keep in mind: Ask Questions: Lots of them. It’s normal to feel lost at first, but with the right guidance, you’ll find your way.


Embrace Discomfort: Don’t cling to your previous position, expecting nothing to change. You’ve stepped out of your comfort zone, so learn to be comfortable with discomfort. Remember, just because you were proficient in FE doesn’t mean you’ll start as proficient in BE. You might begin as a junior again, but your established background might help you become proficient more quickly than the first time.


Enjoy the Journey: Programming is all about challenges, learning, problem-solving, and creativity. Believe in yourself and enjoy the process. Embrace the journey with enthusiasm. Programming is about facing challenges, continuous learning, solving problems, and being creative. Believe in yourself, and good luck on your journey! I entered the programming world without a formal degree, relying on a Udemy course and some personal projects. Four months later, I landed an internship as an FE developer at BMW. For almost two years, I worked as a Front-End (FE) developer. Deep down, I’ve always wanted to dive into Back-End (BE) development, but my team’s leadership believed it wasn’t the right time yet. The beginning was tough. I had no real understanding of programming; I just knew I was passionate about it. My initial interest wasn’t in creating websites or dealing with the visual aspects users see. Instead, I was drawn to the underlying technology and processes. Starting with FE was a practical choice, as friends in the field advised me it was a more accessible entry point for beginners. In retrospect, they were right, at least for that time. Don’t get me wrong — FE development isn’t easier than BE. They each come with their own complexities. It took me a long time to get my bearings, to program with fluency, and to become self-sufficient. Once I had established myself in my FE role, I began to express my interest in BE more assertively. I started attending BE refinements and sought out study advice relevant to our project’s BE context. My initiative was welcomed, and within a couple of sprints, I had my first BE task. First Impressions Have you ever walked into a supermarket for the first time and felt completely lost, trying to take everything in but seeing nothing clearly? That’s exactly how I felt when I first saw the BE repository. I couldn’t make sense of how everything interacted or what communicated with what. I tried reading the documentation, but it only added to my confusion. It felt like I was back to day one of programming. This time, however, I had a general understanding of the business context, which made my frustration even greater. As an intern, I didn’t know what I was supposed to do or how to do it. This time, I knew what needed to be done but not how to do it, and I was no longer a novice. I had high expectations for myself. I had to step back, ask for help, and study over the weekends to wrap my head around things. Slowly but surely, with a lot of assistance, I began to understand how things worked (at least within the context of my story) and managed to deliver everything on time. The Complexity Difference It’s not that one role is more or less complex than the other, but my experience suggests that BE developers need to be more aware of all the relationships and interconnections within an application. While FE developers focus on how to present data, BE developers are concerned with ensuring all the necessary data is available, which requires a broader understanding of the business. Important Points to Consider If you’re thinking about making this career shift, here are a few things to keep in mind: Ask Questions: Lots of them. It’s normal to feel lost at first, but with the right guidance, you’ll find your way. Embrace Discomfort: Don’t cling to your previous position, expecting nothing to change. You’ve stepped out of your comfort zone, so learn to be comfortable with discomfort. Remember, just because you were proficient in FE doesn’t mean you’ll start as proficient in BE. You might begin as a junior again, but your established background might help you become proficient more quickly than the first time. Enjoy the Journey: Programming is all about challenges, learning, problem-solving, and creativity. Believe in yourself and enjoy the process. Ask Questions: Lots of them. It’s normal to feel lost at first, but with the right guidance, you’ll find your way. Ask Questions: Lots of them. It’s normal to feel lost at first, but with the right guidance, you’ll find your way. Embrace Discomfort: Don’t cling to your previous position, expecting nothing to change. You’ve stepped out of your comfort zone, so learn to be comfortable with discomfort. Remember, just because you were proficient in FE doesn’t mean you’ll start as proficient in BE. You might begin as a junior again, but your established background might help you become proficient more quickly than the first time. Embrace Discomfort: Don’t cling to your previous position, expecting nothing to change. You’ve stepped out of your comfort zone, so learn to be comfortable with discomfort. Remember, just because you were proficient in FE doesn’t mean you’ll start as proficient in BE. You might begin as a junior again, but your established background might help you become proficient more quickly than the first time. Enjoy the Journey: Programming is all about challenges, learning, problem-solving, and creativity. Believe in yourself and enjoy the process. Enjoy the Journey: Programming is all about challenges, learning, problem-solving, and creativity. Believe in yourself and enjoy the process. Embrace the journey with enthusiasm. Programming is about facing challenges, continuous learning, solving problems, and being creative. Believe in yourself, and good luck on your journey!

From Front-End to Back-End: A Personal Journey Through the Programming Field

Performance management season always brings a push for data-driven approaches, and it’s usually not lost on people managers the truly subjective nature of the process. As previously written, I think embracing this subjectivity can be more beneficial than attempting to mask a subjective process with vanity metrics. There still is an opportunity to improve flawed data-driven approaches by being data-informed when data supports good arguments rather than replaces them. The Pitfall of Vanity Metrics Vanity metrics are appealing because they provide quantifiable data that seems to offer clear insights. However, they often obscure the subjective nature of performance management. This reminds me of the ongoing discussion on the distinction between math and science. Math relies on deductive reasoning, starting with basic principles and logically deducing consequences. If you accept the initial premises, the conclusions are certain.
In contrast, science uses inductive reasoning, drawing general conclusions from observations.For example, if every swan observed is white, one might conclude all swans are white, but a single black swan disproves this theory (1). The Advantages of Being Data-Informed Being data-informed means using data as a valuable tool to support well-reasoned arguments, not as the sole determinant of decisions, e.g.: Comprehensive Understanding: Data provides insights, but a comprehensive understanding of the factors driving performance is essential. Are all factors driving your metric well understood? If not, data may be misleading (2). By being data-informed, you use data to enhance your understanding rather than replace it.


Supporting Good Arguments: Good arguments are often based on a mix of data, context, and expertise. By supporting arguments with data, you ensure that decisions are well-rounded and consider multiple perspectives. This holistic approach is more robust than relying solely on data, which might overlook important nuances (2).


Encouraging Experimentation: Can you run an experiment to test your assumptions? Controlled experiments are invaluable for establishing causality. When data is used to inform hypotheses and guide experiments, it becomes a powerful tool for validation and learning (2).


Enhancing Communication: Data can help communicate complex ideas clearly and persuasively. Good data storytelling combines narrative with explanatory visuals, making insights more accessible and actionable (3).


Maintaining Motivation: Metrics that align with an individual’s or team’s convictions can enhance motivation. By focusing on meaningful metrics that truly reflect performance and contribution, organizations can boost morale and intrinsic motivation (2). Challenges with Overemphasis on Data While data is invaluable, an overemphasis on data-driven decisions can lead to: Misinterpretation: Without proper context, data can be misinterpreted. It’s crucial to ensure that data is understood correctly and used appropriately (2).


Streetlight Effect: Organizations may focus on easily measurable improvements, neglecting more significant but harder-to-measure changes. This can lead to superficial enhancements rather than meaningful progress (2).


Suspension of Disbelief: Introducing metrics in areas where they don’t belong can lead to a culture of ignoring the limitations and meaninglessness of certain data points (2).


Weak Leadership: Relying too heavily on data can be a sign of ineffective leadership. Strong leaders should be able to use their judgment and observations to make decisions, using data to support rather than dictate those decisions (2). The Value of Good Arguments Good arguments, even without extensive data, can often lead to better decisions. Here’s why: Correlation vs. Causation: Basic statistical principles remind us that correlation does not imply causation. It’s possible to make great decisions with good arguments and minimal data, but bad arguments can easily lead to poor decisions, even with good data (3,2).


Cultural Impact: An overemphasis on data can harm organizational culture. Metrics can lead to superficial improvements at the expense of more nuanced, meaningful changes (2).


Holistic Decision Making: Good arguments consider a wider range of factors, including context, expertise, and judgment, which are often overlooked in data-driven approaches (2). Embracing a Balanced Approach Data has its place and can be a useful tool for supporting arguments. However, it should not be the sole basis for decision-making. Strong arguments, grounded in observation and theory, often provide a more reliable foundation for decisions. Resist the temptation to rely solely on data, and maintain a healthy skepticism towards metrics that promise easy answers. Being data-informed rather than data-driven means using data to enhance and support well-reasoned arguments. This balanced approach ensures that decisions are both informed by data and enriched by context and expertise. Performance management benefits from a balanced approach that leverages data to support strong arguments. By being data-informed, organizations can make better decisions that consider both quantitative and qualitative factors. Embrace the subjectivity of the process and use data as a tool to enhance, not replace good arguments. This approach will lead to more meaningful, well-rounded decisions that align with organizational goals and values. For further reading, check out my other writings: Embracing Followership and Technical Fluency in Modern Leadership
Rethinking Performance Management: Embracing Subjectivity Over Objectivity
Improving Engineering Culture by Building Products, Not Projects University of Houston, THE ENGINES OF OUR INGENUITY: Math Versus Science ↩︎
Richard Marmorstein: Be good-argument-driven, not data-driven ↩︎
Forbes: 8 Pitfalls In The Data-Driven Decision-Making (DDDM) Process ↩︎ Performance management season always brings a push for data-driven approaches, and it’s usually not lost on people managers the truly subjective nature of the process. As previously written, I think embracing this subjectivity can be more beneficial than attempting to mask a subjective process with vanity metrics. There still is an opportunity to improve flawed data-driven approaches by being data-informed when data supports good arguments rather than replaces them. I think embracing this subjectivity I think embracing this subjectivity The Pitfall of Vanity Metrics Vanity metrics are appealing because they provide quantifiable data that seems to offer clear insights. However, they often obscure the subjective nature of performance management. This reminds me of the ongoing discussion on the distinction between math and science. Math relies on deductive reasoning, starting with basic principles and logically deducing consequences. If you accept the initial premises, the conclusions are certain. In contrast, science uses inductive reasoning, drawing general conclusions from observations. For example, if every swan observed is white, one might conclude all swans are white, but a single black swan disproves this theory (1). Math relies on deductive reasoning, starting with basic principles and logically deducing consequences. If you accept the initial premises, the conclusions are certain. Math relies on deductive reasoning, starting with basic principles and logically deducing consequences. If you accept the initial premises, the conclusions are certain. In contrast, science uses inductive reasoning, drawing general conclusions from observations. In contrast, science uses inductive reasoning, drawing general conclusions from observations. For example, if every swan observed is white, one might conclude all swans are white, but a single black swan disproves this theory (1). For example, if every swan observed is white, one might conclude all swans are white, but a single black swan disproves this theory (1). The Advantages of Being Data-Informed Being data-informed means using data as a valuable tool to support well-reasoned arguments, not as the sole determinant of decisions, e.g.: Comprehensive Understanding: Data provides insights, but a comprehensive understanding of the factors driving performance is essential. Are all factors driving your metric well understood? If not, data may be misleading (2). By being data-informed, you use data to enhance your understanding rather than replace it. Supporting Good Arguments: Good arguments are often based on a mix of data, context, and expertise. By supporting arguments with data, you ensure that decisions are well-rounded and consider multiple perspectives. This holistic approach is more robust than relying solely on data, which might overlook important nuances (2). Encouraging Experimentation: Can you run an experiment to test your assumptions? Controlled experiments are invaluable for establishing causality. When data is used to inform hypotheses and guide experiments, it becomes a powerful tool for validation and learning (2). Enhancing Communication: Data can help communicate complex ideas clearly and persuasively. Good data storytelling combines narrative with explanatory visuals, making insights more accessible and actionable (3). Maintaining Motivation: Metrics that align with an individual’s or team’s convictions can enhance motivation. By focusing on meaningful metrics that truly reflect performance and contribution, organizations can boost morale and intrinsic motivation (2). Comprehensive Understanding: Data provides insights, but a comprehensive understanding of the factors driving performance is essential. Are all factors driving your metric well understood? If not, data may be misleading (2). By being data-informed, you use data to enhance your understanding rather than replace it. Comprehensive Understanding: Data provides insights, but a comprehensive understanding of the factors driving performance is essential. Are all factors driving your metric well understood? If not, data may be misleading (2). By being data-informed, you use data to enhance your understanding rather than replace it. Comprehensive Understanding: Supporting Good Arguments: Good arguments are often based on a mix of data, context, and expertise. By supporting arguments with data, you ensure that decisions are well-rounded and consider multiple perspectives. This holistic approach is more robust than relying solely on data, which might overlook important nuances (2). Supporting Good Arguments: Good arguments are often based on a mix of data, context, and expertise. By supporting arguments with data, you ensure that decisions are well-rounded and consider multiple perspectives. This holistic approach is more robust than relying solely on data, which might overlook important nuances (2). Supporting Good Arguments: Encouraging Experimentation: Can you run an experiment to test your assumptions? Controlled experiments are invaluable for establishing causality. When data is used to inform hypotheses and guide experiments, it becomes a powerful tool for validation and learning (2). Encouraging Experimentation: Can you run an experiment to test your assumptions? Controlled experiments are invaluable for establishing causality. When data is used to inform hypotheses and guide experiments, it becomes a powerful tool for validation and learning (2). Encouraging Experimentation: Enhancing Communication: Data can help communicate complex ideas clearly and persuasively. Good data storytelling combines narrative with explanatory visuals, making insights more accessible and actionable (3). Enhancing Communication: Data can help communicate complex ideas clearly and persuasively. Good data storytelling combines narrative with explanatory visuals, making insights more accessible and actionable (3). Enhancing Communication: Maintaining Motivation: Metrics that align with an individual’s or team’s convictions can enhance motivation. By focusing on meaningful metrics that truly reflect performance and contribution, organizations can boost morale and intrinsic motivation (2). Maintaining Motivation: Metrics that align with an individual’s or team’s convictions can enhance motivation. By focusing on meaningful metrics that truly reflect performance and contribution, organizations can boost morale and intrinsic motivation (2). Maintaining Motivation: Challenges with Overemphasis on Data While data is invaluable, an overemphasis on data-driven decisions can lead to: Misinterpretation: Without proper context, data can be misinterpreted. It’s crucial to ensure that data is understood correctly and used appropriately (2). Streetlight Effect: Organizations may focus on easily measurable improvements, neglecting more significant but harder-to-measure changes. This can lead to superficial enhancements rather than meaningful progress (2). Suspension of Disbelief: Introducing metrics in areas where they don’t belong can lead to a culture of ignoring the limitations and meaninglessness of certain data points (2). Weak Leadership: Relying too heavily on data can be a sign of ineffective leadership. Strong leaders should be able to use their judgment and observations to make decisions, using data to support rather than dictate those decisions (2). Misinterpretation: Without proper context, data can be misinterpreted. It’s crucial to ensure that data is understood correctly and used appropriately (2). Misinterpretation: Without proper context, data can be misinterpreted. It’s crucial to ensure that data is understood correctly and used appropriately (2). Misinterpretation: Streetlight Effect: Organizations may focus on easily measurable improvements, neglecting more significant but harder-to-measure changes. This can lead to superficial enhancements rather than meaningful progress (2). Streetlight Effect: Organizations may focus on easily measurable improvements, neglecting more significant but harder-to-measure changes. This can lead to superficial enhancements rather than meaningful progress (2). Streetlight Effect: Suspension of Disbelief: Introducing metrics in areas where they don’t belong can lead to a culture of ignoring the limitations and meaninglessness of certain data points (2). Suspension of Disbelief: Introducing metrics in areas where they don’t belong can lead to a culture of ignoring the limitations and meaninglessness of certain data points (2). Suspension of Disbelief: Weak Leadership: Relying too heavily on data can be a sign of ineffective leadership. Strong leaders should be able to use their judgment and observations to make decisions, using data to support rather than dictate those decisions (2). Weak Leadership: Relying too heavily on data can be a sign of ineffective leadership. Strong leaders should be able to use their judgment and observations to make decisions, using data to support rather than dictate those decisions (2). Weak Leadership: The Value of Good Arguments Good arguments, even without extensive data, can often lead to better decisions. Here’s why: Here’s why: Correlation vs. Causation: Basic statistical principles remind us that correlation does not imply causation. It’s possible to make great decisions with good arguments and minimal data, but bad arguments can easily lead to poor decisions, even with good data (3,2). Cultural Impact: An overemphasis on data can harm organizational culture. Metrics can lead to superficial improvements at the expense of more nuanced, meaningful changes (2). Holistic Decision Making: Good arguments consider a wider range of factors, including context, expertise, and judgment, which are often overlooked in data-driven approaches (2). Correlation vs. Causation: Basic statistical principles remind us that correlation does not imply causation. It’s possible to make great decisions with good arguments and minimal data, but bad arguments can easily lead to poor decisions, even with good data (3,2). Correlation vs. Causation: Basic statistical principles remind us that correlation does not imply causation. It’s possible to make great decisions with good arguments and minimal data, but bad arguments can easily lead to poor decisions, even with good data (3,2). Cultural Impact: An overemphasis on data can harm organizational culture. Metrics can lead to superficial improvements at the expense of more nuanced, meaningful changes (2). Cultural Impact: An overemphasis on data can harm organizational culture. Metrics can lead to superficial improvements at the expense of more nuanced, meaningful changes (2). Holistic Decision Making: Good arguments consider a wider range of factors, including context, expertise, and judgment, which are often overlooked in data-driven approaches (2). Holistic Decision Making: Good arguments consider a wider range of factors, including context, expertise, and judgment, which are often overlooked in data-driven approaches (2). Embracing a Balanced Approach Data has its place and can be a useful tool for supporting arguments. However, it should not be the sole basis for decision-making. Strong arguments, grounded in observation and theory, often provide a more reliable foundation for decisions. Resist the temptation to rely solely on data, and maintain a healthy skepticism towards metrics that promise easy answers. Being data-informed rather than data-driven means using data to enhance and support well-reasoned arguments. This balanced approach ensures that decisions are both informed by data and enriched by context and expertise. Performance management benefits from a balanced approach that leverages data to support strong arguments. By being data-informed, organizations can make better decisions that consider both quantitative and qualitative factors. Embrace the subjectivity of the process and use data as a tool to enhance, not replace good arguments. This approach will lead to more meaningful, well-rounded decisions that align with organizational goals and values. For further reading, check out my other writings: For further reading, check out my other writings: Embracing Followership and Technical Fluency in Modern Leadership Rethinking Performance Management: Embracing Subjectivity Over Objectivity Improving Engineering Culture by Building Products, Not Projects Embracing Followership and Technical Fluency in Modern Leadership Embracing Followership and Technical Fluency in Modern Leadership Embracing Followership and Technical Fluency in Modern Leadership Rethinking Performance Management: Embracing Subjectivity Over Objectivity Rethinking Performance Management: Embracing Subjectivity Over Objectivity Rethinking Performance Management: Embracing Subjectivity Over Objectivity Improving Engineering Culture by Building Products, Not Projects Improving Engineering Culture by Building Products, Not Projects Improving Engineering Culture by Building Products, Not Projects University of Houston, THE ENGINES OF OUR INGENUITY: Math Versus Science ↩︎ Richard Marmorstein: Be good-argument-driven, not data-driven ↩︎ Forbes: 8 Pitfalls In The Data-Driven Decision-Making (DDDM) Process ↩︎ University of Houston, THE ENGINES OF OUR INGENUITY: Math Versus Science ↩︎ Math Versus Science Math Versus Science Richard Marmorstein: Be good-argument-driven, not data-driven ↩︎ Be good-argument-driven, not data-driven Be good-argument-driven, not data-driven Forbes: 8 Pitfalls In The Data-Driven Decision-Making (DDDM) Process ↩︎ 8 Pitfalls In The Data-Driven Decision-Making (DDDM) Process 8 Pitfalls In The Data-Driven Decision-Making (DDDM) Process

How to Balance Data and GoodJudgment in Effective Performance Management

Authors:
(1) Jianzhu Yao, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology;
(2) Ziqi Liu, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology;
(3) Jian Guan, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology;
(4) Minlie Huang, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology. Table of Links Abstract and Intro Related Works DIALSTORY Dataset Proposed Tasks Methodology Experiments Discussion Future Work Conclusion Limitations and References 6 Experiments 6.1 Masked Dialogue Generation Implementation Details To conduct experiments on the masked dialogue generation task, we decide the hyper-parameters based on the performance of the validation set. We train Shao et al. (2021)’s BART model for 4.6 epochs with a 1e-4 learning rate for 1 day, and for our model, we train it for 5.6 epochs with a 1e-4 learning rate for 1 day. All baselines and our model are trained using the Adam optimizer. During the training process for our method, we computed the selection coverage of characters within a single story. And it showed that in every 1000 training steps, the coverage of different characters ranged from 98.64% to 99.00%, which meant nearly all the characters are selected during training, and all the characters contributed to the generated dialogue. It further proved that the argmax in Eq. 3 operation doesn’t break the gradient progress when training for this task. Automatic Evaluation Following previous works, we use several standard, widely used automatic evaluation metrics. We use BLEUn (Papineni et al., 2002) to measure the average word overlap between each generated and groundtruth dialogue turn (n=1,2), and Distinct-n (Li et al., 2015) to evaluate n-gram diversity of generated dialogue turns (n=2,3,4). To be more specific, for the coherence classifier, we construct the training and validation sets by randomly shuffling the order of dialogue turns and keeping other content in the correct order. We regard the perturbed story as a negative example and the original story as a positive example. We sample another 195k stories (except those in DIALSTORY) from the novels of Guan et al. (2022) to construct the training set (190k examples) and the validation set (5k examples). We train the model for 4 epochs with a 2e-5 learning rate and a 16- batch size, using the Adam optimizer. During the evaluation, we consider an example coherent when the probability of being coherent predicted by the classifier is greater than 0.5. We use the ratio of outputs (along with the input) that are classified as coherent by the classifier to all generated outputs as the coherence score. The result of the automatic evaluation is presented in Table 3. According to the table, compared to the BART baseline, our model consistently generates more word overlaps with ground truth and achieves better diversity under the guidance of character representations, which means our model can generate more diverse but not commonplace responses. Figure 3 plots the coherence score varying with the number of masked dialogue turns. The result shows that our model gets a higher coherence score than BART when required to generate more than seven turns of dialogue in one story. Manual Evaluation We conduct a pairwise comparison between our model and the BART baseline. We randomly select 100 examples from the test set. For each pair of outputs along with the input, we ask three annotators to give a preference (win, lose and tie) in terms of fluency, coherence, and informativeness. All the annotations are native Chinese speakers. We adopt majority voting to make final decisions among the annotators. The three aspects of manual evaluation are as follows: Fluency: Grammatical correctness and intra-sentence linguistic quality.


Coherence: Inter-sentence relatedness, causal and temporal dependencies. We judge the coherence between the story and a dialogue turn by following the criterion in Table 5. We add the scores of all the generated dialogue turns in a story to get the overall coherence score of the story, which is then used to compare with each other.


Informativeness: Interesting, diverse and rich details. As shown in Table 4, all the results show moderate (κ > 0.4) agreement, which shows our model outperforms the BART baseline significantly in dialogue informativeness and coherence. Case Study Figure 5 showed two examples to investigate how learning character representations can help our model generate more coherent dialogue. We found that our model can better model the relationship between different characters and the direction of the storyline. For example, in the first case, we can see that the BART’s generation confuses different characters’ fathers, while our model captures the relationship between different characters, and generates proper responses for the corresponding characters, which also moves the plot forward. And in the second case, we can see that BART’s generation is commonplace and contradicts the plot development. In contrast, our model captures the intentions of the speaker and the development trend of the plot, generating an appropriate and coherent response. Since these two models use the same pretrained weight, we can infer that the character modeling module leverages the coherent and reasonable generation. We also summarize four error types of the generated dialogue turn for the DialGen task: (1) Intersentence Contradiction; (2) Inter-sentence Repetition; (3) Intra-sentence Contradiction; (4) Intrasentence Repetition. We show the typical corresponding cases in Figure 4. We conducted a quantitative analysis of those 4 error types on our model’s generation. We analyzed 20 stories with 103 dialog turns and the results are shown in Figure 6. We found that both our model and BART suffer from these errors, suggesting that there is still space for model improvement, especially in the inter-sentence repetition. 6.2 Dialogue Speaker Recognition Implementation Details To conduct experiments on the speaker recognition task, we decide the hyper-parameters based on the performance of the validation set. For the BART baseline and our approach, we insert a mask token before each dialogue needed to be predicted, and a person id token before and after each character name span. Then, we insert all the unique person id tokens before the input stories as different options, and make predictions based on the cos similarity of option tokens and mask tokens. We train Shao et al. (2021)’s BART model for 30 epochs with a 5e-5 learning rate for 3 days, For encoder-only baselines, we implemented BERT, RoBERTa, and MacBERT and trained them for 15 epochs wity a le-5 learning rate for 2 days. For our model, we train it for 22 epochs with a le-6 learning rate for 2 days. All baselines and our model are trained using the Adam optimizer. Metrics We evaluate the DialSpk task using two automatic metrics including dialogue-level accuracy (DAC) and story-level accuracy (SAC). DAC is calculated as the ratio of the correct predictions to the total number of specifies dialogue turns, while SAC is the ratio of the number of stories where all dialogue turns are correctly predicted to the number of all test examples. These two metrics provide the evaluation for dialogue understanding with different granularities. Results As shown in Table 7, our model outperforms all the baselines significantly (p< 0.01, Wilcoxon signed-rank test) on both DAS and SAC scores, suggesting the benefit of learning character representations. We tested the accuracy of automatic training set annotations, and the DAC/SAC scores are 86.78%/67.80%. Together with the model’s performance on the test set, we can see the automatic annotation for the training set is of good quality. We also conducted the human prediction experiment, and the DAC/SAC scores are 97.90%/90.70%, which are much higher than the best model. So there is much room for further improvement for machine-based approaches. This paper is available on arxiv under CC 4.0 DEED license. [2] https://huggingface.co/fnlp/ bart-base-chinese [3] https://huggingface.co/bert-base-chinese [4] https://huggingface.co/hfl/ chinese-roberta-wwm-ext [5] https://huggingface.co/hfl/ chinese-macbert-base Authors: (1) Jianzhu Yao, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology; (2) Ziqi Liu, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology; (3) Jian Guan, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology; (4) Minlie Huang, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology. Authors: Authors: (1) Jianzhu Yao, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology; (2) Ziqi Liu, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology; (3) Jian Guan, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology; (4) Minlie Huang, The CoAI group, Tsinghua University, Beijing, China Department of Computer Science and Technology, Tsinghua University, Beijing, China Beijing National Research Center for Information Science and Technology. Table of Links Abstract and Intro Abstract and Intro Related Works Related Works DIALSTORY Dataset DIALSTORY Dataset Proposed Tasks Proposed Tasks Methodology Methodology Experiments Experiments Discussion Discussion Future Work Future Work Conclusion Conclusion Limitations and References Limitations and References 6 Experiments 6.1 Masked Dialogue Generation Implementation Details To conduct experiments on the masked dialogue generation task, we decide the hyper-parameters based on the performance of the validation set. We train Shao et al. (2021)’s BART model for 4.6 epochs with a 1e-4 learning rate for 1 day, and for our model, we train it for 5.6 epochs with a 1e-4 learning rate for 1 day. All baselines and our model are trained using the Adam optimizer. Implementation Details During the training process for our method, we computed the selection coverage of characters within a single story. And it showed that in every 1000 training steps, the coverage of different characters ranged from 98.64% to 99.00%, which meant nearly all the characters are selected during training, and all the characters contributed to the generated dialogue. It further proved that the argmax in Eq. 3 operation doesn’t break the gradient progress when training for this task. Automatic Evaluation Following previous works, we use several standard, widely used automatic evaluation metrics. We use BLEUn (Papineni et al., 2002) to measure the average word overlap between each generated and groundtruth dialogue turn (n=1,2), and Distinct-n (Li et al., 2015) to evaluate n-gram diversity of generated dialogue turns (n=2,3,4). Automatic Evaluation To be more specific, for the coherence classifier, we construct the training and validation sets by randomly shuffling the order of dialogue turns and keeping other content in the correct order. We regard the perturbed story as a negative example and the original story as a positive example. We sample another 195k stories (except those in DIALSTORY) from the novels of Guan et al. (2022) to construct the training set (190k examples) and the validation set (5k examples). We train the model for 4 epochs with a 2e-5 learning rate and a 16- batch size, using the Adam optimizer. During the evaluation, we consider an example coherent when the probability of being coherent predicted by the classifier is greater than 0.5. We use the ratio of outputs (along with the input) that are classified as coherent by the classifier to all generated outputs as the coherence score. The result of the automatic evaluation is presented in Table 3. According to the table, compared to the BART baseline, our model consistently generates more word overlaps with ground truth and achieves better diversity under the guidance of character representations, which means our model can generate more diverse but not commonplace responses. Figure 3 plots the coherence score varying with the number of masked dialogue turns. The result shows that our model gets a higher coherence score than BART when required to generate more than seven turns of dialogue in one story. Manual Evaluation We conduct a pairwise comparison between our model and the BART baseline. We randomly select 100 examples from the test set. For each pair of outputs along with the input, we ask three annotators to give a preference (win, lose and tie) in terms of fluency, coherence, and informativeness. All the annotations are native Chinese speakers. We adopt majority voting to make final decisions among the annotators. The three aspects of manual evaluation are as follows: Manual Evaluation Fluency: Grammatical correctness and intra-sentence linguistic quality. Coherence: Inter-sentence relatedness, causal and temporal dependencies. We judge the coherence between the story and a dialogue turn by following the criterion in Table 5. We add the scores of all the generated dialogue turns in a story to get the overall coherence score of the story, which is then used to compare with each other. Informativeness: Interesting, diverse and rich details. Fluency: Grammatical correctness and intra-sentence linguistic quality. Fluency : Grammatical correctness and intra-sentence linguistic quality. Fluency Coherence: Inter-sentence relatedness, causal and temporal dependencies. We judge the coherence between the story and a dialogue turn by following the criterion in Table 5. We add the scores of all the generated dialogue turns in a story to get the overall coherence score of the story, which is then used to compare with each other. Coherence : Inter-sentence relatedness, causal and temporal dependencies. We judge the coherence between the story and a dialogue turn by following the criterion in Table 5. We add the scores of all the generated dialogue turns in a story to get the overall coherence score of the story, which is then used to compare with each other. Coherence Informativeness: Interesting, diverse and rich details. Informativeness : Interesting, diverse and rich details. Informativeness As shown in Table 4, all the results show moderate (κ > 0.4) agreement, which shows our model outperforms the BART baseline significantly in dialogue informativeness and coherence. Case Study Figure 5 showed two examples to investigate how learning character representations can help our model generate more coherent dialogue. We found that our model can better model the relationship between different characters and the direction of the storyline. For example, in the first case, we can see that the BART’s generation confuses different characters’ fathers, while our model captures the relationship between different characters, and generates proper responses for the corresponding characters, which also moves the plot forward. And in the second case, we can see that BART’s generation is commonplace and contradicts the plot development. In contrast, our model captures the intentions of the speaker and the development trend of the plot, generating an appropriate and coherent response. Since these two models use the same pretrained weight, we can infer that the character modeling module leverages the coherent and reasonable generation. Case Study We also summarize four error types of the generated dialogue turn for the DialGen task: (1) Intersentence Contradiction; (2) Inter-sentence Repetition; (3) Intra-sentence Contradiction; (4) Intrasentence Repetition. We show the typical corresponding cases in Figure 4. We conducted a quantitative analysis of those 4 error types on our model’s generation. We analyzed 20 stories with 103 dialog turns and the results are shown in Figure 6. We found that both our model and BART suffer from these errors, suggesting that there is still space for model improvement, especially in the inter-sentence repetition. 6.2 Dialogue Speaker Recognition Implementation Details To conduct experiments on the speaker recognition task, we decide the hyper-parameters based on the performance of the validation set. For the BART baseline and our approach, we insert a mask token before each dialogue needed to be predicted, and a person id token before and after each character name span. Then, we insert all the unique person id tokens before the input stories as different options, and make predictions based on the cos similarity of option tokens and mask tokens. We train Shao et al. (2021)’s BART model for 30 epochs with a 5e-5 learning rate for 3 days, For encoder-only baselines, we implemented BERT, RoBERTa, and MacBERT and trained them for 15 epochs wity a le-5 learning rate for 2 days. For our model, we train it for 22 epochs with a le-6 learning rate for 2 days. All baselines and our model are trained using the Adam optimizer. Metrics We evaluate the DialSpk task using two automatic metrics including dialogue-level accuracy (DAC) and story-level accuracy (SAC). DAC is calculated as the ratio of the correct predictions to the total number of specifies dialogue turns, while SAC is the ratio of the number of stories where all dialogue turns are correctly predicted to the number of all test examples. These two metrics provide the evaluation for dialogue understanding with different granularities. Metrics Results As shown in Table 7, our model outperforms all the baselines significantly (p< 0.01, Wilcoxon signed-rank test) on both DAS and SAC scores, suggesting the benefit of learning character representations. We tested the accuracy of automatic training set annotations, and the DAC/SAC scores are 86.78%/67.80%. Together with the model’s performance on the test set, we can see the automatic annotation for the training set is of good quality. We also conducted the human prediction experiment, and the DAC/SAC scores are 97.90%/90.70%, which are much higher than the best model. So there is much room for further improvement for machine-based approaches. Results This paper is available on arxiv under CC 4.0 DEED license. This paper is available on arxiv under CC 4.0 DEED license. available on arxiv [2] https://huggingface.co/fnlp/ bart-base-chinese [3] https://huggingface.co/bert-base-chinese [4] https://huggingface.co/hfl/ chinese-roberta-wwm-ext [5] https://huggingface.co/hfl/ chinese-macbert-base

Understanding and Generating Dialogue between Characters in Stories: Experiments

From Chat-Bots to Killer-Bots?

The launch of generative artificial intelligence (genAI) tools such as ChatGPT and DALL-E, making businesses explore how this technology can spark creativity and drive innovation. In fact, a 2021 survey by found that 86% of executives believe AI will become a mainstream technology in their company within the next 3-5 years. PwC But what exactly is Generative AI, and how can it help in the creative field? Generative AI isn't just about automating tasks, it inspires creativity and has practical applications in fields like drug discovery, virtual reality, and content creation. These AI systems can use deep learning algorithms to create entirely new content, from images and videos to music and text, that closely resembles human work. In this post, we'll explore how Generative AI is changing creativity and problem-solving in the business world. Generative AI as a creative assistant While automation is a one of the aspects of Generative AI, it is also used as a creative assistant; let's explore how it is changing creative processes across various fields. Brainstorm new product ideas Brainstorming sessions often need help to overcome creative blocks and create diverse ideas. Companies can use Generative AI tools like MarcomAI. By using vast datasets, these tools can create surreal product ideas based on keywords provided by the user. For example, PepsiCo uses Generative AI to create new flavor combinations. The AI analyzes consumer preferences, flavor profiles, and market trends to suggest novel and potentially delicious options humans might not have considered. Product and graphic design Generative AI is also changing the way designers explore concepts. Tools like AutoDesk Generative Design allow designers to input parameters and receive multiple design variations based on those constraints. It helps them to explore possibilities they might not have considered before for more innovative and efficient product development. Generates personalized marketing copy For a successful marketing campaign, writing compelling copy that resonates with different audiences can be a time-consuming task. Generative AI tools like Copy.ai analyze target audiences and product details to craft personalized marketing copy for different platforms. It allows marketers to tailor their messages for maximum impact without sacrificing creativity. Similarly, creating catchy headlines is also important for creating content assets. AI tools like Jasper can analyze your content and target audience to suggest attention-grabbing headlines.Thus, marketers can focus on refining the overall strategy while the AI does the creative heavy lifting. Compose music Generative AI is also making waves in the music industry. Amper Music can analyze a musician's style and generate melodic snippets to jumpstart the creative process. These AI-generated parts inspire musicians to create unique masterpieces. Moreover, platforms like LANDR analyze and build upon existing music to create new song sections or even full instrumental accompaniments. It allows musicians to explore new sonic parts and make music that might not have been possible alone. Popular GenAI platforms Generative AI tools like GPT-4, DALL-E 3, Sora, Apple's Realm, Google Gemini, and Anthropic's Claude are changing the creative landscape. Let's see how these platforms are being applied across various industries. by OpenAI, is used for many applications, from generating creative writing prompts to code development. For example, Duolingo, the popular language learning app, has integrated GPT-4 to create more personalized user learning experiences. GPT-4 understands significantly more nuance and detail, allowing you to translate your ideas into exceptionally accurate images easily. DALL·E 3 is built natively on ChatGPT, which lets you use ChatGPT as a brainstorming partner and refiner of your prompts. Just ask ChatGPT what you want to see in anything from a simple sentence to a detailed paragraph. DALL·E 3 Developed by OpenAI, Sora focuses on creating high-quality videos. Unlike prior models, which are limited by video size and resolution, Sora can generate impressive visuals in various formats. Google's Gemini excels at responding to your questions and requests in a comprehensive way, potentially serving as an effective research and writing assistant. Given Google's focus on global accessibility, Gemini can provide information and communication across multiple languages. Anthropic Claude exhibits near-human levels of fluency in complex tasks. It has increased capabilities in analysis and forecasting, nuanced content creation, code generation, and conversing in non-English languages like Spanish, Japanese, and French. These examples show how GenAI is becoming an invaluable assistant in the creative process. As AI continues to evolve, it's important to remember that it is not meant to replace human creativity but to augment it. By taking care of repetitive tasks, AI frees humans to focus on higher-level strategic thinking and emotional connection in the creative process. Building a Responsible AI Future: Challenges and Opportunities As we explore the incredible applications of Generative AI, it's also vital to address its challenges. Data privacy, bias, and the need for skilled professionals are key issues that require our attention. Data privacy Generative AI models are trained on vast amounts of data. To protect personal information, we need robust data governance frameworks and strict adherence to privacy regulations like GDPR and CCPA. The Partnership on AI (PAI) is a collaborative initiative that brings together diverse voices from the AI community, including leading technology companies, academic institutions, and civil society organizations. Their mission is to address important questions about our future with AI and to advance responsible AI practices. In January 2024, released a comprehensive report titled "AI Rules of the Road". This report provides a holistic approach to safety that includes bias, overreliance on AI systems, privacy, and worker treatment. It aligns with their ongoing efforts to establish best practices for responsible data collection and usage in AI development. PAI Bias in AI systems If the data used to train Generative AI models is biased, the outputs generated by these models can perpetuate and even amplify those biases. Ensuring diversity and inclusivity in the data sets used for training AI models is essential to mitigate this risk. Organizations like OpenAI have been working on techniques like "constrained language models" to reduce bias in generated content. Additionally, regularly auditing AI systems for potential biases and making necessary adjustments is necessary to build trust. The shortage of skilled professionals As the demand for AI talent continues to grow, there is a widening gap between the supply and demand of skilled professionals. To bridge this gap, we need to invest in education and training programs that equip individuals with the necessary skills to work with Generative AI technologies. Initiatives like the AI4ALL program, which focuses on increasing diversity and inclusion in AI education, are making great strides in this direction. Focusing on ethical AI development practices is another aspect of building a responsible AI future. It involves integrating ethical considerations into each stage of AI development, from data collection and model training to deployment and monitoring. The IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems has developed a comprehensive framework called "Ethically Aligned Design" to guide the responsible development of AI systems. To implement these principles, organizations can establish internal ethics committees to oversee the development and deployment of AI systems. These committees can ensure that AI projects align with the organization's values and adhere to ethical guidelines. Regular audits and impact assessments can also identify and mitigate potential risks associated with AI systems. Wrapping up Generative AI is pushing human creativity and innovation like never before. We can accomplish many tasks in less time and create imaginative products with these tools and technology. However, we must also address AI's challenges. We can ensure AI systems are developed and deployed responsibly by prioritizing data privacy, mitigating bias, and nurturing a skilled workforce. Our best chance of building AI systems that support our values and benefit humanity is through collaboration and transparency. By working with researchers, developers, and policymakers, we can create a future where Generative AI is used for good. So, let's experiment with Generative AI tools, learn about the latest advancements, and contribute a unique perspective to AI adoption.

Generative AI: The Future of Creativity and Problem-Solving is Here

100 Complex LLM Terminology Explained in One Single & One Simple Sentence

Today, we're diving deep into the world of anchoring, a cognitive bias that influences our choices far more pervasively than we realize. If you’ve heard of Anchors before, you’ve probably heard it in a sales context. I’ll touch on that, but it goes way deeper than this. Understanding Anchoring isn't just about dodging sneaky sales tactics; it's about understanding how our brains construct reality from imperfect information. Anchors Aweigh: What Is This Thing, Anyway? Picture this: You're browsing for a new jacket. The first one you see is beautiful, but it's a whopping $400. Way out of your budget. However, as you wander through the store, every other jacket now seems like a relative bargain. Suddenly, that $150 option looks downright appealing. That initial $400 price tag served as an anchor. It distorted your perception of what constitutes a "reasonable" price for a jacket. This mental glitch, my friends, is the anchoring effect in action. Another (non sales, example), was shown that if people are first asked whether Gandhi died before or after age 9, and are then asked to guess Gandhi's actual age at death, their guesses will be lower than if they had first been asked whether Gandhi lived past age 200. The first anchor of 9 or 200 pulls their final estimate in that direction, despite it having no logical relevance. The anchoring bias describes our mind's tendency to become overly influenced by the first piece of information (the anchor) it receives on a topic. Once that anchor's dropped, subsequent judgments get made relative to it, even if the initial anchor is totally arbitrary or irrelevant. Anchoring in the Wild Anchoring isn't just about shopping sprees and impulse buys. It permeates all sorts of life decisions: The person who throws out a number first (salary, sale price, you name it) often has an advantage because it sets the psychological stage for the rest of the negotiation. Negotiations: Get too fixated on the price you originally paid for a stock? You might miss signs it's time to sell, even if the fundamentals have changed (that's sunk cost fallacy, a close cousin of anchoring). Investing: From job interviews to dating, that initial impression (positive or negative) can color how people perceive everything else you do. First Impressions: Anchors: The Cognitive Comfort Blanket Think of an anchor as a mental shortcut. Our brains crave stability and order – they don't enjoy floating around in a sea of uncertainty. Anchors, even irrelevant ones, give us a point of reference, a starting line from which to judge everything else. The tricky part? Once an anchor is planted in our minds, it has a cascading effect: We start noticing details that confirm the established anchor and downplay information that contradicts it. This is a form of confirmation bias. Selective Attention: An anchor can make something seem outrageously expensive or like an incredible deal, even if the objective value remains unchanged. Our point of reference becomes skewed. Distorted Valuation: Anchors don't just affect isolated decisions. They can have a long-term influence on our spending habits, expectations, and even our beliefs about the world. The Ripple Effect: Why Are Our Brains So Easily Swayed? Researchers have a few theories: Our brains are cognitive misers—they love shortcuts. Instead of thoroughly evaluating every decision from scratch, referencing an anchor is a quick, albeit imperfect, decision-making hack. Effort Reduction: When clueless about a fair price or course of action, any initial data point seems better than none. The anchor gives us some place to start. Uncertainty: The World is Your Anchor Playground Let's be real, anchors are everywhere: That initial "Free" frames future pricing decisions. Suddenly the monthly subscription feels way less appealing, even if the total cost is reasonable. Companies know this. The "Free Trial" Trap: Have you noticed how tip options on those card readers keep getting higher? They're anchoring your perception of what constitutes a "normal" tip amount. The Power of Suggestion: The first opinion you hear on a controversial topic, or even just the initial word choice ("riot" vs. "protest"), can shape how you process subsequent information. Anchoring in the Social Realm: The Achilles Heel of Anchoring The most insidious thing about anchors is that they work even when we KNOW they exist. Studies show that even when people are explicitly warned about anchoring bias, it still influences their estimation and judgment. Why does this happen? Here's where things get interesting: System 1 thinking (fast, intuitive) is easily swayed by anchors. But while System 2 (slow, deliberate reasoning) can recognize the bias, it takes significant effort to fully counteract its effects. Anchors Tap Into Deeper Thinking: Our brains are wired to assess things comparatively. Even if you intellectually recognize that an anchor is arbitrary, your emotional response is primed to the comparison. Relative, Not Absolute: Anchors create a sense of cognitive fluency. Options that feel familiar or easy to process often seem intuitively correct, even if they aren't the most rational choice. The Fluency Factor: Fight Back Against the Invisible Hand So, are we doomed to be cognitive puppets manipulated by random anchors? Not quite. Here's your toolkit: Impulse buys are the best friend of anchoring. Give yourself time. Before reacting to a price, offer, or information, impose a self-mandated waiting period. That pause gives your logical brain space to catch up. The Art of Delayed Gratification: Never rely on a single data point. Actively seek out alternative pricing, multiple opinions, or different perspectives on an issue. This helps break the spell of the initial anchor. Seek Counter-Anchors: Especially in negotiations, sometimes it pays to be the one who avoids stating the first number. Strategic ambiguity shifts the dynamic in your favor. Embrace Strategic Ambiguity: Beyond the Obvious – Anchoring in Unexpected Places Anchoring isn't just a salesperson's trick. It seeps into areas where we least expect it, subtly shaping our choices, opinions, and even our self-perception: Ever catch yourself longing for "the good old days" when things seemed cheaper, simpler, or better? Those nostalgic memories can become anchors, making us resistant to change or less receptive to the advantages of the present. Anchoring to the Past: Our initial impressions of ourselves (in childhood through early adulthood) have incredible staying power. That old "I'm bad at math" label can linger for decades, leading to missed opportunities because we're anchored to a past limitation. The Self-Perception Anchor: Studies suggest that even something as trivial as the severity of a sentencing demand by the prosecution can unconsciously anchor the judge's ultimate sentence, regardless of the actual evidence. Anchoring in the Courtroom: Anchors: A Double-Edged Sword It's important to remember, anchoring isn't inherently good or bad. Our brains evolved this shortcut for a reason. Here's where it can be helpful: Ever heard the advice to "visualize success"? It's partly an anchoring effect. Setting a high initial goal (sales revenue, fitness target, creative output) creates a reference point that can motivate us, even if we don't fully reach the first ambitious mark. Goal Setting as Anchoring: Consciously reframing negative thought patterns with positive affirmations can be a form of beneficial self-anchoring. Over time, it helps shift your internal reference point. Positive Self-Talk: Charities that display suggested donation amounts or non-profits highlighting the impact of even small contributions are leveraging anchoring principles to nudge people towards greater generosity. Anchors for Social Good: The Ultimate Goal: Cognitive Flexibility It's not about eliminating anchors – they're an inevitable part of how our brains work. The key is becoming mindful of when they're operating, knowing when to deliberately set our own anchors, and developing the flexibility to shift our reference points when necessary. Understanding anchoring is a step towards greater objectivity, better decision-making, and a healthy skepticism towards the hidden forces trying to shape our reality. Let's get meta - could this newsletter be one big anchor to get you thinking more critically about decision-making? Absolutely 😉 Scott Also appears . here

Unveiling the Anchors: How Your Brain Makes Sense of the World (And Gets It Wrong)

অ্যাঙ্করগুলি উন্মোচন করা: কীভাবে আপনার মস্তিষ্ক বিশ্বের অনুভূতি তৈরি করে (এবং এটি ভুল হয়ে যায়)

Die Anker enthüllen: Wie Ihr Gehirn die Welt versteht (und es falsch versteht)

Revelando las anclas: cómo su cerebro le da sentido al mundo (y lo entiende mal)

Dévoiler les ancres : comment votre cerveau donne un sens au monde (et se trompe)

एंकरों का अनावरण: आपका मस्तिष्क दुनिया को कैसे समझता है (और इसे गलत बनाता है)

アンカーを明らかにする: あなたの脳は世界をどのように理解するのか (そして間違ってしまうのか)

앵커 공개: 뇌가 세상을 어떻게 이해하고 잘못 이해하는지

Revelando as âncoras: como seu cérebro dá sentido ao mundo (e erra)

Раскрытие якорей: как ваш мозг воспринимает мир (и делает это неправильно)

Çapaların Açıklanması: Beyniniz Dünyayı Nasıl Anlıyor (Ve Yanlış Anlıyor)

Tiết lộ những điểm neo: Bộ não của bạn nhận thức thế giới như thế nào (và hiểu sai)

揭开锚点：你的大脑如何理解世界（以及理解错误）

We are building an exciting startup that aims to revolutionize mobile app development with our new AI-powered platform. Usually, creating mobile apps requires extensive coding skills and resources; with Swiftspeed, anyone can build fully functional apps simply by answering a few questions. app builder The key to our offering is the newly released groundbreaking AI technology. Our platform can take basic inputs and preferences from users and then automatically generate complete mobile apps. This has the potential to open up app development to a whole new segment of entrepreneurs, small businesses, and individuals with ideas for useful apps but without the means to code them. While website builders like WordPress and no-code platforms already exist, Swiftspeed is the first comprehensive app development solution of its kind. Our early reviews have cited Swiftspeed's app development tool's ease of use, flexible features, and speed as major advantages to competitors. The AI builder allows error-free app development in days or weeks rather than the months of work typically required. This gives Swiftspeed the potential to truly democratize app development. With mobile apps becoming integral for businesses and brands of all sizes, we aim to make app development accessible for organizations that have previously struggled with the costs and complexities involved. We will represent a breakthrough in enabling any idea to become a fully functional app quickly, affordably, and intelligently. The Problem With Traditional App Development Traditional app development can be an extremely time-consuming and expensive process. Developers must manually code apps from scratch, which requires substantial technical expertise and resources. Here are some of the key challenges with traditional app development: It is a lengthy process, often taking months or even years to build an app with a traditional approach. Developers must code each feature and function, integrate APIs, build the user interface, and test extensively. This makes launching a new app slow and difficult. The costs are very high, with the average mobile app costing $250,000 to $500,000 or more to develop. Complex enterprise apps can cost millions of dollars. These high costs put app development out of reach for many organizations. Significant technical expertise is required, so companies must build large development teams with skills across multiple platforms and coding languages. This level of talent is expensive and in short supply. Iterating and adding features is time-consuming, as developers must manually update code each time. This makes it hard to adapt apps quickly to changing user needs. Human error can cause bugs, crashes, and security vulnerabilities. Extensive testing is required to ensure quality and reliability. Designing a polished, intuitive user interface requires a specialized skill set. This further increases costs and project timelines. Summarily, while traditional coding yields extensive customization options, it comes with steep price tags, long timelines, scarce developer talent, and significant risks. There is an urgent need for faster, more affordable app development. How Swiftspeed's AI App Builder Works Swiftspeed's AI allows anyone to create custom mobile apps without coding. The proprietary technology was developed over 5 years by a team of engineers and AI experts. app creator At the core of the platform is a powerful AI agent that transforms app requirements into full-fledged apps. Users simply describe what they want their app to do through a conversational interface. The AI asks clarifying questions and suggests relevant features to build a comprehensive app specification. Once the specification is complete, the AI generates optimized code and assets on the fly to build a production-ready cross-platform mobile app. The system handles all the complexities of mobile programming, from database setup to UI design, to deliver a polished app tailored to the user's needs. Our sophisticated mobile app features and templates also allow for manual customizations, enabling app developers to fine-tune the app to their branding taste. Swiftspeed combines rule-based reasoning, statistical machine learning, and neural networks to "learn" app development. Our knowledge base comes pre-trained with thousands of app examples across every app category imaginable. The platform converts basic app descriptions into sophisticated software. The AI handles everything from wireframing and prototyping to writing, testing, and compiling code. It integrates with popular and services to enable advanced functionality. APIs The end result is a unique custom app that matches the user's vision without requiring any specialized skills. Key Benefits and Advantages Here are the several key benefits of using our AI tool compared to traditional app development methods: Faster App Development Swiftspeed develops mobile apps in a fraction of the time required for traditional coding. Whereas manual coding can take months or years for complex apps, our tool can generate full apps in minutes and deploy them across Google Play and the App Stores in Minutes. This accelerated timeframe is possible because the AI handles much of the grunt work involved in coding, freeing up developers to focus on higher-level design and customization. Lower Costs By automating parts of the development process, we drastically reduce the manhours required for coding. This directly translates into lower costs compared to outsourced or in-house development teams. Swiftspeed offers fixed-price app development with full transparency into pricing. There are no hidden fees or surprise overages. Higher Quality Apps AI-generated code often has fewer defects than human-written code. Our AI app creator also integrates best practices and avoids common coding pitfalls, resulting in more stable, higher-quality apps with fewer bugs and security vulnerabilities. The AI can also ensure apps meet the latest standards and app store regulations. Examples and Case Studies - Apps Built With Swiftspeed Swiftspeed app builder has already been used to develop some impressive apps across a variety of industries and use cases. Here are just a few examples: Educational App An education technology company , has launched an e-learning app tailored for students and educators, offering features like interactive course material, real-time progress tracking, virtual classrooms, and a collaborative forum for discussions. aptLearn The app enables enhanced interaction and learning continuity between students and educators, regardless of their location. Role-based access control is implemented to maintain a secure learning environment for all users. Moreover, this app's backend is fortified with advanced integrations to ensure. Ride-Hailing App A budding rideshare startup utilized Swiftspeed's design capabilities and ride-hailing templates to effortlessly create a custom app emulating popular services like Uber. This app interfaces with GPS tracker systems to monitor user trips and integrates with third-party payment gateways for seamless transactions. Users can access various ride options, estimate costs, track their rides in real-time, and share their experiences on social media. An E-Commerce App An e-commerce business launched a sophisticated app designed to optimize the online shopping experience. The app encompasses features such as real-time inventory updates, order tracking, customer support, streamlined shipping schedules, and hassle-free returns processing. This innovative app fosters improved communication and operational fluency between e-commerce operations and fulfillment warehouses. Role-based access control guarantees secure and specified functionality across the company, while the app's backend, powered by robust integrations, ensures top-tier data security and consistent availability for users. Custom Community App A local community organization used Swiftspeed to build a custom app to connect members. Features include event listings, discussion forums, member directories, shared files and photos, push notifications, and volunteer signup management. The easy-to-use app improves community engagement and organizers manage it themselves thanks to Swiftspeed's intuitive interface. Swiftspeed allows users to quickly build fully featured custom apps for any industry or need with minimal technical expertise. These examples showcase the power of Swiftspeed's AI app builder in creating tailored mobile experiences. Conclusion Swiftspeed's AI App Builder is revolutionizing mobile app development. With its innovative technology, We are dramatically changing the landscape of app development and disrupting the status quo. Whereas traditional app development is slow, expensive, and requires extensive coding skills, our platform allows anyone to create a custom app in minutes. We handle all the coding in the background, helping users focus on what matters most: their business. By democratizing app development and making it accessible to more people, we have the potential to release a wave of app innovation. Our technology could also enable new apps to be created for niches and audiences not previously served, such as specialized apps for local communities, small businesses, hobby groups, and more. As AI capabilities continue to advance, we are positioned at the vanguard to enhance automated app development further. The technology holds great promise for making app development faster, easier, and more powerful. About Swiftspeed Swiftspeed was founded in 2018 with the mission to make mobile app development faster, easier, and more accessible. Based in Bell Yard, London, and led by CEO , Swiftspeed aims to democratize app development so anyone can bring their ideas to life. Akinola Abdulakeem Our humble beginnings trace back to a simple question - why should building an app require months of development and six-figure budgets? Our founder saw an opportunity to leverage AI and machine learning to simplify and automate much of the app-building process. After initial research and prototyping, We released the first version of our App Builder in 2018 and . This new innovation allows users with no coding skills to create custom mobile apps by simply describing their needs and design preferences. announced AI Integration in January 2024 The AI generates full native app source code optimized for iOS and Android and makes it accessible to customers. Within our first year, over 200,000 apps were created with the platform. Today, Our App Builder is used by small businesses, entrepreneurs, non-profits, and Fortune 500 companies alike. Our company aspires to become the global leader in no-code mobile app development, empowering ideas and enabling digital transformation. With the pace of Swiftspeed's growth, this vision may soon become a reality.

Swiftspeed: Revolutionalizing App Development With Their New AI App Builder: How It Works

সুইফ্টস্পিড: তাদের নতুন এআই অ্যাপ নির্মাতার সাথে অ্যাপ ডেভেলপমেন্টকে বিপ্লবী করা: এটি কীভাবে কাজ করে

Swiftspeed: Revolutionierung der App-Entwicklung mit ihrem neuen AI App Builder: So funktioniert es

Swiftspeed: revolucionando el desarrollo de aplicaciones con su nuevo creador de aplicaciones de IA: cómo funciona

Swiftspeed : révolutionner le développement d'applications avec leur nouveau générateur d'applications AI : comment cela fonctionne

स्विफ्टस्पीड: अपने नए एआई ऐप बिल्डर के साथ ऐप डेवलपमेंट में क्रांतिकारी बदलाव: यह कैसे काम करता है

Swiftspeed: 新しい AI アプリ ビルダーでアプリ開発に革命を起こす: その仕組み

Swiftspeed: 새로운 AI App Builder를 통한 앱 개발 혁신: 작동 방식

Swiftspeed: revolucionando o desenvolvimento de aplicativos com seu novo AI App Builder: como funciona

Swiftspeed: революция в разработке приложений с помощью нового конструктора приложений с искусственным интеллектом: как это работает

Swiftspeed: Yeni Yapay Zeka Uygulama Oluşturucusu ile Uygulama Geliştirmede Devrim Yaratıyor: Nasıl Çalışır?

Swiftspeed: Cách mạng hóa việc phát triển ứng dụng với Trình tạo ứng dụng AI mới của họ: Cách thức hoạt động

Swiftspeed：通过新的 AI App Builder 彻底改变应用程序开发：它是如何工作的

As companies welcome the exciting benefits of Generative AI for getting more work done, they also face a tricky problem: Gen AI Hallucinations. Ideally, AI-generated results should be reliable without human assistance. But sometimes, AI can hallucinate or generate fictitious content that poses a great risk. When wrong information gets out, it doesn't just cause embarrassment—it can hurt a company's reputation and erode people’s trust. So, while everyone wants to use Gen AI to be more efficient, they must also be careful to avoid these mistakes that could damage their image. What is a Gen AI Hallucination? , seen in Large-Language Models (LLMs) like ChatGPT, often leads to inaccurate information and responses, ranging from slight deviations from facts to entirely fabricated content. Gen AI hallucination LLMs operate by predicting the next word in a response based on the user's input, lacking independent reasoning abilities, which can result in errors. Given the natural fluency and coherence of the generated text, fact-checking is essential to prevent the . dissemination of false information Example: An infamous incident of AI hallucination occurred with Bard when asked about discoveries from the James Webb Space Telescope. Bard claimed the telescope was the first to capture images of an exoplanet outside our solar system. But this information was false and was quickly debunked through fact-checking. Why Do Gen AI Hallucinations Happen? Hallucinations in generative AI models often stem from the following: Irrelevant or subpar information fetched by the retrieval system can degrade output quality, causing errors or misinformed responses. Inaccurate context retrieval: Poor user prompts can mislead the LLM, resulting in responses based on wrong or unsuitable context. Additionally, if the underlying app prompt is ineffective this could also hurt the GenAI app’s information retrieval process. Ineffective prompts: Difficulty with idioms, slang, or non-English languages may result in incorrect or illogical responses, especially if the retrieval system fails to obtain or decipher relevant context in these languages. Complex language challenges: Why Are Gen AI Hallucinations a Problem? AI hallucinations pose serious risks beyond technical errors, . Let’s briefly examine all the problems caused by Generative AI hallucinations. impacting your brand's reputation and consumer trust Reputational Damage The reputation and status of the impacted party may be seriously damaged when AI-generated content is deliberately used to propagate false information or defamatory statements. This involves a bad reputation, a decline in trust, and financial consequences. It can be difficult to reverse or undo the harm caused once false information has been disseminated, which can have a long-lasting impact on relationships with partners, consumers, and the general public. Legal Consequences If we look at the legal aspects of AI hallucinations, we’ll find tons of them. These legal consequences range from defamation, invasion of privacy, or copyright infringement. This could also happen unintentionally. Imagine you’re just exploring AI but end up creating something that someone doesn’t. In that case, you’ll have to face legal charges if it becomes viral. A few years back in 2019, one such event happened when a went viral. He was caught on camera making offensive and untrue remarks. The company's stock price dropped significantly, but it was a joke. After that, the business took the makers of the deepfake video to court. deepfake video claiming to be Mark Zuckerberg Medical Mistakes Hallucinations caused by generative AI can potentially cause medical errors by presenting false or misleading information. For example, errors in the data produced by the AI model may lead to a misdiagnosis or an ineffective course of therapy if medical personnel rely on AI-generated content for diagnostic or therapeutic suggestions. A study published in raised issues with commercial AI algorithms used to diagnose dermatology. The study discovered that these algorithms, intended to evaluate pictures of skin lesions, showed notable errors, especially when differentiating between benign and malignant lesions. This demonstrates how crucial it is to thoroughly validate and continuously assess AI algorithms in the healthcare industry. JAMA Network Open in 2020 Compliance Challenges Compliance issues can occur when AI hallucinations result in the creation of content that does not adhere to industry norms or legal constraints. For instance, businesses may face legal and financial repercussions if by unintentionally revealing private information. AI-generated material violates data privacy laws Creating biased or discriminating information against equality and fairness could be one way to do this. Organizations must implement robust governance frameworks and quality assurance processes to address compliance challenges associated with AI hallucinations. Financial Loss Hallucination in Artificial Intelligence can also result in financial loss. That occurs if inaccurate content creation causes expensive mistakes or interferes with corporate processes. For instance, if financial reports produced by AI are erroneous or contain misleading information. It forces the investors or companies to make poor investment decisions that result in losses. Similarly, inaccurate or misinterpreted customer preferences might cause AI-generated marketing initiatives to fall flat with target demographics. Thus wasting money and missing out on sales possibilities. Disruption of User Trust AI hallucinations can create fake information that closely mimics reality, which can seriously undermine consumer trust. This may result in disseminating false information, swaying public opinion, and growing fake news. Users will find it easier to trust sources and separate fact from fiction when AI-generated material is identical to real information. This erodes public confidence in disseminating information and may significantly affect public trust. Are Artificial Intelligence Hallucinations Always Bad? Hallucinations, despite their potential risks, can hold value. LLMs exhibit a unique set of strengths and weaknesses. While they may struggle with tasks traditionally associated with computer proficiency, such as search capabilities, they excel in storytelling, creativity, and aesthetics. Hallucinations in AI should be reframed as a feature rather than a flaw within AI systems. Marketers can leverage this by prompting AI to hallucinate its interface, enabling it to perform tasks that are otherwise challenging or costly to measure. For instance, marketers could assign scores to various objects based on their alignment with the brand and then task AI with identifying potential lifelong consumers based on those scores. Ultimately, manipulating these hallucinations could lead to innovative solutions and enhanced outcomes in the advertising and marketing space.

Gen AI Hallucinations: The Good, the Bad, and the Costly

221 Stories To Learn About User Experience

Have you ever stumbled upon a meme that made you pause, not just because it was funny, but because it felt like a direct commentary on your life? Picture the "This Is Fine" dog, sitting calmly amidst flames, a cup of coffee in hand. It's funny, yes, but isn't it also a striking metaphor for how we often cope with chaos in our own lives? Memes have transcended their origins as internet jokes to become mirrors reflecting society's psyche. They're making us laugh, think, feel, and sometimes question. Let's dive into how these digital snippets have become the cultural artifacts of our time. Memes as Cultural Artifacts Consider the "Distracted Boyfriend" meme: a man turning his attention away from his partner to admire another. At first glance, it's just a funny take on infidelity. But if you stare at it for one too many seconds, you'll find layers of social commentary on loyalty, distraction, and the human tendency to desire what we don't have. Basically, it’s a modern take on ‘your grass is greener on the other side of the fence’. Memes like these capture the zeitgeist of our era, encapsulating complex societal norms and behaviors in a single image. Read: Elmo Checks In, Everything Is Not Okay: A Lesson in Empathy for Brands Virality and Relatability Why do certain memes, like the "Bernie Sanders at Inauguration" with his mittens, capture our collective imagination? It's their relatability. In Bernie's case, it was the universal feeling of wanting to be comfortable anywhere, even at a presidential inauguration. These moments go viral because they tap into shared emotions and experiences, making us feel part of a larger narrative. And we love to be part of something more than just ourselves. Memes and Social Commentary Memes often serve as a platform for subtle social critique. Take the "Area 51 Raid" meme, which started as a joke but soon evolved into a commentary on government transparency and collective action. What does it say about our society when a meme mobilizes hundreds to "see them aliens"? Dissecting Why Memes Work From a psychological standpoint, memes do more than just entertain; they fulfill several intrinsic human needs. At their core, memes can reinforce a sense of belonging by creating an 'in-group' where shared knowledge and humor signify membership. This aligns with Abraham Maslow's hierarchy of needs, where belongingness is a crucial step toward achieving esteem and self-actualization. Memes also serve as a vehicle for catharsis, allowing individuals to express frustrations, fears, and desires in a socially acceptable manner. The popularity of memes expressing existential dread or the absurdity of adult responsibilities speaks to a collective need to voice and validate our anxieties in a world that often demands stoicism. Furthermore, the virality of memes can be attributed to their simplicity and the cognitive ease with which they can be processed and shared. This simplicity, however, belies their complexity as cultural artifacts capable of influencing public opinion and shaping discourse. They act as a barometer for societal values, norms, and tensions, offering a snapshot of the zeitgeist in a way that is accessible and immediately relatable. The Future of Memes: Deciphering the New Lexicon Today, it is evident that memes have evolved into a language of their own—a lexicon so rich and nuanced that understanding a single tweet might require fluency in multiple meme dialects. Consider the depth of knowledge required to navigate the layers of meaning in a tweet that references not one, but five different memes. Each meme, a building block of a larger narrative, contributes to a complex web of cultural references that are instantly recognizable to the initiated. This phenomenon has created a unique in-group—those "chronically online"—who communicate in shorthand so dense that someone not versed in the latest internet lore might find themselves adrift in a sea of seemingly nonsensical exchanges. Take, for example, the statement about an unkillable mushroom: " ." Out of context, it's an absurd declaration. Yet, within the meme culture, it represents so much more. Reddit You cannot kill me in a way that matters It’s like memes have become the Rosetta Stone of the internet age, allowing us to share jokes, insights, and solidarity across linguistic and geographic boundaries. This way, we are basically compressing intricate thoughts and emotions into a format that is digestible and relatable on a universal scale. Conclusion So, next time you come across a meme that makes you chuckle or pause in thought, remember: you're engaging with a piece of cultural history, a snapshot of our collective psyche. What does that meme say about us, and what does our reaction to it reveal about our values, fears, and hopes? In the meme economy, every share, every like, every comment is a contribution to the ongoing discourse of our time. What will your contribution be?

Understanding Internet Culture: Why Are Memes Such a Hit?

Today I am speaking with Ilan Rakhmanov, the Founder & CEO of ChainGPT and we'll discuss the innovative integration of artificial intelligence within the Web3 and blockchain landscapes, highlighting the current state, challenges, and future opportunities of this fusion. We will also discuss the critical skillsets needed for professionals in this evolving field, the genesis and mission of his company in bridging AI with blockchain, and the potential of AI-driven tools to transform digital art, news aggregation, and blockchain analytics. Please introduce yourself and tell us what you do? Hi, my name is Ilan Rakhmanov, I am the CEO & Founder of ChainGPT, the AI-powered Web3 infrastructure providing a diverse suite of tools and services. I am a serial entrepreneur and businessman, and for the past couple of years since founding ChainGPT, I have been working to bring the company and its AI services to Web3 and blockchain to help fill the void of real AI and Web3 integration. My experience leading multiple 7-9 figure companies has included roles across various industries and sectors ranging from software and IT to business, law, design, and marketing. I’m never bored. Leading the ChainGPT team, we have collaborated with more than 70 Web3 companies and industry frontrunners. Can you describe the current state of integration between AI and blockchain technologies in the Web3 space? As of now, integration is still quite nascent, but we are seeing an influx of innovative applications. We’re witnessing continuous R&D being done behind the scenes, so we can certainly anticipate more use cases and integrations in the pipeline. What skillsets are becoming increasingly important for professionals working at the intersection of AI, blockchain, and Web3? As with any new industry, the skillsets we see quickly becoming a prerequisite for professionals working at the intersection of AI, blockchain, and Web3 are an increased proficiency in existing industry skills converging with new knowledge and technical prowess tailored to this fresh industry. Existing skills such as coding language, large dataset analysis, and machine learning, normally reserved for computer developers and IT departments are quickly becoming a general skill required across the board. Similarly, soft problem-solving skills, teamwork, and effective communication have evolved to become similarly critical. However, these base-level requirements must also embody novel or more advanced capabilities unique to this new industry. These capabilities include skills and knowledge surrounding smart contract creation, development and auditing, advanced natural language processing (NLP), reinforcement learning models, consensus mechanisms, advanced cryptography, and zero-knowledge proofs (ZK-proofs), alongside familiarity with blockchain development platforms such as Ethereum, Solana, and Polygon. This might sound incredibly specific now, but all of this knowledge and skillsets will inevitably take the place of the previous “Microsoft Office proficiency” requirement of the early 2000s. A compilation of many, if not the vast majority, of these skills will become standard requirements for any professional operating in the intersection of AI, blockchain, and Web3. Can you describe the genesis of ChainGPT and its mission within the Web3 and blockchain space? We established ChainGPT in 2022 as we recognized the continuously growing role of AI across every sector, and the gaping infrastructure void to facilitate and support this. ChainGPT is an AI infrastructure platform focusing on the Web3 blockchain and crypto space. Our mission is to provide individuals and startups with a comprehensive suite of AI solutions tailored specifically to the evolving needs of this industry. Since launching, we’re thankful to be backed by industry leaders such as Google, to be a part of the prestigious NVIDIA Inception Program for Startups, and to have forged partnerships with more than 70 leading Web3 companies, including BNB Chain and Tron. That being said, ChainGPT is not just about AI—it's about community engagement and empowerment, and we embody that being ensuring that our staking and governance models allow $CGPT holders to shape the ecosystem's future. Ultimately, ChainGPT's genesis and mission revolve around providing accessible and advanced AI solutions to fuel innovation and success in the Web3 and blockchain spaces, supported by a strong community and strategic alliances. Could you elaborate on the technology behind ChainGPT’s AI News Auto-Model and how it ensures the relevance and accuracy of the curated content? Our fully automated AI Generated News service aggregates breaking news from only the most trusted sources within crypto and Web3, and then extracts valuable insights from these publications. Given the unique nature of how crypto and blockchain news is disseminated, that does include reliable or verified social media accounts. After aggregating, the tool generates short, to-the-point summaries that are easy for readers to consume on the go. Aside from that, the service employs real-time learning capable of catering to diverse use cases, stories, interactive trading guides, and more. How does the AI NFT Generator align with current digital art and blockchain technology trends? Even before the explosive launch of ChatGPT in November 2022, we had already begun to see growing use of AI-art generators like Dall-E and Midjourney—despite not being as advanced as the tools we have now. The NFT market itself has always been largely composed of unique digital art, so our use of generative AI to mint unique art collections and allow users to mint their own NFTs is just the natural evolution of the NFT space. In your view, what are the most promising opportunities that arise from the convergence of AI and blockchain in the Web3 era? At face value, they are quite different technologies with different strengths, but there are countless ways for them to co-exist. The clearest intersection between AI and blockchain can be seen in AI-arbitrage bots, AI tools creating a better user experience, and NFT generation. However, there are more expansive, infrastructure-level possibilities or applications that cover areas like boosting data ownership and transparency. Aside from that, generative AI’s approachability and fluency can help guide new Web3 users through a fairly complex interface, helping them avoid mistakes and possible scams. But this is definitely an area where we will continue to see a steady stream of innovation and interesting applications. Can you discuss the capabilities of the ChainGPT Chatbot, especially in addressing technical blockchain and crypto queries? The ChainGPT Chatbot allows users to access information and answers to any question related to or within blockchain and crypto. Not unlike the news service, the chatbot utilizes the most trustworthy sources within crypto and blockchain to gather, analyze, process, and distill information while extracting valuable and relevant insights. By doing so, our chatbot provides a fast source of information to any query no matter how general or technical it may be in a way understandable to the user who might not have as much experience with blockchain or crypto jargon. What are the most significant challenges and opportunities in AI-driven blockchain analytics? Given the immense reliance on data to power large language models in generative AI, the biggest challenge will be trying to ensure the integrity of the data being used to train these analytics tools. In Web3 especially, where there can sometimes be a financial stake, it is critical to ensure that those designing these models don’t use them to serve their financial interests. With most of your team being developers, how does ChainGPT foster a culture of innovation and collaboration? First of all, one of my most important responsibilities as CEO is ensuring that our developers, product teams, and everyone across the company understand the value in collaborating. It’s on me to foster that as a startup leader overseeing a team that works across borders. Thankfully we have an incredible group of people who are creative, thoughtful, and bring a lot of different perspectives and experiences to the table. How does ChainGPT plan to grow and engage its user community further? We just need to keep doing what we are doing—building meaningful products and tools that regular users and developers can use. We are invested deeply in our IDO launchpad and incubator which helps solidify our ecosystem and we are deeply connected with our community across multiple social media platforms. By continuing to build and engage, we will see our community grow. Where do you see ChainGPT in the next five years, especially in the rapidly evolving Web3 environment? I fully anticipate Web3 becoming the current internet iteration in five years. With that said, we see ourselves as an integral component of a robust Web3 ecosystem by continuing to provide unique and innovative AI solutions that either tackle space challenges or enhance other aspects. There’s no question the industry will look different in five years. Still, as early innovators of AI in blockchain, we will remain at the forefront of all industry developments, finding a niche or mainstream issue to improve or solve. What future trends do you anticipate in integrating AI and blockchain technologies, and how might these shape the Web3 landscape? I anticipate many future trends, but chief among them will be the use of blockchain to make AI more transparent and secure. This concept is still very academic but many of the brightest minds are working on applying it in numerous ways. For instance, imagine using blockchain to give individuals complete control over their personal data. With AI models becoming increasingly powerful as we compile more data, enabling Web3 to govern these models transparently can enable them to scale in a way that benefits the entire population. Of course, we’ll need to wait and see where regulations also come into play in this realm. On a micro level, I fully anticipate AI being used to help developers build, scale, and deploy apps rapidly. I also fully expect to see AI being used to track market shifts and trends more accurately and transparently.

Exploring AI's Impact on Blockchain with Ilan Rakhmanov

This year has been important in so many ways. I’ve been publishing more content than ever. I had a part-time personal assistant. Got access to many wonderful opportunities. I raised my net worth by 60%. I’ve built a great team and moved the product forward at the cybersec SaaS I work at. I almost finished my master’s in management (thesis postponed to the next year). I designed and furnished my first (and hopefully last) apartment. Hell, I even started using separate products for hair and body. And there was a lot of personal change as well. Join me in this short recap if you want to learn where a path like mine can lead you. I’ll share mostly blog and online presence-related details with some professional (engineering), financial, and private information sprinkled on top. Happy to tell more to those interested. Why am I here? I like sharing knowledge, writing, and helping people. I believe that publishing is a vital part of being a professional. Being here, rushing some content out even though it’s not perfect, being vulnerable - that’s a great way to get to know more people and open one’s mind to new points of view. I think one person - Austin Kleon - put this in words much better than me. Check out his short book, and you’ll know everything. is about why generosity trumps genius. It's about getting , about using the network instead of wasting time "networking." It's not self-promotion, it's self-discovery--let others into your process, then let them steal from you. Show Your Work! findable I had a blast putting effort into this blog and other projects this year. Some people enjoyed it as well, as seen in stats from various platforms. Stats Here’s the most interesting numbers i dug out! My blog was visited a total of 43 thousand times: I also got an additional 23k visits to my articles republished on Hackernoon. My LinkedIn content generated 65k impressions, similar to views on the blog. I’ve been posting quite irregularly, as seen in the graph below. Near the end of the year, I started listening to the numbers, resulting in more engagement than ever. Posts mentioning others and with my photos are noticeably more popular. Other numbers: My newsletter got 50 subscribers (up from 24 in July). On Twitter, my posts generated ~300k impressions. I spoke at eight conferences. I was a guest in 5 podcast episodes. I have published 12 blog articles. Best content If you missed some of my articles, here’s a chance to catch up. It’s clear that my most popular content was not opinions, summaries, or anything related to my publishing online. Most viewed articles were software engineering tutorials. A lot of traffic came from Google search, where I ranked quite high for some keywords. I also think I figured out a good way to create thumbnails. 11k views: Extracting and Generating JSON Data with GPTs, LangChain, and Node.js 9K views: How I Effortlessly Create PowerPoint Presentations with ChatGPT and MARP (as a Software Engineer) 2K views: Fine-tuning OpenAI GPT 3.5: Practical example with Python Tip for me for the next year: write more engineering tutorials. Other projects simple message from me here: don’t build a house or buy a new unfinished apartment if you’re not into it. I had to learn interior design from the ground up. Get to know materials, design solutions, too many new technical terms. It interesting, it can be fun, but it is a lot of work. Even though I’m very happy with the outcome, I think I could have spent my time better. Real estate investing / buying a new apartment: really is kicked off 2, shipped none. Excuses? Master thesis and moving apartments. Picking them up again in 2024. One was MeetingFusion: , the other is still unpublished. MF got several sign-ups on the waiting list. Side projects/startups: https://meetingfusion.com : Struggling, will post an update soon. My old AI Startup, Sentimatic : I’ve been able to provide some small-scale but well-paid software engineering services on the side. Consulting : applied for government funding and spent a lot of time. My assistant spent 175 hours on it! The application was of a master’s thesis size. Funding for new AI project Opportunities Writing here gave me more opportunities than ever before. Apart from everything mentioned above, here are some highlights. I’ve been invited a few times to run a workshop about topics I write about by several engineering companies. All my offers were rejected due to the price. I am okay with this, as I only considered it if I was able to provide the absolute best quality of the workshop. I’ve got more interesting job offers than ever. I did not proceed with any, as I’m committed to my current work at Netacea. I’ve been asked to develop a Vue.JS course for a popular course platform. The contract signing is in progress, so hopefully, next year, some people will be able to learn something new using the course. I’ve been invited to many more conferences than I spoke at. Had to choose where to spend my time. Money My work around this blog and anything professional that is not related to Netacea (the cybersec startup I’ve mentioned) only me money. cost Apart from this money pit, I am quite okay at managing my finances, saving it, and investing. I’m also fluent in tax codes. I don’t let money go to waste. I am following the stock market to a limited extent as well. All of this pays dividends and helps me feel more secure in life. I recommend everyone to have basic financial fluency. I won’t get into much detail and exact numbers, but I’ll share some stats to give you some insight into how I think about money and investing. I can share some guidelines for engineers in a separate article. Let me know. My net worth (assets less liabilities) is allocated between cash, real estate, and stocks. I am definitely real estate heavy but I have access to really good advice. Going forward, I’d prefer to have my net worth allocated in some business I run or contribute to in the form of work or investment. I am tracking my savings and investments as a whole by doing a monthly net worth evaluation (it’s a quick exercise at this point). I’ve been doing it since April 2019 and have been able to maintain a good pace of growth, albeit with some ups and downs. Don’t get all excited; the starting point was fairly low! My short advice for friends: use retirement tax-advantaged accounts buy the simplest stock market investments - index funds, S&P500, whole US stock market get a mortgage; it’s a low-interest rate leveraged capital, and overall properties keep up with inflation rates have a safety net with some small cash (as in a bank) pile to bail you out in emergencies Plans My plans regarding this blog and my online presence are still in the making. I want to aim for something that will keep giving me satisfaction while being relevant and impactful. Here are the goals I am thinking about: Launch and grow a newsletter up to at least 1000 subscribers. Shipping one successful SaaS tool on my own. And then grow or sell it. Grow my global (online) and local (at Netacea) influence as an engineering leader. Write one blog post per week. Publish video content and get a reasonable following. Regardless of the details of these, you’ll likely see much more from me next year! Stay tuned. Summary So yeah, a lot happened, and I want it to continue to happen. This year has been another great period of my life which I’m truly grateful for. As for this blog, I want to keep writing and being a part of the global community. After years of learning and working, I finally think like I have something to say. For me, writing has always been this elusive thing that I never quite got to do. I stopped hiding behind perfectionism and started sharing. And many people found it useful, which just makes me happy. I keep learning but this time, I make some of my private notes public. This year, I’ve learned a lot, made mistakes, and made new friends. The future looks exciting. See you there. Have the most wonderful year of your life. Don't forget to follow me on Twitter and subscribe to my blog’s for more tips and insights! @ horosin_ newsletter If you don't have Twitter, you can also follow me on . LinkedIn Also published . here

Reflecting On My 2023: A Year of Growth, Opportunities, and Financial Insights

How are you, hacker?
🪐 has got you covered with fresh content from our trending stories of the day! Set email preference . Want to know what's trending right now?: The Techbeat by HackerNoon here Investigating Internet Freedom: Unseen Aspects of Our Online Existence By [ 2 Min read ]
Join award-winning journalist Nicole Scott in her latest video series, exploring digital democracy and our collective role in shaping the online world. @ionos Read More. A Beginner's Guide to Understanding SQL Window Functions - Part 2 By [ 8 Min read ]
let us explore more advanced SQL concepts. It may sound a bit complicated at first glance, but I will provide simple examples suitable for beginners... @yonatansali Read More. Meeting Customer Needs With User-Centric Product Development By [ 6 Min read ]
In a nutshell, user-centric product development implies attentiveness to the client's needs. This approach puts the customer at the heart of the product @tonythevoit Read More. Unveiling the Best of 2023: Top Stories, Writers, Tech Categories, and Templates! 📚 By [ 5 Min read ]
We've compiled a 2023 roundup of the top most-read stories, top active writers, top tech categories, and even the top most popular templates. Take a look! @editingprotocol Read More. Geofencing Made Easy: Implementing Geofencing With ipstack API By [ 6 Min read ]
Would you like to discover the advantages of using geofencing with the ipstack IP API? Learn geofencing in depth with ipstack API. @apilayerapimarketplace Read More. TypeScript Might Not Be Your God: A Case Study of Using JSDoc for JavaScript Projects By [ 10 Min read ]
Unlock JavaScript typing potential using JSDoc & TypeScript synergy. Bridge the gap between clarity and flexibility for seamless development. @wha1s1ove Read More. Web3 Gaming: 7 Predictions for 2024 By [ 6 Min read ]
The resilience of Web3 gaming, bolstered by an emerging crypto bull market, is paving the way for what industry experts are forecasting as a defining year. @ggmetagame Read More. Using Witnet to Overcome The Challenges of Developing A Truly Multichain Oracle Network By [ 4 Min read ]
Tomas Ruiz on overcoming the challenges of developing a truly multi-chain oracle network using Witnet. @oraclesummit Read More. Everything I Achieved With My Mozilla Contribution Sprint By [ 2 Min read ]
After years of supporting Mozilla financially and spreading messages with my peers, I returned as an active contributor. @bogomil Read More. Reverse Proxying — the Backbone of Microservices Architecture By [ 5 Min read ]
Understand how reverse proxies powers almost every service that we use today from monolith to microservices. It is a vital piece of 21st century infrastructure! @infinity Read More. How to Shed Pounds Off Your Docker Image By [ 4 Min read ]
Checkout the best practices for making your Docker images slim and fast. Be Kubernetes ready and streamline your development process! @infinity Read More. Creating the Perfect Dashboard: Visualizing Your Metrics By [ 10 Min read ]
A well-designed dashboard can become a key element in forecasting trends and anticipating outcomes, impacting the efficiency and motivation of the team. @vanpelz Read More. I Created a React Utility Component for Animations With Tailwind and CSS: AnimateIn By [ 5 Min read ]
<AnimateIn/> is a reusable React component that I’ve made to drop in whenever I want to quickly add some animation effects to my projects. @johnpolacek Read More. Become a Productivity Beast With These 16 Killer Work-Flow Tools By [ 4 Min read ]
I've curated 16 tools from website management and content creation to design apps, coding utilities, automation tools, and much more that will save you time. @madzadev Read More. AI's Dirty Secret: The Hidden Cost of its Environmental Impact By [ 6 Min read ]
As AI's carbon footprint expands, the environmental impact becomes an increasingly critical issue. Check out 5 ways  to build a sustainable AI-powered future. @viceasytiger Read More. How to Build a Decentralized Charity Platform with Next.js, TypeScript, and Solidity By [ 8 Min read ]
Build a decentralized application (Dapp) for a charity marketplace. Learn Solidity, Ethereum, OpenZeppelin, Next.js, TypeScript, Hardhat, EthersJs, Redux-Toolki @daltonic Read More. Designing a Video Game: Principles and Rules of Modern Level Design By [ 12 Min read ]
Today's gamers expect levels that are not only visually stunning but also richly interactive and immersive. How can we reach these expectations as developers? @beatboyninja Read More. How I Improved My English Speaking Skills With AIs That Should Work For You Too By [ 9 Min read ]
With AI language learning apps, such as speaking assistant Pronounce, text checker Grammarly, and ChatGPT, I boosted fluency, grammar, pronunciation, and more. @pronounce Read More. Python's Testing Playbook: Building Bulletproof Code By [ 11 Min read ]
Master Python testing: From unit tests to TDD and BDD, learn to build resilient code with pytest and unittest in this in-depth guide. @nikolaibabkin Read More. Dopamine Pumps: Neurohacking Habits By [ 7 Min read ]
The marketing industry operates on a spectrum of neurohormones, with dopamine at its peak and corticoliberin at its nadir. 🧑‍💻 What happened in your world this week? It's been said that , , . Feeling stuck? We got you covered ⬇️⬇️⬇️ We hope you enjoy this worth of free reading material. Feel free to forward this email to a nerdy friend who'll love you for it.
See you on Planet Internet! With love,
The HackerNoon Team ✌️ @cryptobro Read More. writing can help consolidate technical knowledge establish credibility and contribute to emerging community standards ANSWER THESE GREATEST INTERVIEW QUESTIONS OF ALL TIME

The TechBeat: Investigating Internet Freedom: Unseen Aspects of Our Online Existence (1/24/2024)

Stories, News & Mentions about FLUENCY

Fluency’s Stories on HackerNoon