Your attention span has shrunk by nearly two minutes in the past 20 years, driven in part by technology’s hold over your work and home life, experts say. Here’s how to stay focused so you can get things done.

Your attention span is shrinking, studies say. Here’s how to stay focused

Cannes Film Festival officials are notoriously strict when it comes to enforcing protocols for black-tie ensembles on guests before they are allowed to ascend the iconic Palais steps. However, at Tuesday’s glamorous opening night ceremony, one guest managed to sneak by security on four legs ...

‘Anatomy of a Fall’ Breakout Messi the Dog Laps Up Attention at Cannes

Rebecca Joynes, 30, has pleaded not guilty to six charges of sexual activity with two teenagers while she was a teacher. She told a court she &#x27;liked&#x27; the attention but denies having sex with the boys

Teacher Rebecca Joynes says she was 'lonely and craved attention' and admits boy, 15, spent night at her flat - Mirror Online

A teacher denies the allegations of sexual activity with boys claiming she only wanted attention.

Salford: Teacher 'liked attention' but denies sex with boys

Rebecca Joynes denies four charges of sexual activity with two teenagers and two counts of sexual activity while being in position of trust

Teacher 'caved in' to attention from pupils but denies having sex whilst teaching them

Rebecca Joynes wept as she told jurors that her baby was taken away ‘24 hours after she was born’

Teacher accused of sex with children tells court she was ‘lonely and craved attention’ | The Independent

The Metropolitan Police has launched a murder probe in south London after the death of the 26-year-old died of stab wounds.Separate stabbing incidents » · Conservative leadership hopeful Boris Johnson has defended his infamous description of black people in Africa bearing “watermelon smiles” ...

'Lonely' teacher says she liked attention but denies having sex with boys

Rebecca Joynes, 30, denies any sexual activity took place, has pleaded not guilty to six charges of sexual activity with two teenagers and is on trial at Manchester Crown Court.

Teacher denies having sex with boys as she tells court she 'liked' attention after break-up | UK News | Sky News

Will Senate Democrats hold hearings into Donald Trump&#x27;s controversial pitch to oil industry executives? It&#x27;s “highly likely,” one key member said.

Trump’s pitch to Big Oil execs sparks Senate Democrats’ attention

The Blueshirts need to channel what they had the last time they ended the Hurricanes’ season, in Game 7 of their 2022 second-round series.

Rangers’ attention turns to Game 5 for second shot at Hurricanes knockout punch

attenio

Authors:
(1) Jianhui Pang, from the University of Macau, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (nlp2ct.pangjh3@gmail.com);
(2) Fanghua Ye, University College London, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (fanghua.ye.19@ucl.ac.uk);
(3) Derek F. Wong, University of Macau;
(4) Longyue Wang, Tencent AI Lab, and corresponding author. Table of Links Abstract and 1 Introduction 2 Related Work 3 Anchor-based Large Language Models 3.1 Background 3.2 Anchor-based Self-Attention Networks 3.3 Anchor-based Inference 4 Experiments and 4.1 Our Implementation 4.2 Data and Training Procedure 4.3 Evaluation 5 Results 6 Analysis 7 Conclusion, Limitations, Ethics Statement, and References A More Experimental Results B Data Settings 3.2 Anchor-based Self-Attention Networks Anchor-based Attention Masks. To accomplish this, we devise anchor-based attention masks, as illustrated in Figure 2. Assuming that the current token in the sequence is a non-anchor token, we allow attention towards previous non-anchor tokens within the same sequence and anchor tokens from preceding sequences, while blocking attention towards non-anchor tokens from previous sequences. This approach ensures that non-anchor tokens can only access information from anchor tokens in previous sequences and the current sequence’s information. Conversely, when the current token is an anchor token, which is the last token in the sequence, we exclusively permit its attention towards previous non-anchor tokens within the same sequence, blocking all other attention. This constraint forces the anchor token to aggregate information solely from its current sequence. Consequently, we replace Eq. (3) with anchor-based attention masks in Eq. (4) to determine the mask of the i-th token in the input text concerning the j-th token (assuming that the i-th token belongs to the k-th sequence). Anchor Token Selection. By implementing the AnSAN mechanism for training LLMs, we can compel the model to compress sequence information into the anchor token and generate new tokens based on the anchor token information from previous sequences and non-anchor token information from the current sequence. The challenge now lies in selecting an appropriate anchor token. In our experiment, we propose two implementation methods: one using the endpoint as the anchor token, and the other appending a new token specifically as the anchor token. This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Jianhui Pang, from the University of Macau, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (nlp2ct.pangjh3@gmail.com); (2) Fanghua Ye, University College London, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (fanghua.ye.19@ucl.ac.uk); (3) Derek F. Wong, University of Macau; (4) Longyue Wang, Tencent AI Lab, and corresponding author. Authors: Authors: (1) Jianhui Pang, from the University of Macau, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (nlp2ct.pangjh3@gmail.com); (2) Fanghua Ye, University College London, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (fanghua.ye.19@ucl.ac.uk); (3) Derek F. Wong, University of Macau; (4) Longyue Wang, Tencent AI Lab, and corresponding author. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Related Work 2 Related Work 3 Anchor-based Large Language Models 3.1 Background 3.1 Background 3.2 Anchor-based Self-Attention Networks 3.2 Anchor-based Self-Attention Networks 3.3 Anchor-based Inference 3.3 Anchor-based Inference 4 Experiments and 4.1 Our Implementation 4 Experiments and 4.1 Our Implementation 4.2 Data and Training Procedure 4.2 Data and Training Procedure 4.3 Evaluation 4.3 Evaluation 5 Results 5 Results 6 Analysis 6 Analysis 7 Conclusion, Limitations, Ethics Statement, and References 7 Conclusion, Limitations, Ethics Statement, and References A More Experimental Results A More Experimental Results B Data Settings B Data Settings 3.2 Anchor-based Self-Attention Networks Anchor-based Attention Masks. To accomplish this, we devise anchor-based attention masks, as illustrated in Figure 2. Assuming that the current token in the sequence is a non-anchor token, we allow attention towards previous non-anchor tokens within the same sequence and anchor tokens from preceding sequences, while blocking attention towards non-anchor tokens from previous sequences. This approach ensures that non-anchor tokens can only access information from anchor tokens in previous sequences and the current sequence’s information. Conversely, when the current token is an anchor token, which is the last token in the sequence, we exclusively permit its attention towards previous non-anchor tokens within the same sequence, blocking all other attention. This constraint forces the anchor token to aggregate information solely from its current sequence. Consequently, we replace Eq. (3) with anchor-based attention masks in Eq. (4) to determine the mask of the i -th token in the input text concerning the j -th token (assuming that the i -th token belongs to the k -th sequence). Anchor-based Attention Masks. i j i k Anchor Token Selection. By implementing the AnSAN mechanism for training LLMs, we can compel the model to compress sequence information into the anchor token and generate new tokens based on the anchor token information from previous sequences and non-anchor token information from the current sequence. Anchor Token Selection. The challenge now lies in selecting an appropriate anchor token. In our experiment, we propose two implementation methods: one using the endpoint as the anchor token, and the other appending a new token specifically as the anchor token. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

The Role of Anchor Tokens in Self-Attention Networks

Authors:
(1) Jianhui Pang, from the University of Macau, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (nlp2ct.pangjh3@gmail.com);
(2) Fanghua Ye, University College London, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (fanghua.ye.19@ucl.ac.uk);
(3) Derek F. Wong, University of Macau;
(4) Longyue Wang, Tencent AI Lab, and corresponding author. Table of Links Abstract and 1 Introduction 2 Related Work 3 Anchor-based Large Language Models 3.1 Background 3.2 Anchor-based Self-Attention Networks 3.3 Anchor-based Inference 4 Experiments and 4.1 Our Implementation 4.2 Data and Training Procedure 4.3 Evaluation 5 Results 6 Analysis 7 Conclusion, Limitations, Ethics Statement, and References A More Experimental Results B Data Settings 3.1 Background Transformers. LLMs are primarily realized as decoder-only transformers (Vaswani et al., 2017; Touvron et al., 2023a,b), incorporating an input embedding layer and multiple decoder layers. Each layer contains a self-attention network and a feedforward network with normalization modules. Crucially, causal attention masks are employed, allowing tokens to attend only to preceding ones. Self-Attention Networks. Typically for decoderonly LLMs like Llama2 (Touvron et al., 2023b), self-attention networks (SANs) map queries Q, keys K, and values V into an output, as delineated in the following equations, where M denotes an L × L masking matrix, facilitating the current i-th token to attend to only preceding tokens whilst disregarding subsequent tokens during the training and inference phases. Keys/Values Caches. In the application of LLMs, the keys/values caches increase with lengthy prefix texts and continuously generated tokens during the inference phase, such as in question-answering (Saad-Falcon et al., 2023), text summarization (Basyal and Sanghvi, 2023), and machine translation (Pang et al., 2024). The key and value matrices associated with tokens of prefix inputs are cached to avoid recomputation and expedite subsequent token prediction (Radford et al., 2019). Additionally, the model generates the output token-by-token in the real-time inference process, which requires more cache memory to store the newly generated sequence. Therefore, addressing the challenges arising from the ever-expanding texts is crucial for enhancing the efficiency of LLM inference. This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Jianhui Pang, from the University of Macau, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (nlp2ct.pangjh3@gmail.com); (2) Fanghua Ye, University College London, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (fanghua.ye.19@ucl.ac.uk); (3) Derek F. Wong, University of Macau; (4) Longyue Wang, Tencent AI Lab, and corresponding author. Authors: Authors: (1) Jianhui Pang, from the University of Macau, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (nlp2ct.pangjh3@gmail.com); (2) Fanghua Ye, University College London, and work was done when Jianhui Pang and Fanghua Ye were interning at Tencent AI Lab (fanghua.ye.19@ucl.ac.uk); (3) Derek F. Wong, University of Macau; (4) Longyue Wang, Tencent AI Lab, and corresponding author. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Related Work 2 Related Work 3 Anchor-based Large Language Models 3.1 Background 3.1 Background 3.2 Anchor-based Self-Attention Networks 3.2 Anchor-based Self-Attention Networks 3.3 Anchor-based Inference 3.3 Anchor-based Inference 4 Experiments and 4.1 Our Implementation 4 Experiments and 4.1 Our Implementation 4.2 Data and Training Procedure 4.2 Data and Training Procedure 4.3 Evaluation 4.3 Evaluation 5 Results 5 Results 6 Analysis 6 Analysis 7 Conclusion, Limitations, Ethics Statement, and References 7 Conclusion, Limitations, Ethics Statement, and References A More Experimental Results A More Experimental Results B Data Settings B Data Settings 3.1 Background Transformers . LLMs are primarily realized as decoder-only transformers (Vaswani et al., 2017; Touvron et al., 2023a,b), incorporating an input embedding layer and multiple decoder layers. Each layer contains a self-attention network and a feedforward network with normalization modules. Crucially, causal attention masks are employed, allowing tokens to attend only to preceding ones. Transformers Self-Attention Networks. Typically for decoderonly LLMs like Llama2 (Touvron et al., 2023b), self-attention networks (SANs) map queries Q , keys K , and values V into an output, as delineated in the following equations, Self-Attention Networks. Q K V where M denotes an L × L masking matrix, facilitating the current i -th token to attend to only preceding tokens whilst disregarding subsequent tokens during the training and inference phases. M L L i Keys/Values Caches . In the application of LLMs, the keys/values caches increase with lengthy prefix texts and continuously generated tokens during the inference phase, such as in question-answering (Saad-Falcon et al., 2023), text summarization (Basyal and Sanghvi, 2023), and machine translation (Pang et al., 2024). The key and value matrices associated with tokens of prefix inputs are cached to avoid recomputation and expedite subsequent token prediction (Radford et al., 2019). Additionally, the model generates the output token-by-token in the real-time inference process, which requires more cache memory to store the newly generated sequence. Therefore, addressing the challenges arising from the ever-expanding texts is crucial for enhancing the efficiency of LLM inference. Keys/Values Caches This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

Unlocking the Mechanics of Decoder-Only Transformers and Self-Attention

In today’s lightning-fast digital landscape, website speed is no longer a luxury — it’s a fundamental requirement. Every developer should possess the knowledge to analyze and optimize web page performance for a seamless user experience. After all, a speedy website translates into higher engagement, lower bounce rates, and ultimately, increased conversions. The High Cost of Slow Websites The detrimental effects of sluggish websites are well-documented by numerous studies: Upward of 25% of users abandon a site if it takes longer than 4 seconds to load Akami Study.
A 3-second delay can lead to a significant drop in engagement, with 22% fewer page views and a 50% higher bounce rate Strangeloop Networks Study.
Conversions take a major hit for websites taking 5 seconds to load, experiencing a decrease of 38% Strangeloop Networks Study.
Pinterest increased search engine traffic and sign-ups by 15% when they reduced perceived wait times by 40%.
The BBC found they lost an additional 10% of users for every extra second their site took to load. Prioritizing Core Web Vitals Forget outdated metrics — Google prioritizes Core Web Vitals for website performance evaluation. These metrics measure real-world user experience and directly impact search engine rankings. Here’s a breakdown of the three key Core Web Vitals: Largest Contentful Paint (LCP): This tracks the time it takes for the largest content element to load. Optimize images and preload content to improve LCP (ideally under 2.5 seconds). Learn more about LCP
Interaction to Next Paint (INP): This metric measures the user’s perceived responsiveness. Aim for an INP of under 50 milliseconds. Learn more about INP
Cumulative Layout Shift (CLS): This metric assesses how much your page layout shifts as elements load. Use pre-defined dimensions for images and avoid lazy loading critical content to minimize CLS (ideally below a score of 0.1). Learn more about CLS Optimizing for Interactivity Beyond loading speed, interactivity matters. Here’s how to ensure your page feels responsive: Time to Interactive (TTI): This measures the time it takes for your page to become fully interactive. Reduce unnecessary JavaScript and optimize critical rendering paths to achieve a TTI under 3.1 seconds. Learn more about TTI
Total Blocking Time (TBT): This metric focuses on how long your main thread is blocked by JavaScript execution. Minimize render-blocking JavaScript and leverage code splitting to keep TBT below 3.1 seconds. Learn more about TBT Actionable Steps for Improvement Leverage a CDN: Consider a content delivery network (CDN) to improve content delivery speed for geographically dispersed users. Monitor CDN performance, including cache hit rate and first-byte time. Remember to carefully consider the Time-to-Live (TTL) of your content. A longer TTL can improve performance by reducing the number of requests to your origin server, but it can also lead to stale content if not managed properly.
Minify and Optimize Resources: Reduce file sizes and optimize images for web delivery.
Implement Lazy Loading: Load non-critical content below the fold only when the user scrolls down to improve initial page load.
Utilize Browser Caching: Enable browser caching for static assets to reduce server requests on subsequent visits. Other Considerations While Core Web Vitals and interactivity metrics provide a solid foundation, there are other factors to consider for comprehensive website performance optimization: Network Performance: Although not directly measured by Lighthouse, network response times significantly impact user experience. Tools like Google PageSpeed Insights can help identify network bottlenecks.
Server-Side Optimization: Optimizing server response times and resource processing can significantly improve perceived website performance. Continuous Monitoring and Improvement Remember, website performance is an ongoing process. Regularly monitor your website’s performance metrics using tools like Google PageSpeed Insights and Lighthouse. Continuously analyze and optimize your code, content, and infrastructure to ensure a top-notch user experience. In today’s lightning-fast digital landscape, website speed is no longer a luxury — it’s a fundamental requirement. Every developer should possess the knowledge to analyze and optimize web page performance for a seamless user experience. After all, a speedy website translates into higher engagement, lower bounce rates, and ultimately, increased conversions. The High Cost of Slow Websites The detrimental effects of sluggish websites are well-documented by numerous studies: Upward of 25% of users abandon a site if it takes longer than 4 seconds to load Akami Study. A 3-second delay can lead to a significant drop in engagement, with 22% fewer page views and a 50% higher bounce rate Strangeloop Networks Study. Conversions take a major hit for websites taking 5 seconds to load, experiencing a decrease of 38% Strangeloop Networks Study. Pinterest increased search engine traffic and sign-ups by 15% when they reduced perceived wait times by 40%. The BBC found they lost an additional 10% of users for every extra second their site took to load. Upward of 25% of users abandon a site if it takes longer than 4 seconds to load Akami Study . Akami Study A 3-second delay can lead to a significant drop in engagement, with 22% fewer page views and a 50% higher bounce rate Strangeloop Networks Study . Strangeloop Networks Study Conversions take a major hit for websites taking 5 seconds to load, experiencing a decrease of 38% Strangeloop Networks Study . Strangeloop Networks Study Pinterest increased search engine traffic and sign-ups by 15% when they reduced perceived wait times by 40%. Pinterest increased search engine traffic and sign-ups by 15% The BBC found they lost an additional 10% of users for every extra second their site took to load. The BBC found they lost an additional 10% of users Prioritizing Core Web Vitals Forget outdated metrics — Google prioritizes Core Web Vitals for website performance evaluation. These metrics measure real-world user experience and directly impact search engine rankings. Here’s a breakdown of the three key Core Web Vitals: Core Web Vitals Largest Contentful Paint (LCP): This tracks the time it takes for the largest content element to load. Optimize images and preload content to improve LCP (ideally under 2.5 seconds). Learn more about LCP Interaction to Next Paint (INP): This metric measures the user’s perceived responsiveness. Aim for an INP of under 50 milliseconds. Learn more about INP Cumulative Layout Shift (CLS): This metric assesses how much your page layout shifts as elements load. Use pre-defined dimensions for images and avoid lazy loading critical content to minimize CLS (ideally below a score of 0.1). Learn more about CLS Largest Contentful Paint (LCP): This tracks the time it takes for the largest content element to load. Optimize images and preload content to improve LCP (ideally under 2.5 seconds). Learn more about LCP Largest Contentful Paint (LCP): Learn more about LCP Interaction to Next Paint (INP): This metric measures the user’s perceived responsiveness. Aim for an INP of under 50 milliseconds. Learn more about INP Interaction to Next Paint (INP): Learn more about INP Cumulative Layout Shift (CLS): This metric assesses how much your page layout shifts as elements load. Use pre-defined dimensions for images and avoid lazy loading critical content to minimize CLS (ideally below a score of 0.1). Learn more about CLS Cumulative Layout Shift (CLS): Learn more about CLS Optimizing for Interactivity Beyond loading speed, interactivity matters. Here’s how to ensure your page feels responsive: Time to Interactive (TTI): This measures the time it takes for your page to become fully interactive. Reduce unnecessary JavaScript and optimize critical rendering paths to achieve a TTI under 3.1 seconds. Learn more about TTI Total Blocking Time (TBT): This metric focuses on how long your main thread is blocked by JavaScript execution. Minimize render-blocking JavaScript and leverage code splitting to keep TBT below 3.1 seconds. Learn more about TBT Time to Interactive (TTI): This measures the time it takes for your page to become fully interactive. Reduce unnecessary JavaScript and optimize critical rendering paths to achieve a TTI under 3.1 seconds. Learn more about TTI Time to Interactive (TTI): Learn more about TTI Total Blocking Time (TBT): This metric focuses on how long your main thread is blocked by JavaScript execution. Minimize render-blocking JavaScript and leverage code splitting to keep TBT below 3.1 seconds. Learn more about TBT Total Blocking Time (TBT): Learn more about TBT Actionable Steps for Improvement Leverage a CDN: Consider a content delivery network (CDN) to improve content delivery speed for geographically dispersed users. Monitor CDN performance, including cache hit rate and first-byte time. Remember to carefully consider the Time-to-Live (TTL) of your content. A longer TTL can improve performance by reducing the number of requests to your origin server, but it can also lead to stale content if not managed properly. Minify and Optimize Resources: Reduce file sizes and optimize images for web delivery. Implement Lazy Loading: Load non-critical content below the fold only when the user scrolls down to improve initial page load. Utilize Browser Caching: Enable browser caching for static assets to reduce server requests on subsequent visits. Leverage a CDN: Consider a content delivery network (CDN) to improve content delivery speed for geographically dispersed users. Monitor CDN performance, including cache hit rate and first-byte time. Remember to carefully consider the Time-to-Live (TTL) of your content. A longer TTL can improve performance by reducing the number of requests to your origin server, but it can also lead to stale content if not managed properly. Leverage a CDN: Minify and Optimize Resources: Reduce file sizes and optimize images for web delivery. Minify and Optimize Resources: Implement Lazy Loading: Load non-critical content below the fold only when the user scrolls down to improve initial page load. Implement Lazy Loading: Utilize Browser Caching: Enable browser caching for static assets to reduce server requests on subsequent visits. Utilize Browser Caching: Other Considerations While Core Web Vitals and interactivity metrics provide a solid foundation, there are other factors to consider for comprehensive website performance optimization: Network Performance: Although not directly measured by Lighthouse, network response times significantly impact user experience. Tools like Google PageSpeed Insights can help identify network bottlenecks. Server-Side Optimization: Optimizing server response times and resource processing can significantly improve perceived website performance. Network Performance: Although not directly measured by Lighthouse, network response times significantly impact user experience. Tools like Google PageSpeed Insights can help identify network bottlenecks. Network Performance: Server-Side Optimization: Optimizing server response times and resource processing can significantly improve perceived website performance. Server-Side Optimization: Continuous Monitoring and Improvement Remember, website performance is an ongoing process. Regularly monitor your website’s performance metrics using tools like Google PageSpeed Insights and Lighthouse. Continuously analyze and optimize your code, content, and infrastructure to ensure a top-notch user experience.

What Metrics Should You Pay Attention to When Optimizing Web Page Performance?

As a content creator and marketer, my journey started in 2016 when I began creating short comedy videos for Instagram. While my early efforts didn’t achieve the viral success I hoped for, these initial experiments set the foundation for a deeper exploration into video content and advertising. Over time, I realized that creating videos wasn't just about making people laugh — it was a research project in itself, a way to study how to captivate attention, optimize viewer retention, and ultimately, market products and services through storytelling. In 2020, I pivoted towards TikTok during the global lockdown, which marked a turning point. My experiment with life hacks quickly gained traction, with my first video reaching half a million views. By the time I had posted 25 videos, I had amassed over 100,000 followers. By staying true to a simple concept, I was able to reach millions of viewers. The format was clear: Hook the viewer within the first few seconds with a statement like "I've always wanted to try this".
Show the hack in action.
Deliver results in a short, snappy video format. Soon, I had over a million followers, and my videos consistently received over a billion views. Around this time, I also began creating content on YouTube, which further expanded my audience. This was when I began receiving offers from brands to incorporate advertising into my content. This marked the beginning of my journey into advertising integrations that aligned with my research into attention management. In this article, I'd like to share some of my favorite case studies from brand integrations that succeeded. Each one required careful thought, creativity, and experience to make the right decisions and achieve the desired results. Without years of trial and error, pulling off these campaigns would have been nearly impossible. Xiaomi Phone Advertisement: Using Voice to Stand Out One of the more interesting projects I worked on was promoting Xiaomi's new phone. Usually, unboxing videos — especially for less popular models — don’t always grab attention without something special. I knew I had to do things differently this time. Instead of my usual no-speech style, I decided to add voice narration for the first time. This caught people off guard since they hadn’t heard me talk in videos before, which made them curious. As a result, the video got over 13 million views. This wasn’t just luck — it was about understanding how to grab attention in the first few seconds and keep people watching. Adele's "Oh My God": Subtle Product Placement Promoting Adele’s new track “Oh My God” was another fun challenge. The idea was to bring attention to the song without making it too obvious. So, I used one of my go-to formats — testing ideas from the internet — but added her track as the background music. This subtle move worked really well because it blended naturally with the content and didn’t feel forced. It helped the video get great engagement, all while pushing the track to a wider audience. It’s a simple idea, but it proves that with a bit of creativity, you can turn product placement into something that feels more like part of the story. KumoSpace: Promoting a Complex Product with Simplicity KumoSpace, a Zoom alternative with cool entertainment features, was tricky to promote. It’s hard to explain a platform like that in a short video. So I had to think of a way to keep people watching without overwhelming them with information. The solution? I used one of their features — the ability to share YouTube videos — and showed a viral clip with over 100 million views. This helped grab attention right from the start, and once people were hooked, I introduced KumoSpace. The video performed well because it didn’t feel like an ad—it felt like content people wanted to watch. Conclusion: The Key to Successful Advertising Through thousands of videos and countless advertising integrations, I've learned that the key to success in both blogging and traditional marketing lies in understanding how to retain audience attention. Whether through humor, unexpected twists, or innovative formats, the first 5-10 seconds are crucial. If viewers stay until the end, the likelihood of success skyrockets. For those in the advertising space, I encourage you to think outside the box. Humor, alternative approaches, and a focus on engagement rather than just views or likes can transform any video into a powerful marketing tool. As my passion for marketing has only grown, I have set a new goal for myself: to open my own media agency and work on advertising projects worldwide. I’m especially excited about the opportunity to work with the Asian market, where I believe there is incredible potential for creative campaigns. I’m curious, what kinds of advertising — whether it’s traditional or digital — really grab your attention? Do you have any favorite ads that you think nailed it, or maybe ones that blew your mind with their creativity? I’d love to hear about any campaigns that stood out to you, whether it’s because they were funny, clever, or just really well executed. Drop your thoughts in the comments — I’m always looking for fresh ideas and inspiration! As a content creator and marketer, my journey started in 2016 when I began creating short comedy videos for Instagram . While my early efforts didn’t achieve the viral success I hoped for, these initial experiments set the foundation for a deeper exploration into video content and advertising. Over time, I realized that creating videos wasn't just about making people laugh — it was a research project in itself, a way to study how to captivate attention, optimize viewer retention, and ultimately, market products and services through storytelling. Instagram In 2020, I pivoted towards TikTok during the global lockdown, which marked a turning point. My experiment with life hacks quickly gained traction, with my first video reaching half a million views. By the time I had posted 25 videos, I had amassed over 100,000 followers. By staying true to a simple concept, I was able to reach millions of viewers. TikTok The format was clear: Hook the viewer within the first few seconds with a statement like "I've always wanted to try this". Show the hack in action. Deliver results in a short, snappy video format. Hook the viewer within the first few seconds with a statement like "I've always wanted to try this". Show the hack in action. Deliver results in a short, snappy video format. Soon, I had over a million followers, and my videos consistently received over a billion views. Around this time, I also began creating content on YouTube , which further expanded my audience. This was when I began receiving offers from brands to incorporate advertising into my content. This marked the beginning of my journey into advertising integrations that aligned with my research into attention management. YouTube In this article, I'd like to share some of my favorite case studies from brand integrations that succeeded. Each one required careful thought, creativity, and experience to make the right decisions and achieve the desired results. Without years of trial and error, pulling off these campaigns would have been nearly impossible. Xiaomi Phone Advertisement: Using Voice to Stand Out Xiaomi Phone Advertisement: Using Voice to Stand Out One of the more interesting projects I worked on was promoting Xiaomi's new phone . Usually, unboxing videos — especially for less popular models — don’t always grab attention without something special. I knew I had to do things differently this time. Instead of my usual no-speech style, I decided to add voice narration for the first time. This caught people off guard since they hadn’t heard me talk in videos before, which made them curious. As a result, the video got over 13 million views. This wasn’t just luck — it was about understanding how to grab attention in the first few seconds and keep people watching. promoting Xiaomi's new phone Adele's "Oh My God": Subtle Product Placement Adele's "Oh My God": Subtle Product Placement Promoting Adele’s new track “Oh My God” was another fun challenge. The idea was to bring attention to the song without making it too obvious. So, I used one of my go-to formats — testing ideas from the internet — but added her track as the background music. This subtle move worked really well because it blended naturally with the content and didn’t feel forced. It helped the video get great engagement, all while pushing the track to a wider audience. It’s a simple idea, but it proves that with a bit of creativity, you can turn product placement into something that feels more like part of the story. Promoting Adele’s new track “Oh My God” KumoSpace: Promoting a Complex Product with Simplicity KumoSpace: Promoting a Complex Product with Simplicity KumoSpace, a Zoom alternative with cool entertainment features, was tricky to promote . It’s hard to explain a platform like that in a short video. So I had to think of a way to keep people watching without overwhelming them with information. The solution? I used one of their features — the ability to share YouTube videos — and showed a viral clip with over 100 million views. This helped grab attention right from the start, and once people were hooked, I introduced KumoSpace. The video performed well because it didn’t feel like an ad—it felt like content people wanted to watch. tricky to promote Conclusion: The Key to Successful Advertising Conclusion: The Key to Successful Advertising Through thousands of videos and countless advertising integrations, I've learned that the key to success in both blogging and traditional marketing lies in understanding how to retain audience attention. Whether through humor, unexpected twists, or innovative formats, the first 5-10 seconds are crucial. If viewers stay until the end, the likelihood of success skyrockets. For those in the advertising space, I encourage you to think outside the box. Humor, alternative approaches, and a focus on engagement rather than just views or likes can transform any video into a powerful marketing tool. As my passion for marketing has only grown, I have set a new goal for myself: to open my own media agency and work on advertising projects worldwide. I’m especially excited about the opportunity to work with the Asian market, where I believe there is incredible potential for creative campaigns. I’m curious, what kinds of advertising — whether it’s traditional or digital — really grab your attention? Do you have any favorite ads that you think nailed it, or maybe ones that blew your mind with their creativity? I’d love to hear about any campaigns that stood out to you, whether it’s because they were funny, clever, or just really well executed. Drop your thoughts in the comments — I’m always looking for fresh ideas and inspiration!

Advertising Integrations That "Go Viral": my Journey in Attention Management and Marketing Research

Integraciones publicitarias que se vuelven virales: mi recorrido por la gestión de la atención y la investigación de mercados

שילובי פרסום ש"הופכים לוויראליים": המסע שלי בניהול קשב ומחקר שיווקי

「バイラルになる」広告統合: 注目度管理とマーケティング調査の旅

"바이럴"이 되는 광고 통합: 주의 관리 및 마케팅 조사에서의 나의 여정

Ba Intégrations Publicités Oyo "Ekende Viral": Voyage na ngai na Gestion ya Attention na Recherche ya Marketing

Fifandraisana amin'ny dokam-barotra "Mandeha Viral": ny diako amin'ny fitantanana ny saina sy ny fikarohana ara-barotra

Рекламни интеграции што се „вирални“: моето патување во управувањето со вниманието и маркетинг истражувањето

"Вирус" болж буй зар сурталчилгааны интеграцчлал: Анхаарлын менежмент ба маркетингийн судалгааны миний аялал

د اعلاناتو ادغام چې "وایرس ته ځي": زما سفر د پاملرنې مدیریت او بازار موندنې څیړنې کې

"வைரலாகும்" விளம்பர ஒருங்கிணைப்புகள்: கவனம் மேலாண்மை மற்றும் சந்தைப்படுத்தல் ஆராய்ச்சியில் எனது பயணம்

Рекламні інтеграції, які стають вірусними: моя подорож у сфері управління увагою та маркетингових досліджень

Ukuhlanganiswa Kokukhangisa Okuthi "Go Viral": Uhambo lwami Ekuphathweni Kokunaka kanye Nocwaningo Lwezokumaketha

Authors:
(1) Lenore Blum (lblum@cs.cmu.edu);
(2) Manuel Blum (mblum@cs.cmu.edu). Table of Links Abstract and 1 Introduction 2 Brief Overview of CtmR, a Robot with a CTM Brain 2.1 Formal Definition of CtmR 2.2 Conscious Attention in CtmR 2.3 Conscious Awareness and the Feeling of Consciousness in CtmR 2.4 CtmR as a Framework for Artificial General Intelligence (AGI) 3 Alignment of CtmR with Other Theories of Consciousness 4 Addressing Kevin Mitchell’s questions from the perspective of CtmR 5 Summary and Conclusions 6 Acknowledgements 7 Appendix 7.1 A Brief History of the Theoretical Computer Science Approach to Computation 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it References 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it We tried to make the CtmR Up-Tree competition deterministic, but it turns out to necessarily be probabilistic. This is because any deterministic competition must be made increasingly complex to realize essential properties. For example, this is the case if we want chunks of near equal |weight| to have near equal chances of getting onto the STM stage. Consider a deterministic competition which makes decisions based on a chunks weight. Suppose chunk A is pinned to weight 11, chunk B to weight 9, and all other chunks to weight 0. In the deterministic CtmR competition, A always wins and B never does. And as long as weights don’t change, CtmR remains totally unconscious of chunk B. In the probabilistic CtmR competition described below (with the parameter d = 0), chunk A wins with probability 11/20 while B win with probability 9/20. So A and B each have roughly equal probability of winning a competition, and a new independent competition is begun at every clock tick. In that case, CtmR will likely become conscious of both. In the probabilistic CtmR competition (a variant of the standard tennis or chess tournament) it is proved that a chunk wins the competition with probability proportional to (a function of) its weight. As a consequence, the winning chunk is independent of where processors are located! This is a property of CtmR’s probabilistic competition. (It is a property that would be difficult if not impossible to achieve in tennis and chess tournaments.) We now describe the probabilistic competition. First recall that a chunk is a tuple, <pointer, time, gist, weight, auxiliary information> consisting of a pointer to the originating processor, the time the chunk was put into the competition, a succinct Brainish gist of information, a valenced weight (to indicate the importance/ value/ confidence the originating processor assigns its gist), and some auxiliary information. For the probabilistic CtmR, the auxiliary information is a pair of numbers which we call (intensity, mood). At the start of the competition, each LTM processor enters a chunk into its leaf node with intensity = |weight| and mood = weight. In the probabilistic competition, each non-leaf node of the Up-Tree contains a coin-toss neuron. The coin-toss neuron probabilistically chooses the local winner of the two (competing) chunks in the node’s two children based on their f values, where f(chunk) = intensity + d•(mood) and -1 ≤ d ≤ +1. Here d is a constant called CtmR’s disposition. If C1 and C2 are the two competing chunks, then the coin toss neuron will choose Ci to be the local winner with probability f(Ci)/f(C1)+f(C2). The local winner will move up one level in a single clock tick. The first four parameters of this new chunk are the same as the local winner’s. But its intensity is the sum of two competing chunks’ intensities. Similarly for its mood. Thus, as a chunk moves up the tree, the intensity never decreases. This is not (necessarily) the case for the mood. In this way, the winning chunk’s auxiliary information at the end of the competition will contain the sum of all submitted chunks’ intensities (|weights|) and the sum of all submitted moods (weights). If the disposition is d > 0, CtmR will be “upbeat” in the sense that positively valenced chunks will have a higher probability of winning than negatively weighted chunks of the same |weight|. If its disposition is d = +1, CtmR is manic: only positively valenced chunks can win the competition. The CtmR knows only what is positive in its life, as long as anything is positive. If the disposition is d < 0, CtmR will be “downbeat”. If d = -1, CtmR is “hopelessly depressed”, only negatively valenced chunks can win the competition. There is no way out of this horrible state except with a reboot, i.e., to “shock” the system to get a less extreme disposition.[35] This paper is available on arxiv under CC BY 4.0 DEED license. [35] If d=+1, CtmR is in the manic state, and CtmR theory suggests that there too, a reboot is warranted. In humans, electroconvulsive therapy (ECT) is used primarily for depression. The CtmR model suggests that ECT might work as well for mania. Despite the fact that “ECT is a rapid and highly effective treatment of manic episodes”, current guidelines only endorse ECT “for pharmacotherapy-resistant mania but often as second- or third-line treatment.” (Elias, Thomas, & Sackeim, 2021). Authors: (1) Lenore Blum (lblum@cs.cmu.edu); (2) Manuel Blum (mblum@cs.cmu.edu). Authors: Authors: (1) Lenore Blum (lblum@cs.cmu.edu); (2) Manuel Blum (mblum@cs.cmu.edu). Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Brief Overview of CtmR, a Robot with a CTM Brain 2.1 Formal Definition of CtmR 2.1 Formal Definition of CtmR 2.2 Conscious Attention in CtmR 2.2 Conscious Attention in CtmR 2.3 Conscious Awareness and the Feeling of Consciousness in CtmR 2.3 Conscious Awareness and the Feeling of Consciousness in CtmR 2.4 CtmR as a Framework for Artificial General Intelligence (AGI) 2.4 CtmR as a Framework for Artificial General Intelligence (AGI) 3 Alignment of CtmR with Other Theories of Consciousness 3 Alignment of CtmR with Other Theories of Consciousness 4 Addressing Kevin Mitchell’s questions from the perspective of CtmR 4 Addressing Kevin Mitchell’s questions from the perspective of CtmR 5 Summary and Conclusions 5 Summary and Conclusions 6 Acknowledgements 6 Acknowledgements 7 Appendix 7.1 A Brief History of the Theoretical Computer Science Approach to Computation 7.1 A Brief History of the Theoretical Computer Science Approach to Computation 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it References References 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it We tried to make the CtmR Up-Tree competition deterministic, but it turns out to necessarily be probabilistic. This is because any deterministic competition must be made increasingly complex to realize essential properties. For example, this is the case if we want chunks of near equal |weight| to have near equal chances of getting onto the STM stage. Consider a deterministic competition which makes decisions based on a chunks weight. Suppose chunk A is pinned to weight 11, chunk B to weight 9, and all other chunks to weight 0. In the deterministic CtmR competition, A always wins and B never does. And as long as weights don’t change, CtmR remains totally unconscious of chunk B. In the probabilistic CtmR competition described below (with the parameter d = 0), chunk A wins with probability 11/20 while B win with probability 9/20. So A and B each have roughly equal probability of winning a competition, and a new independent competition is begun at every clock tick. In that case, CtmR will likely become conscious of both. In the probabilistic CtmR competition (a variant of the standard tennis or chess tournament) it is proved that a chunk wins the competition with probability proportional to (a function of) its weight. As a consequence, the winning chunk is independent of where processors are located! This is a property of CtmR’s probabilistic competition. (It is a property that would be difficult if not impossible to achieve in tennis and chess tournaments.) We now describe the probabilistic competition. First recall that a chunk is a tuple, <pointer, time, gist, weight, auxiliary information> consisting of a pointer to the originating processor, the time the chunk was put into the competition, a succinct Brainish gist of information, a valenced weight (to indicate the importance/ value/ confidence the originating processor assigns its gist), and some auxiliary information. For the probabilistic CtmR, the auxiliary information is a pair of numbers which we call (intensity, mood) . auxiliary information (intensity, mood) At the start of the competition , each LTM processor enters a chunk into its leaf node with At the start of the competition intensity = |weight| and mood = weight. intensity = |weight| mood = weight. In the probabilistic competition, each non-leaf node of the Up-Tree contains a coin-toss neuron. coin-toss neuron. The coin-toss neuron probabilistically chooses the local winne r of the two (competing) chunks in the node’s two children based on their f values, where local winne f f(chunk) = intensity + d•(mood) and -1 ≤ d ≤ +1. f(chunk) = intensity + d•(mood) and -1 ≤ d ≤ +1. Here d is a constant called CtmR’s disposition . d disposition If C1 and C2 are the two competing chunks, then the coin toss neuron will choose Ci to be the local winner with probability f(Ci)/f(C1)+f(C2). The local winner will move up one level in a single clock tick. The first four parameters of this new chunk are the same as the local winner’s. But its intensity is the sum of two competing chunks’ intensities. Similarly for its mood. Thus, as a chunk moves up the tree, the intensity never decreases. This is not (necessarily) the case for the mood. In this way, the winning chunk’s auxiliary information at the end of the competition will contain the sum of all submitted chunks’ intensities (|weights|) and the sum of all submitted moods (weights). If the disposition is d > 0 , CtmR will be “upbeat” in the sense that positively valenced chunks will have a higher probability of winning than negatively weighted chunks of the same |weight|. If its disposition is d = +1 , CtmR is manic: only positively valenced chunks can win the competition. The CtmR knows only what is positive in its life, as long as anything is positive. d > 0 d = +1 If the disposition is d < 0 , CtmR will be “downbeat”. If d = -1 , CtmR is “hopelessly depressed”, only negatively valenced chunks can win the competition. There is no way out of this horrible state except with a reboot, i.e., to “shock” the system to get a less extreme disposition.[35] d < 0 d = -1 This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv [35] If d=+1 , CtmR is in the manic state, and CtmR theory suggests that there too, a reboot is warranted. d=+1 In humans, electroconvulsive therapy (ECT) is used primarily for depression. The CtmR model suggests that ECT might work as well for mania. Despite the fact that “ECT is a rapid and highly effective treatment of manic episodes”, current guidelines only endorse ECT “for pharmacotherapy-resistant mania but often as second- or third-line treatment.” (Elias, Thomas, & Sackeim, 2021).

The Probabilistic Competition for Conscious Attention and the Influence of Disposition on It

Authors:
(1) Lenore Blum (lblum@cs.cmu.edu);
(2) Manuel Blum (mblum@cs.cmu.edu). Table of Links Abstract and 1 Introduction 2 Brief Overview of CtmR, a Robot with a CTM Brain 2.1 Formal Definition of CtmR 2.2 Conscious Attention in CtmR 2.3 Conscious Awareness and the Feeling of Consciousness in CtmR 2.4 CtmR as a Framework for Artificial General Intelligence (AGI) 3 Alignment of CtmR with Other Theories of Consciousness 4 Addressing Kevin Mitchell’s questions from the perspective of CtmR 5 Summary and Conclusions 6 Acknowledgements 7 Appendix 7.1 A Brief History of the Theoretical Computer Science Approach to Computation 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it References 2.2 Conscious Attention in CtmR Formal definition 1. Conscious attention in CtmR is the reception by all LTM processors of the broadcast of CtmR’s current conscious content. [KM2] [KM5] In other words, CtmR pays conscious attention at time t+h+1 to the winner of the competition that commenced at time t.[15] LTM processors make predictions and get feedback from CtmR’s inner and outer worlds. Based on this feedback, learning algorithms internal to each processor improve that processor’s behavior. These learning algorithms include each processor’s built-in Sleeping Experts Learning algorithm that helps the processor adjust weights it gives its gist. See (Blum A. , 1995) and (Blum, Hopcroft, & Kannan, 2015). [KM7] Thus, we can already see some basic predictive dynamics (prediction + testing + feedback + learning/correction) occurring within CtmR. [KM2] In time, some LTM processors become connected via bi-directional links.[16] Such links enable conscious communication, i.e. communication that goes through STM, to be replaced by more direct unconscious communication through links. Thus, when CtmR initially learns to ride a bike, most communication is done consciously until relevant processor links have formed. Then, for the most part, riding a bike is done unconsciously until an obstacle is encountered, forcing CtmR to pay conscious attention again. [KM6] CtmR’s competition, broadcast, attention and immediate direct communication via links is reminiscent of a process that Dehaene and Jean-Pierre Changeux call ignition (Dehaene & Changeux, 2005). But for feelings of consciousness, attention is not all you need. More is required. This paper is available on arxiv under CC BY 4.0 DEED license. [15] Thus, there is a (h+1) delay between the time the chunks are entered into the competition by the “unconscious” LTM processors and when CtmR pays conscious attention to the wining chunk. This is somewhat analogous to behavioral and brain studies going back to (Libet, 1985). [16] The CtmR has no links at birth. In our earlier “What’s her name?” scenario, when processor p sees that processors p’’ and p’’’ have useful information, p may form a bi-directional link with p’’ and perhaps also with p’’’. Authors: (1) Lenore Blum (lblum@cs.cmu.edu); (2) Manuel Blum (mblum@cs.cmu.edu). Authors: Authors: (1) Lenore Blum (lblum@cs.cmu.edu); (2) Manuel Blum (mblum@cs.cmu.edu). Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Brief Overview of CtmR, a Robot with a CTM Brain 2.1 Formal Definition of CtmR 2.1 Formal Definition of CtmR 2.2 Conscious Attention in CtmR 2.2 Conscious Attention in CtmR 2.3 Conscious Awareness and the Feeling of Consciousness in CtmR 2.3 Conscious Awareness and the Feeling of Consciousness in CtmR 2.4 CtmR as a Framework for Artificial General Intelligence (AGI) 2.4 CtmR as a Framework for Artificial General Intelligence (AGI) 3 Alignment of CtmR with Other Theories of Consciousness 3 Alignment of CtmR with Other Theories of Consciousness 4 Addressing Kevin Mitchell’s questions from the perspective of CtmR 4 Addressing Kevin Mitchell’s questions from the perspective of CtmR 5 Summary and Conclusions 5 Summary and Conclusions 6 Acknowledgements 6 Acknowledgements 7 Appendix 7.1 A Brief History of the Theoretical Computer Science Approach to Computation 7.1 A Brief History of the Theoretical Computer Science Approach to Computation 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it 7.2 The Probabilistic Competition for Conscious Attention and the Influence of Disposition on it References References 2.2 Conscious Attention in CtmR Formal definition 1. Conscious attention in CtmR is the reception by all LTM processors of the broadcast of CtmR’s current conscious content. [KM2] [KM5] Formal definition 1. In other words, CtmR pays conscious attention at time t+h+1 to the winner of the competition that commenced at time t.[15] LTM processors make predictions and get feedback from CtmR’s inner and outer worlds. Based on this feedback, learning algorithms internal to each processor improve that processor’s behavior. These learning algorithms include each processor’s built-in Sleeping Experts Learning algorithm that helps the processor adjust weights it gives its gist. See (Blum A. , 1995) and (Blum, Hopcroft, & Kannan, 2015). [KM7] Thus, we can already see some basic predictive dynamics (prediction + testing + feedback + learning/correction) occurring within CtmR. [KM2] In time, some LTM processors become connected via bi-directional links .[16] Such links enable conscious communication, i.e. communication that goes through STM, to be replaced by more direct unconscious communication through links. Thus, when CtmR initially learns to ride a bike, most communication is done consciously until relevant processor links have formed. Then, for the most part, riding a bike is done unconsciously until an obstacle is encountered, forcing CtmR to pay conscious attention again. [KM6] links CtmR’s competition, broadcast, attention and immediate direct communication via links is reminiscent of a process that Dehaene and Jean-Pierre Changeux call ignition (Dehaene & Changeux, 2005). But for feelings of consciousness, attention is not all you need. More is required. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv [15] Thus, there is a (h+1) delay between the time the chunks are entered into the competition by the “unconscious” LTM processors and when CtmR pays conscious attention to the wining chunk. This is somewhat analogous to behavioral and brain studies going back to (Libet, 1985). [16] The CtmR has no links at birth. In our earlier “What’s her name?” scenario, when processor p sees that processors p’’ and p’’’ have useful information, p may form a bi-directional link with p’’ and perhaps also with p’’’.

Conscious Attention in Conscious Turing Machine Robots

The Noonification: Transformers: Age of Attention (8/27/2024)

Transformers: Age of Attention

ট্রান্সফরমার: মনোযোগের বয়স

Transformers: Zeitalter der Aufmerksamkeit

Transformers: La era de la atención

Transformers : L'Âge de l'Attention

ट्रांसफॉर्मर्स: एज ऑफ अटेंशन

トランスフォーマー：注目の時代

트랜스포머: 주의의 시대

Transformers: Era da Atenção

Трансформеры: Эпоха внимания

Transformers: Dikkat Çağı

Transformers: Thời đại của sự chú ý

变形金刚：注意力时代

How Memecoins Are Redefining Crypto Culture and the Attention Economy I was one of the few females in the early days of cryptocurrency, and the culture was as wild and untamed as the markets themselves. VIP events catering to "whales," the term for big-money investors, often took place in strip clubs, where flashy displays of wealth were the norm. The focus was on speculation and excess, with an almost frenzied atmosphere fueled by the promise of quick riches. The notorious "strippers and Lambos" image became synonymous with crypto, painting a picture of a subculture obsessed with instant gratification and conspicuous consumption. But as the cryptocurrency space has matured, so too has its culture. The rise of memecoins—digital tokens created initially as jokes or internet memes—has marked a significant shift, not just in how value is perceived and exchanged, but in the very ethos of the industry. While still a playground for speculative excess and troll humor, some tokens have migrated towards loftier goals and are now evolving into a platform for meaningful engagement and social change. The memecoin transformation into PolitiFi by more serious politically driven projects embodies a set of beliefs and ideals and attracts investors with those same beliefs and ideals. Nowhere was this transformation more evident than at this year’s #Bitcoin2024, which I attended as official media. The exclusive invite-only event hosted by MAGAA (Make America Great Again Again) was the most surprising. Unlike the boisterous, male-dominated gatherings of the past, this event was held at a high-end rooftop bar, where the atmosphere was one of sophistication and purpose. Guests were treated to exquisite food, craft cocktails, and a lineup of exceptional speakers, from David Bailey, the founder of the Bitcoin Conference, Don Trump Jr, and Tucker Carlson, all reflecting crypto's new, more refined image and meme culture’s ability to attract and inspire and audience. Perhaps most striking was the diversity of the attendees. Nearly half of the crowd were women—industry professionals shaping blockchain technology's future. Gone were the "booth babes" and escorts who once symbolized the male-centric culture of early crypto. In their place were leaders, innovators, and changemakers, contributing to discussions that ranged from the commoditization of attention via memecoins to the broader societal implications of blockchain, privacy, and regulation. Speaking with some of the largest MAGAA holders, a universal theme emerged: the desire to create an uncensorable mechanism to fuel grassroots activism. This shift is emblematic of the changing narrative around cryptocurrency. The initial hype, driven by the allure of easy money and extravagant lifestyles, has given way to a more nuanced understanding of what blockchain technology can achieve. Memecoins, once dismissed as mere jokes, are now at the forefront of this new narrative. They represent a digital currency that thrives on attention, turning likes, shares, and viral content into tangible assets. But beyond their economic potential, memecoins are being harnessed for social good, funding political movements, and enabling grassroots activism in ways traditional financial systems cannot.  **This inspiring use case for memecoins is a testament to the potential of cryptocurrency to bring about positive change. Mike Miglio, the co-founder of MAGAA, is one of the pioneers of this new wave. As a blockchain attorney, Miglio has witnessed the evolution of crypto culture firsthand. "The early days were about capturing attention at any cost, often in ways that were more about spectacle than substance," he says. "But now, we're seeing a shift. Attention is still the currency, but it's being used to create real value—value that can drive social change and empower communities." The transition from strip clubs to rooftop bars is more than just a change of venue. It signifies a broader transformation in the industry's values and objectives. Where once the goal was to flaunt wealth, now it is to wield mainstream influence. Where the culture was once exclusive, catering to a select few, it is now inclusive, inviting diverse voices to the table. And where the focus was once on short-term gains, it is now on long-term impact through steady growth. This evolution reflects the maturing of the crypto industry as it moves away from its hedonistic beginnings and toward a future where digital assets are not just financial tools but instruments for building a better world. Memecoins, with their unique ability to commoditize attention, are leading this charge, proving that in the digital age, we value not only money but also the ability to create change. As the industry continues to grow, the stories we tell about it will evolve as well. The crypto narrative is no longer confined to the old tropes of wealth and excess. As blockchain technology penetrates more deeply into the mainstream, it is becoming a story about empowerment, innovation, and the redefinition of value itself. And in this new era, the currency of attention is proving to be as powerful as any financial asset. How Memecoins Are Redefining Crypto Culture and the Attention Economy I was one of the few females in the early days of cryptocurrency, and the culture was as wild and untamed as the markets themselves. VIP events catering to "whales," the term for big-money investors, often took place in strip clubs, where flashy displays of wealth were the norm. The focus was on speculation and excess, with an almost frenzied atmosphere fueled by the promise of quick riches. The notorious " strippers and Lambos " image became synonymous with crypto, painting a picture of a subculture obsessed with instant gratification and conspicuous consumption. strippers and Lambos But as the cryptocurrency space has matured, so too has its culture. The rise of memecoins—digital tokens created initially as jokes or internet memes—has marked a significant shift, not just in how value is perceived and exchanged, but in the very ethos of the industry. While still a playground for speculative excess and troll humor, some tokens have migrated towards loftier goals and are now evolving into a platform for meaningful engagement and social change. The memecoin transformation into PolitiFi by more serious politically driven projects embodies a set of beliefs and ideals and attracts investors with those same beliefs and ideals. Nowhere was this transformation more evident than at this year’s #Bitcoin2024, which I attended as official media. The exclusive invite-only event hosted by MAGAA (Make America Great Again Again) was the most surprising. Unlike the boisterous, male-dominated gatherings of the past, this event was held at a high-end rooftop bar, where the atmosphere was one of sophistication and purpose. Guests were treated to exquisite food, craft cocktails, and a lineup of exceptional speakers , from David Bailey, the founder of the Bitcoin Conference, Don Trump Jr, and Tucker Carlson, all reflecting crypto's new, more refined image and meme culture’s ability to attract and inspire and audience. exceptional speakers Perhaps most striking was the diversity of the attendees. Nearly half of the crowd were women—industry professionals shaping blockchain technology's future. Gone were the "booth babes" and escorts who once symbolized the male-centric culture of early crypto. In their place were leaders, innovators, and changemakers, contributing to discussions that ranged from the commoditization of attention via memecoins to the broader societal implications of blockchain, privacy, and regulation. Speaking with some of the largest MAGAA holders, a universal theme emerged: the desire to create an uncensorable mechanism to fuel grassroots activism. This shift is emblematic of the changing narrative around cryptocurrency. The initial hype, driven by the allure of easy money and extravagant lifestyles, has given way to a more nuanced understanding of what blockchain technology can achieve. Memecoins, once dismissed as mere jokes, are now at the forefront of this new narrative. They represent a digital currency that thrives on attention, turning likes, shares, and viral content into tangible assets. But beyond their economic potential, memecoins are being harnessed for social good, funding political movements, and enabling grassroots activism in ways traditional financial systems cannot.  **This inspiring use case for memecoins is a testament to the potential of cryptocurrency to bring about positive change. Mike Miglio, the co-founder of MAGAA , is one of the pioneers of this new wave. As a blockchain attorney, Miglio has witnessed the evolution of crypto culture firsthand. "The early days were about capturing attention at any cost, often in ways that were more about spectacle than substance," he says. "But now, we're seeing a shift. Attention is still the currency, but it's being used to create real value—value that can drive social change and empower communities." MAGAA The transition from strip clubs to rooftop bars is more than just a change of venue. It signifies a broader transformation in the industry's values and objectives. Where once the goal was to flaunt wealth, now it is to wield mainstream influence. Where the culture was once exclusive, catering to a select few, it is now inclusive, inviting diverse voices to the table. And where the focus was once on short-term gains, it is now on long-term impact through steady growth. This evolution reflects the maturing of the crypto industry as it moves away from its hedonistic beginnings and toward a future where digital assets are not just financial tools but instruments for building a better world. Memecoins, with their unique ability to commoditize attention, are leading this charge, proving that in the digital age, we value not only money but also the ability to create change. As the industry continues to grow, the stories we tell about it will evolve as well. The crypto narrative is no longer confined to the old tropes of wealth and excess. As blockchain technology penetrates more deeply into the mainstream, it is becoming a story about empowerment, innovation, and the redefinition of value itself. And in this new era, the currency of attention is proving to be as powerful as any financial asset.

The Commoditization of Attention

Authors:
(1) P Aditya Sreekar, Amazon and these authors contributed equally to this work {sreekarp@amazon.com};
(2) Sahil Verm, Amazon and these authors contributed equally to this work {vrsahil@amazon.com;}
(3) Varun Madhavan, Indian Institute of Technology, Kharagpur. Work done during internship at Amazon {varunmadhavan@iitkgp.ac.in};
(4) Abhishek Persad, Amazon {persadap@amazon.com}. Table of Links Abstract and Introduction
Related Works
Methodology
Experiment
Conclusion and Future Work
References 2. Related Works Tree-based algorithms are widely used in machine learning for tabular data. Decision trees recursively split the data into multiple parts based on axis-aligned hyper-planes (Hastie et al., 2009). Random Forests (RF) (Breiman, 2001) and Gradient Boosted Decision Trees (GBDT) (Friedman, 2001) are the most commonly used tree based ensembles. RF fits multiple decision trees on random subsets of the data and averages/polls the predictions to alleviate the overfitting characteristic of decision trees. GBDT, XGBoost (Chen and Guestrin, 2016), and CatBoost (Prokhorenkova et al., 2018) are boosted ensemble models that sequentially build decision trees to correct errors made by previous trees, leading to improved performance on complex datasets with non-linear relations. Recently, there has been a lot of interest in deep learning models for tabular data. Some methods introduce differentiable approximations of decision functions used in decision trees to make them differentiable (Hazimeh et al., 2020; Popov et al., 2019). These methods outperform pure tree based problem for some problem statements, however, they are not consistently better (Gorishniy et al., 2021). Other methods have used attention mechanisms to adapt DL methods to tabular data (Arik et al., 2019; Huang et al., 2020; Gorishniy et al., 2021; Somepalli et al., 2021; Chen et al., 2022). TabNet (Arik et al., 2019) proposes a sparse attention mechanism that is stacked in multiple layers to mimic the recursive splitting of decision trees. Inspired from the success of self-attention transformers (Vaswani et al., 2017) in many domains (Devlin et al., 2019; Dosovitskiy et al., 2021; Gong et al., 2021) methods like TabTransformer (Huang et al., 2020), FT-Transformer (Gorishniy et al., 2021) and SAINT (Somepalli et al., 2021) were proposed. TabTransformer embeds all categorical variables into a unified embedding space, and a sentence of categorical embeddings is passed through self-attention transformer layers. FT-Transformer further extends this by attending to numerical features as well, by using continuous embedding. SAINT builds on FT-Transformer by proposing a new kind of attention which captures interactions between samples of a batch. However, SAINT does not provide any advantage over FT-Transformer for our problem statement, because intersample attention is only effective when the number of dimensions is higher in comparision to the number of samples, thus we do not compare RCT against SAINT (Somepalli et al., 2021). This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. Authors: (1) P Aditya Sreekar, Amazon and these authors contributed equally to this work {sreekarp@amazon.com}; (2) Sahil Verm, Amazon and these authors contributed equally to this work {vrsahil@amazon.com;} (3) Varun Madhavan, Indian Institute of Technology, Kharagpur. Work done during internship at Amazon {varunmadhavan@iitkgp.ac.in}; (4) Abhishek Persad, Amazon {persadap@amazon.com}. Authors: Authors: (1) P Aditya Sreekar, Amazon and these authors contributed equally to this work {sreekarp@amazon.com}; (2) Sahil Verm, Amazon and these authors contributed equally to this work {vrsahil@amazon.com;} (3) Varun Madhavan, Indian Institute of Technology, Kharagpur. Work done during internship at Amazon {varunmadhavan@iitkgp.ac.in}; (4) Abhishek Persad, Amazon {persadap@amazon.com}. Table of Links Abstract and Introduction Related Works Methodology Experiment Conclusion and Future Work References Abstract and Introduction Abstract and Introduction Related Works Related Works Methodology Methodology Experiment Experiment Conclusion and Future Work Conclusion and Future Work References References 2. Related Works Tree-based algorithms are widely used in machine learning for tabular data. Decision trees recursively split the data into multiple parts based on axis-aligned hyper-planes (Hastie et al., 2009). Random Forests (RF) (Breiman, 2001) and Gradient Boosted Decision Trees (GBDT) (Friedman, 2001) are the most commonly used tree based ensembles. RF fits multiple decision trees on random subsets of the data and averages/polls the predictions to alleviate the overfitting characteristic of decision trees. GBDT, XGBoost (Chen and Guestrin, 2016), and CatBoost (Prokhorenkova et al., 2018) are boosted ensemble models that sequentially build decision trees to correct errors made by previous trees, leading to improved performance on complex datasets with non-linear relations. Recently, there has been a lot of interest in deep learning models for tabular data. Some methods introduce differentiable approximations of decision functions used in decision trees to make them differentiable (Hazimeh et al., 2020; Popov et al., 2019). These methods outperform pure tree based problem for some problem statements, however, they are not consistently better (Gorishniy et al., 2021). Other methods have used attention mechanisms to adapt DL methods to tabular data (Arik et al., 2019; Huang et al., 2020; Gorishniy et al., 2021; Somepalli et al., 2021; Chen et al., 2022). TabNet (Arik et al., 2019) proposes a sparse attention mechanism that is stacked in multiple layers to mimic the recursive splitting of decision trees. Inspired from the success of self-attention transformers (Vaswani et al., 2017) in many domains (Devlin et al., 2019; Dosovitskiy et al., 2021; Gong et al., 2021) methods like TabTransformer (Huang et al., 2020), FT-Transformer (Gorishniy et al., 2021) and SAINT (Somepalli et al., 2021) were proposed. TabTransformer embeds all categorical variables into a unified embedding space, and a sentence of categorical embeddings is passed through self-attention transformer layers. FT-Transformer further extends this by attending to numerical features as well, by using continuous embedding. SAINT builds on FT-Transformer by proposing a new kind of attention which captures interactions between samples of a batch. However, SAINT does not provide any advantage over FT-Transformer for our problem statement, because intersample attention is only effective when the number of dimensions is higher in comparision to the number of samples, thus we do not compare RCT against SAINT (Somepalli et al., 2021). This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. available on arxiv

Stories, News & Mentions about ATTENIO

Attenio’s Stories on HackerNoon

Attenio's News & Mentions