Vividly

vividly

In-messenger games, a digital era phenomenon born in Asia and rapidly adopted across the globe, have fundamentally reshaped social interactions and digital economies since their emergence in the early 2010s. These games, integrated into popular instant messaging platforms like WeChat, Kakao, and Telegram, have evolved from simple entertainment to complex ecosystems involving blockchain technology and digital assets. The question now is: Which player will reshape the landscape with decentralized technologies and take the lead in this evolving space? Drawing from my experience at TonTon Games. Asian In-Messenger Games: WeChat vs. Kakao WeChat, developed by Tencent, has positioned itself as a leader in integrating mini-games within its messaging platform since 2017. This initiative has proven highly successful, with WeChat’s mini-games attracting over 400 million monthly active users and generating approximately $2.3 billion in revenue in the first half of 2024 alone. Popular titles include "Jump Jump," known for its simplicity and addictiveness, and "The Pirate Game," a strategy-based offering. Meanwhile, "Xundao Daquian" has shown firm performance, leading Interactive Entertainment rankings with 162.41 million monthly active users in February 2024. WeChat's success can be attributed to the seamless integration of these games, their accessibility, and high user retention rates. The platform's mini-games have become integral to daily user engagement, demonstrating the massive potential of in-messenger gaming. KakaoTalk, founded by Kakao Corp in 2010, dominates the South Korean market with a 93% market penetration. Leveraging this extensive user base, Kakao Games has implemented in-messenger gaming features. Their strategy revolves around utilizing KakaoTalk's social network to enhance player connectivity, driving user engagement and retention. Kakao Games has taken a strategic approach to its growth, forming partnerships with global developers like Ocean Drive Studio, Chrono Studio, Candy Digital, and METABORA. This collaborative strategy has diversified its gaming portfolio, focusing on high-quality, narrative-driven experiences. Key offerings like "Guardian Tales" and "Eversoul" testify to the platform's commitment to quality and innovation, instilling confidence in its future. Innovative Approach to In-Messenger Games: Telegram's Web3 Strategy In the Western market, Telegram has emerged as a frontrunner in integrating blockchain technology within its messaging platform, positioning itself at the forefront of the Web3 movement. As of 2024, Telegram boasts a substantial global user base of 950 million active users, largely driven by the growing popularity of games, creating a significant market for its Web3 initiatives. Telegram's strategy leverages The Open Network (TON) to facilitate secure and rapid transactions within the app, supporting in-game economies and decentralized applications (dApps). This aligns with industry projections from DappRadar, which anticipate the blockchain gaming market to reach a valuation of $39.7 billion by 2025, indicating substantial growth potential and exciting opportunities in this sector. However, as Daniil Shcherbakov, CEO of TonTon Games, cautions, merely porting over successful Web2 games won’t resonate with today’s Web3 audience on Telegram. Shcherbakov explains, “It’s not enough to take a Web2 game, even a successful Web2 game, and put it on Telegram and expect money to grow on trees. There is no audience currently for that. It should have a proper, well-developed Web3 narrative. Without it, people won’t play it, at least the people that are playing now.” This focus on Web3-native gaming mechanics is evident in Telegram's gaming portfolio, which includes titles designed to leverage blockchain integration. For instance, Notcoin is a tap-to-earn game where users can accumulate Notcoin (NOT) through simple gameplay mechanics, while SnakeTON is a survival game that rewards players with cryptocurrency as they navigate obstacles. Similarly, Catizen is a strategic, community-based game allowing players to earn and trade digital assets, offering an experience that appeals to Web3-minded users looking for both entertainment and economic participation. By addressing these demands, Telegram is building an ecosystem that supports not only traditional gaming enjoyment but also deeper engagement with GameFi (Game Finance) elements. Asian Super Apps vs. Telegram: Key Differences Cultural background undoubtedly shapes game design, impacting every step in the process — from narrative themes to character development. Game developers often draw inspiration from their cultural heritage, enriching gameplay and fostering a deeper connection between players through storytelling. Telegram games focus more on individualism, freedom, and open-world exploration, while Asian games often emphasize community, structured progression, and deep storytelling. This approach reflects players' cultural values and preferences in each region, vividly illustrated through game mechanics, aesthetics, and engagement strategies. The adoption of blockchain and Web3 technologies further highlights the divergence between these approaches. WeChat faced significant hurdles in this area due to strict government oversight of blockchain and cryptocurrency activities in China. Kakao, despite joining the Bank of Korea's digital currency project through its subsidiary Ground X, was positioned more as a follower than a leader in blockchain innovation. Telegram's approach differs markedly from that of its Eastern counterparts. The platform enhanced user engagement by integrating blockchain technology via TON, implementing features such as collectible gifts and NFTs, and opening new monetization channels. This proactive approach positioned it as a pioneer in the Web3 space within messaging platforms, yet it also poses some challenges. Potential Challenges for Telegram's Web3 Ecosystem Achieving mass adoption of Web3 technologies, including decentralized applications (dApps) on Telegram, will be critical for the long-term success of its ecosystem. However, widespread adoption demands significant efforts in user education and onboarding, as blockchain technology remains unfamiliar to many users. Ensuring clear, accessible information alongside user-friendly interfaces is essential for driving engagement and building a broader user base within the Web3 environment. Integrating Web3 features within a messaging platform also presents challenges related to interoperability with TON. Maintaining scalability, performance, and security will be essential to support high transaction volumes without compromising user experience. Additionally, Telegram must navigate regulatory compliance carefully, as evolving legal landscapes across jurisdictions could impact the development and deployment of Web3 features. Final Thoughts: The Future of In-Messenger Games and Web3 Integration Blockchain technology will likely reshape the gaming industry in several exciting ways. Virtual worlds and economies will become more interconnected, allowing players to own and trade in-game assets across different games truly. This will create new revenue streams for both developers and players. As in-messenger gaming continues to evolve, a few key technologies are paving the way for broader adoption across global audiences. HTML5 and Cocos are emerging as essential tools that can support mass-market growth and enhance game quality, accessibility, and lightweight functionality — crucial elements for games hosted within messaging platforms. These technologies provide the flexibility and performance needed to deliver responsive, visually rich experiences while keeping the file size low and eliminating the need for complex installations, making them ideal for today’s mobile-first world. The social dynamics within these games are shifting, too. Where early in-messenger games were primarily focused on single-player or light social elements, modern games within platforms like WeChat, Kakao, and Telegram are deeply interactive, with live chat channels and in-game communities allowing hundreds of thousands of players to engage, give real-time feedback, and build a sense of camaraderie. This is a departure from the one-way ratings and reviews seen on app stores, adding another dimension to user experience and engagement. Finally, the integration of blockchain technology on platforms like Telegram brings a new level of innovation through GameFi elements. In-game assets are becoming valuable digital items, tradable and secure through NFTs, while in-game staking and blockchain-based payments offer users new ways to invest, earn, and participate in the ecosystem. This trend echoes a global shift toward decentralized ownership and economies, signaling a new era of in-messenger gaming where players can interact with, own, and trade game assets freely. In-messenger games, a digital era phenomenon born in Asia and rapidly adopted across the globe, have fundamentally reshaped social interactions and digital economies since their emergence in the early 2010s. These games, integrated into popular instant messaging platforms like WeChat, Kakao, and Telegram, have evolved from simple entertainment to complex ecosystems involving blockchain technology and digital assets. The question now is: Which player will reshape the landscape with decentralized technologies and take the lead in this evolving space? Drawing from my experience at TonTon Games. TonTon Games. Asian In-Messenger Games: WeChat vs. Kakao Asian In-Messenger Games: WeChat vs. Kakao WeChat, developed by Tencent, has positioned itself as a leader in integrating mini-games within its messaging platform since 2017. This initiative has proven highly successful, with WeChat’s mini-games attracting over 400 million monthly active users and generating approximately $2.3 billion in revenue in the first half of 2024 alone. 400 million monthly active users 400 million monthly active users $2.3 billion in revenue Popular titles include "Jump Jump," known for its simplicity and addictiveness, and "The Pirate Game," a strategy-based offering. Meanwhile, "Xundao Daquian" has shown firm performance, leading Interactive Entertainment rankings with 162.41 million monthly active users in February 2024. WeChat's success can be attributed to the seamless integration of these games, their accessibility, and high user retention rates. The platform's mini-games have become integral to daily user engagement, demonstrating the massive potential of in-messenger gaming. KakaoTalk, founded by Kakao Corp in 2010, dominates the South Korean market with a 93% market penetration . Leveraging this extensive user base, Kakao Games has implemented in-messenger gaming features. Their strategy revolves around utilizing KakaoTalk's social network to enhance player connectivity, driving user engagement and retention. 93% market penetration 93% market penetration Kakao Games has taken a strategic approach to its growth, forming partnerships with global developers like Ocean Drive Studio, Chrono Studio, Candy Digital, and METABORA. This collaborative strategy has diversified its gaming portfolio, focusing on high-quality, narrative-driven experiences. Key offerings like "Guardian Tales" and "Eversoul" testify to the platform's commitment to quality and innovation, instilling confidence in its future. Innovative Approach to In-Messenger Games: Telegram's Web3 Strategy Innovative Approach to In-Messenger Games: Telegram's Web3 Strategy In the Western market, Telegram has emerged as a frontrunner in integrating blockchain technology within its messaging platform, positioning itself at the forefront of the Web3 movement. As of 2024, Telegram boasts a substantial global user base of 950 million active users , largely driven by the growing popularity of games, creating a significant market for its Web3 initiatives. 950 million active users 950 million active users Telegram's strategy leverages The Open Network (TON) to facilitate secure and rapid transactions within the app, supporting in-game economies and decentralized applications (dApps). This aligns with industry projections from DappRadar , which anticipate the blockchain gaming market to reach a valuation of $39.7 billion by 2025, indicating substantial growth potential and exciting opportunities in this sector. industry projections from DappRadar However, as Daniil Shcherbakov, CEO of TonTon Games, cautions, merely porting over successful Web2 games won’t resonate with today’s Web3 audience on Telegram. Shcherbakov explains, “It’s not enough to take a Web2 game, even a successful Web2 game, and put it on Telegram and expect money to grow on trees. There is no audience currently for that. It should have a proper, well-developed Web3 narrative. Without it, people won’t play it, at least the people that are playing now.” “It’s not enough to take a Web2 game, even a successful Web2 game, and put it on Telegram and expect money to grow on trees. There is no audience currently for that. It should have a proper, well-developed Web3 narrative. Without it, people won’t play it, at least the people that are playing now.” “It’s not enough to take a Web2 game, even a successful Web2 game, and put it on Telegram and expect money to grow on trees. There is no audience currently for that. It should have a proper, well-developed Web3 narrative. Without it, people won’t play it, at least the people that are playing now.” “It’s not enough to take a Web2 game, even a successful Web2 game, and put it on Telegram and expect money to grow on trees. There is no audience currently for that. It should have a proper, well-developed Web3 narrative. Without it, people won’t play it, at least the people that are playing now.” This focus on Web3-native gaming mechanics is evident in Telegram's gaming portfolio, which includes titles designed to leverage blockchain integration. For instance, Notcoin is a tap-to-earn game where users can accumulate Notcoin (NOT) through simple gameplay mechanics, while SnakeTON is a survival game that rewards players with cryptocurrency as they navigate obstacles. Similarly, Catizen is a strategic, community-based game allowing players to earn and trade digital assets, offering an experience that appeals to Web3-minded users looking for both entertainment and economic participation. By addressing these demands, Telegram is building an ecosystem that supports not only traditional gaming enjoyment but also deeper engagement with GameFi (Game Finance) elements. Asian Super Apps vs. Telegram: Key Differences Asian Super Apps vs. Telegram: Key Differences Cultural background undoubtedly shapes game design, impacting every step in the process — from narrative themes to character development. Game developers often draw inspiration from their cultural heritage, enriching gameplay and fostering a deeper connection between players through storytelling. Telegram games focus more on individualism, freedom, and open-world exploration, while Asian games often emphasize community, structured progression, and deep storytelling. This approach reflects players' cultural values and preferences in each region, vividly illustrated through game mechanics, aesthetics, and engagement strategies. The adoption of blockchain and Web3 technologies further highlights the divergence between these approaches. WeChat faced significant hurdles in this area due to strict government oversight of blockchain and cryptocurrency activities in China. Kakao, despite joining the Bank of Korea's digital currency project through its subsidiary Ground X, was positioned more as a follower than a leader in blockchain innovation. Telegram's approach differs markedly from that of its Eastern counterparts. The platform enhanced user engagement by integrating blockchain technology via TON, implementing features such as collectible gifts and NFTs, and opening new monetization channels. This proactive approach positioned it as a pioneer in the Web3 space within messaging platforms, yet it also poses some challenges. Potential Challenges for Telegram's Web3 Ecosystem Potential Challenges for Telegram's Web3 Ecosystem Achieving mass adoption of Web3 technologies, including decentralized applications (dApps) on Telegram, will be critical for the long-term success of its ecosystem. However, widespread adoption demands significant efforts in user education and onboarding, as blockchain technology remains unfamiliar to many users. Ensuring clear, accessible information alongside user-friendly interfaces is essential for driving engagement and building a broader user base within the Web3 environment. Integrating Web3 features within a messaging platform also presents challenges related to interoperability with TON. Maintaining scalability, performance, and security will be essential to support high transaction volumes without compromising user experience. Additionally, Telegram must navigate regulatory compliance carefully, as evolving legal landscapes across jurisdictions could impact the development and deployment of Web3 features. Final Thoughts: The Future of In-Messenger Games and Web3 Integration Final Thoughts: The Future of In-Messenger Games and Web3 Integration Blockchain technology will likely reshape the gaming industry in several exciting ways. Virtual worlds and economies will become more interconnected, allowing players to own and trade in-game assets across different games truly. This will create new revenue streams for both developers and players. As in-messenger gaming continues to evolve, a few key technologies are paving the way for broader adoption across global audiences. HTML5 and Cocos are emerging as essential tools that can support mass-market growth and enhance game quality, accessibility, and lightweight functionality — crucial elements for games hosted within messaging platforms. These technologies provide the flexibility and performance needed to deliver responsive, visually rich experiences while keeping the file size low and eliminating the need for complex installations, making them ideal for today’s mobile-first world. The social dynamics within these games are shifting, too. Where early in-messenger games were primarily focused on single-player or light social elements, modern games within platforms like WeChat, Kakao, and Telegram are deeply interactive, with live chat channels and in-game communities allowing hundreds of thousands of players to engage, give real-time feedback, and build a sense of camaraderie. This is a departure from the one-way ratings and reviews seen on app stores, adding another dimension to user experience and engagement. Finally, the integration of blockchain technology on platforms like Telegram brings a new level of innovation through GameFi elements. In-game assets are becoming valuable digital items, tradable and secure through NFTs, while in-game staking and blockchain-based payments offer users new ways to invest, earn, and participate in the ecosystem. This trend echoes a global shift toward decentralized ownership and economies, signaling a new era of in-messenger gaming where players can interact with, own, and trade game assets freely.

The New Era of In-Messenger Games: How Asia’s Innovations and Web3 Technology are Shaping the Future

In today’s digital age, the demand for high-quality video content is skyrocketing across various sectors, from education to entertainment. AI video generators have emerged as essential tools to meet this growing need, offering innovative solutions that simplify and accelerate the video creation process. Enter Pollo AI, a revolutionary AI-driven video generator that promises to transform how we create and share videos with its smart, user-friendly technology. In this review, we will delve into Pollo AI's features, exploring how it stands out in the crowded field of video generation tools. What Is Pollo AI? Pollo AI is a state-of-the-art AI video generator that can take any text prompt you provide and convert it into a high-quality, professional video clip. For example, by entering a prompt like “A chess match between AI and an alien,” Pollo AI can generate a visual clip that vividly brings your text to life. Its text-to-video functionality makes creating engaging content easier than ever. Pollo AI’s innovative image-to-video feature allows users to transform static images into dynamic, high-class visual outcomes effortlessly. Simply upload an image, and Pollo AI will generate a captivating video that brings the image back to reality with motion and effects. This feature is perfect for creating engaging visual content for social media or educational purposes. How Does Pollo AI Work? Pollo AI is an intuitive tool designed for both first-timers and professionals who need to generate creative and engaging AI videos easily. Here is a comprehensive breakdown of the simple steps to get this platform started: Step 1: Input Text Prompt or Photos To create a video with Pollo AI, simply type your text prompt, like “a bomb exploded in a desert” or “a cat is walking toward a flower, sniffing and licking it,” and Pollo AI will visualize it for you. Additionally, you can upload an image, such as a family photo, and Pollo AI will create a video or animation based on that image. Step 2: Customize Your Input With Pollo AI, you can customize your video using Standard or Professional modes, choose an aspect ratio like 16:9 or 9:16, and control the length of your video. You can also adjust the prompt adherence slider for creative flexibility and select camera movements such as Horizontal Pan or Vertical Pan to tailor your video’s specifics. Step 3: Click Create to Start the Process After inputting all your preferred video settings, the final step is to hit the Create button and let Pollo AI start crafting your video. The processing time may vary based on factors like the use of Professional Mode and the video’s length and complexity, so be patient while Pollo AI works its magic. Step 4: Check Out the Ultimate Creation When the visual content is complete, you can view it directly in your browser. You have the option to watch it multiple times, save it, download it, share it, or use it however you like. If you are not satisfied, you can always return to the drawing board and start the process again. Why Choose Pollo AI? While there are other AI video generators available on the market, Pollo AI stands out as the superior option for creating high-class visual content. Let’s dive into its main features in detail. Text-To-Video Transformation: With a simple text prompt, Pollo AI can bring your ideas to life by identifying crucial elements and visualizing them in a compelling video. This tool creatively interprets your instructions, delivering the perfect video style and content for your specific needs. Refresh the Story of Your Image: Give your static images a dynamic makeover. Along with your text inputs, Pollo AI can breathe life into your visuals by creating videos that narrate their stories while preserving the original content. This smart tool transforms simple images into engaging, story-driven animations. Versatile AI Video Creation: Pollo AI can turn your ideas in mind into videos of any style you desire. Whether you envision lifelike cinematic sequences or creative, stylized animations, Pollo AI offers limitless possibilities to bring your concepts to vibrant life. Rapid Video Production: Its state-of-the-art video generation algorithms are engineered for speed, enabling Pollo AI to produce high-quality videos in mere minutes. Whether you need a quick turnaround for a project or want to see your ideas come to life swiftly, Pollo AI ensures you won’t have to wait long. Exceptional Video Quality: Pollo AI leverages cutting-edge technology to transform your creative ideas into stunning, high-definition videos that meet professional standards. The result is a perfect fusion of innovation and excellence, offering you visually captivating outputs that stand out. Effortless Video Creation: Users do not need any technical video-editing skills to utilize Pollo AI. This AI video generator is designed to be user-friendly and intuitive, simplifying the video creation process. With Pollo AI, anyone can easily create professional-grade videos without prior experience or technical knowledge. Summary Pollo AI sets a new benchmark in the world of video generation with its advanced, user-centric technology. By streamlining the video creation process, it allows users to produce high-quality content quickly and effortlessly, making it an invaluable tool for content creators. In a market crowded with video generation tools, Pollo AI stands out not only for its intelligent features but also for its commitment to delivering exceptional results. Whether you are a novice or a professional, Pollo AI provides the perfect blend of innovation and ease of use, ensuring your visual outcomes are both stunning and impactful. This story was distributed by Margrowth under HackerNoon’s Brand As An Author Program. Learn more about the program here: https://business.hackernoon.com/brand-as-author In today’s digital age, the demand for high-quality video content is skyrocketing across various sectors, from education to entertainment. AI video generators have emerged as essential tools to meet this growing need, offering innovative solutions that simplify and accelerate the video creation process. Enter Pollo AI , a revolutionary AI-driven video generator that promises to transform how we create and share videos with its smart, user-friendly technology. In this review, we will delve into Pollo AI's features, exploring how it stands out in the crowded field of video generation tools. Pollo AI Pollo AI What Is Pollo AI? Pollo AI is a state-of-the-art AI video generator that can take any text prompt you provide and convert it into a high-quality, professional video clip. For example, by entering a prompt like “A chess match between AI and an alien,” Pollo AI can generate a visual clip that vividly brings your text to life. Its text-to-video functionality makes creating engaging content easier than ever. text-to-video text-to-video Pollo AI’s innovative image-to-video feature allows users to transform static images into dynamic, high-class visual outcomes effortlessly. Simply upload an image, and Pollo AI will generate a captivating video that brings the image back to reality with motion and effects. This feature is perfect for creating engaging visual content for social media or educational purposes. image-to-video image-to-video How Does Pollo AI Work? Pollo AI is an intuitive tool designed for both first-timers and professionals who need to generate creative and engaging AI videos easily. Here is a comprehensive breakdown of the simple steps to get this platform started: Step 1: Input Text Prompt or Photos To create a video with Pollo AI, simply type your text prompt, like “a bomb exploded in a desert” or “a cat is walking toward a flower, sniffing and licking it,” and Pollo AI will visualize it for you. Additionally, you can upload an image, such as a family photo, and Pollo AI will create a video or animation based on that image. Step 2: Customize Your Input With Pollo AI, you can customize your video using Standard or Professional modes, choose an aspect ratio like 16:9 or 9:16, and control the length of your video. You can also adjust the prompt adherence slider for creative flexibility and select camera movements such as Horizontal Pan or Vertical Pan to tailor your video’s specifics. Step 3: Click Create to Start the Process After inputting all your preferred video settings, the final step is to hit the Create button and let Pollo AI start crafting your video. The processing time may vary based on factors like the use of Professional Mode and the video’s length and complexity, so be patient while Pollo AI works its magic. Step 4: Check Out the Ultimate Creation When the visual content is complete, you can view it directly in your browser. You have the option to watch it multiple times, save it, download it, share it, or use it however you like. If you are not satisfied, you can always return to the drawing board and start the process again. Why Choose Pollo AI? While there are other AI video generators available on the market, Pollo AI stands out as the superior option for creating high-class visual content. Let’s dive into its main features in detail. Text-To-Video Transformation: With a simple text prompt, Pollo AI can bring your ideas to life by identifying crucial elements and visualizing them in a compelling video. This tool creatively interprets your instructions, delivering the perfect video style and content for your specific needs. Text-To-Video Transformation: With a simple text prompt, Pollo AI can bring your ideas to life by identifying crucial elements and visualizing them in a compelling video. This tool creatively interprets your instructions, delivering the perfect video style and content for your specific needs. Refresh the Story of Your Image: Give your static images a dynamic makeover. Along with your text inputs, Pollo AI can breathe life into your visuals by creating videos that narrate their stories while preserving the original content. This smart tool transforms simple images into engaging, story-driven animations. Refresh the Story of Your Image: Give your static images a dynamic makeover. Along with your text inputs, Pollo AI can breathe life into your visuals by creating videos that narrate their stories while preserving the original content. This smart tool transforms simple images into engaging, story-driven animations. Versatile AI Video Creation: Pollo AI can turn your ideas in mind into videos of any style you desire. Whether you envision lifelike cinematic sequences or creative, stylized animations, Pollo AI offers limitless possibilities to bring your concepts to vibrant life. Versatile AI Video Creation: Pollo AI can turn your ideas in mind into videos of any style you desire. Whether you envision lifelike cinematic sequences or creative, stylized animations, Pollo AI offers limitless possibilities to bring your concepts to vibrant life. Rapid Video Production: Its state-of-the-art video generation algorithms are engineered for speed, enabling Pollo AI to produce high-quality videos in mere minutes. Whether you need a quick turnaround for a project or want to see your ideas come to life swiftly, Pollo AI ensures you won’t have to wait long. Rapid Video Production: Its state-of-the-art video generation algorithms are engineered for speed, enabling Pollo AI to produce high-quality videos in mere minutes. Whether you need a quick turnaround for a project or want to see your ideas come to life swiftly, Pollo AI ensures you won’t have to wait long. Exceptional Video Quality: Pollo AI leverages cutting-edge technology to transform your creative ideas into stunning, high-definition videos that meet professional standards. The result is a perfect fusion of innovation and excellence, offering you visually captivating outputs that stand out. Exceptional Video Quality: Pollo AI leverages cutting-edge technology to transform your creative ideas into stunning, high-definition videos that meet professional standards. The result is a perfect fusion of innovation and excellence, offering you visually captivating outputs that stand out. Effortless Video Creation: Users do not need any technical video-editing skills to utilize Pollo AI. This AI video generator is designed to be user-friendly and intuitive, simplifying the video creation process. With Pollo AI, anyone can easily create professional-grade videos without prior experience or technical knowledge. Effortless Video Creation: Users do not need any technical video-editing skills to utilize Pollo AI. This AI video generator is designed to be user-friendly and intuitive, simplifying the video creation process. With Pollo AI, anyone can easily create professional-grade videos without prior experience or technical knowledge. Summary Pollo AI sets a new benchmark in the world of video generation with its advanced, user-centric technology. By streamlining the video creation process, it allows users to produce high-quality content quickly and effortlessly, making it an invaluable tool for content creators. Pollo AI Pollo AI In a market crowded with video generation tools, Pollo AI stands out not only for its intelligent features but also for its commitment to delivering exceptional results. Whether you are a novice or a professional, Pollo AI provides the perfect blend of innovation and ease of use, ensuring your visual outcomes are both stunning and impactful. This story was distributed by Margrowth under HackerNoon’s Brand As An Author Program. Learn more about the program here: https://business.hackernoon.com/brand-as-author This story was distributed by Margrowth under HackerNoon’s Brand As An Author Program. Learn more about the program here: https://business.hackernoon.com/brand-as-author This story was distributed by Margrowth under HackerNoon’s Brand As An Author Program. Learn more about the program here: https://business.hackernoon.com/brand-as-author https://business.hackernoon.com/brand-as-author

Pollo AI Review: Intelligent AI-Based Video Generator

Pollo AI Review: Intelligenter KI-basierter Videogenerator

Reseña de Pollo AI: generador de videos inteligente basado en inteligencia artificial

סקירת Pollo AI: מחולל וידאו חכם מבוסס בינה מלאכותית

Pollo AI Revizyon: Entelijan AI ki baze sou dèlko videyo

Ulasan Pollo AI: Generator Video Cerdas Berbasis AI

Pollo AI Recensione: generatore video intelligente basato sull'intelligenza artificiale

Pollo AI レビュー: インテリジェントな AI ベースのビデオ ジェネレーター

Pollo AI მიმოხილვა: ინტელექტუალური AI-ზე დაფუძნებული ვიდეო გენერატორი

Pollo AI Review: ម៉ាស៊ីនបង្កើតវីដេអូដែលមានមូលដ្ឋានលើ AI ឆ្លាតវៃ

Pollo AI Review: Intelligente AI-gebaseerde videogenerator

Recenzia Pollo AI: Inteligentný videogenerátor založený na AI

Pollo AI İncelemesi: Akıllı AI Tabanlı Video Oluşturucu

Coze: A Love Story

Did you know that 48% of US smartphone users admit they’re addicted to screens? I was one of them. But I was too naive to admit. Screen addiction nearly tore my family apart, turning our vibrant home into a series of disconnected moments. I still remember one dinner where my wife and I sat across from our three children, all glued to their screens, barely uttering a word to each other. It hit me then: if we didn’t address this growing problem, we risked losing what truly mattered—our family bond. Recognizing the need for change, I embarked on a journey to reclaim our connections and restore meaningful communication. Here’s how we took control of our screen habits and strengthened our family ties. We Set A Poor Example For Our Kids In the early days of smartphone usage, my wife and I didn’t think much about the habits we were setting. We were busy with our own lives—balancing work and responsibilities—and we justified our phone usage as a way to stay connected and informed. Little did we know, we were unwittingly teaching them to prioritize screens over family connections. Gradually, screens became ubiquitous in our daily lives. They invaded our breakfast routines, crept into family movie nights, and even made their way into bedtime rituals. I remember how excited we were when we first gave our oldest child a smartphone. It felt like a rite of passage, a way for him to stay connected with friends. But instead of encouraging responsible usage, we fell into a routine where phones were always within reach. Research shows that the average time spent on Instagram’s Android app per user is 15 hours and 50 minutes per month. That’s a significant chunk of time, and I could see it affecting my kids firsthand. They no longer engaged in conversations or shared their thoughts about school and friends. Instead, they communicated through emojis and memes, losing the art of genuine connection. The Impact of Screen Addiction On Our Family As screens took center stage in our household, I noticed a troubling shift in our family dynamics. Excessive screen time began eroding our connections, with conversations dwindling to silence, replaced by the click-clack of keyboards. My children, ages 12 and 13, struggled to engage with each other, having traded playtime for their own virtual worlds. A once-enjoyable family game night became a frustrating reminder of our disconnect; instead of laughter, my kids were absorbed in their phones, barely making eye contact. The emotional distance deepened, leaving me feeling helpless as I watched them drift apart, communicating through texts instead of face-to-face. This realization hit hard—I understood we needed to confront our screen addiction before we lost our family bonds entirely and became disconnected lives living under one roof. Taking Action: Starting the Conversation Recognizing the need for change, I decided to confront the issue of screen addiction with my family. One evening, I gathered everyone in the living room, aiming for an open dialogue. I took a deep breath and said, “We need to talk about our screens and how they’re affecting us as a family.” The initial silence was thick; my kids exchanged skeptical glances, accustomed to rules without understanding the reasons behind them. I shared my observations about how our screen habits impacted our relationships, expressing how much I missed our shared stories and laughter. My daughter rolled her eyes, and my son muttered about needing his phone for school, but I pressed on. I emphasized that this wasn’t about punishment but finding a better balance to strengthen our connections. To my surprise, curiosity sparked. My oldest shared his pressure to respond quickly to messages, while my youngest missed playing games together. We then brainstormed solutions, creating potential ground rules like “no phones at the dinner table.” Involving everyone shifted their attitudes; they felt empowered to contribute, transforming this into a collaborative effort. As we finalized our list, I felt renewed hope—together, we were reclaiming our family bond. Creating Family Rules: Struggles and Resistance After our brainstorming session, we had a solid list of rules to reduce screen time and encourage meaningful interactions. However, implementing them proved challenging. During our first family dinner, everyone was excited about the new guidelines, but the urge to check our phones crept in as we sat down. I insisted on the “no phones at the dinner table” rule, but my youngest son fidgeted and pleaded to check his phone, breaking the calm. The struggle continued during our first family movie night. My daughter suggested watching a film together, but as the movie started, everyone’s attention drifted to their phones, overwhelmed by the familiar pull of notifications. I felt frustration rise and reminded them of our rules. Arguments erupted over “just one more scroll” or “needing phones for homework,” making it feel like a losing battle against screens. Despite these challenges, I recognized the importance of pushing through to reconnect as a family. Small victories emerged, like evenings where we navigated dinner without screens, sharing laughs and stories. These moments became the foundation for building trust and connection, reminding us that we were on this journey together. Gradual Improvement: Building Better Connections Despite initial struggles, we began to see gradual improvements in our family dynamics. Committing to screen-free practices allowed us to rediscover the joy of spending quality time together. I vividly recall one Friday night when we held a board game marathon, gathering around the table with snacks and laughter instead of being glued to our screens. As the games unfolded, conversations flowed freely. The kids shared funny anecdotes from school and engaged in playful banter, filling the house with warmth and laughter. We also explored alternative family activities like hiking, visiting parks, trying new restaurants, and cooking together, reinforcing that life beyond screens could be fulfilling. Over time, I noticed a remarkable shift in communication; my kids began to open up about their thoughts and feelings. These conversations reaffirmed that our efforts to reconnect were about nurturing relationships, and I learned the importance of celebrating our small victories along the way. Tips for Families Struggling with Screen Addiction Open Discussions are Crucial Begin by having candid conversations about screen time and its impact on family dynamics, allowing everyone to share their thoughts and feelings. This fosters a sense of shared responsibility. Involve Everyone in Rule-Making Involve everyone in rule-making, as it makes family members feel invested in following the guidelines. Drafting rules together emphasizes a collective effort to improve family life. Set Realistic Expectations Change takes time and habits won’t shift overnight. Approach setbacks with patience and remind each other that you’re all learning together. Celebrate Small Victories Celebrate small victories, like successful screen-free dinners or enjoyable outings, to reinforce positive changes and create motivation. Be Open to Adjustments Flexibility is key, as what works for one family may not work for another. Adapt your rules as you discover new activities and routines. Final Words: Be Patient, It Takes Time Reflecting on our journey to overcome screen addiction, three key points stand out. Open communication is vital; discussing screen habits fosters understanding and encourages family members to express their feelings. Recognizing that screen addiction is a real issue allows us to confront it together. Lastly, patience is essential; change takes time, and setbacks are part of the process. Celebrating small victories reinforces progress and motivation. With consistent communication, awareness, and patience, you can build stronger family connections. The path may be challenging, but the rewards—deeper bonds and meaningful interactions—are well worth the effort. Take that first step today. Did you know that 48% of US smartphone users admit they’re addicted to screens? Did you know that 48% of US smartphone users admit they’re addicted to screens? 48% of US smartphone users 48% of US smartphone users I was one of them. But I was too naive to admit. Screen addiction nearly tore my family apart, turning our vibrant home into a series of disconnected moments. I still remember one dinner where my wife and I sat across from our three children, all glued to their screens, barely uttering a word to each other. It hit me then: if we didn’t address this growing problem, we risked losing what truly mattered—our family bond. Recognizing the need for change, I embarked on a journey to reclaim our connections and restore meaningful communication. Here’s how we took control of our screen habits and strengthened our family ties. We Set A Poor Example For Our Kids In the early days of smartphone usage, my wife and I didn’t think much about the habits we were setting. We were busy with our own lives—balancing work and responsibilities—and we justified our phone usage as a way to stay connected and informed. Little did we know, we were unwittingly teaching them to prioritize screens over family connections. Gradually, screens became ubiquitous in our daily lives. They invaded our breakfast routines, crept into family movie nights, and even made their way into bedtime rituals. I remember how excited we were when we first gave our oldest child a smartphone. It felt like a rite of passage, a way for him to stay connected with friends. But instead of encouraging responsible usage, we fell into a routine where phones were always within reach. Research shows that the average time spent on Instagram’s Android app per user is 15 hours and 50 minutes per month. That’s a significant chunk of time, and I could see it affecting my kids firsthand. They no longer engaged in conversations or shared their thoughts about school and friends. Instead, they communicated through emojis and memes, losing the art of genuine connection. The Impact of Screen Addiction On Our Family As screens took center stage in our household, I noticed a troubling shift in our family dynamics. Excessive screen time began eroding our connections, with conversations dwindling to silence, replaced by the click-clack of keyboards. My children, ages 12 and 13, struggled to engage with each other, having traded playtime for their own virtual worlds. A once-enjoyable family game night became a frustrating reminder of our disconnect; instead of laughter, my kids were absorbed in their phones, barely making eye contact. The emotional distance deepened, leaving me feeling helpless as I watched them drift apart, communicating through texts instead of face-to-face. This realization hit hard—I understood we needed to confront our screen addiction before we lost our family bonds entirely and became disconnected lives living under one roof. Taking Action: Starting the Conversation Recognizing the need for change, I decided to confront the issue of screen addiction with my family. One evening, I gathered everyone in the living room, aiming for an open dialogue. I took a deep breath and said, “We need to talk about our screens and how they’re affecting us as a family.” The initial silence was thick; my kids exchanged skeptical glances, accustomed to rules without understanding the reasons behind them. I shared my observations about how our screen habits impacted our relationships, expressing how much I missed our shared stories and laughter. My daughter rolled her eyes, and my son muttered about needing his phone for school, but I pressed on. I emphasized that this wasn’t about punishment but finding a better balance to strengthen our connections. To my surprise, curiosity sparked. My oldest shared his pressure to respond quickly to messages, while my youngest missed playing games together. We then brainstormed solutions, creating potential ground rules like “no phones at the dinner table.” Involving everyone shifted their attitudes; they felt empowered to contribute, transforming this into a collaborative effort. As we finalized our list, I felt renewed hope—together, we were reclaiming our family bond. Creating Family Rules: Struggles and Resistance After our brainstorming session, we had a solid list of rules to reduce screen time and encourage meaningful interactions. However, implementing them proved challenging. During our first family dinner, everyone was excited about the new guidelines, but the urge to check our phones crept in as we sat down. I insisted on the “no phones at the dinner table” rule, but my youngest son fidgeted and pleaded to check his phone, breaking the calm. The struggle continued during our first family movie night. My daughter suggested watching a film together, but as the movie started, everyone’s attention drifted to their phones, overwhelmed by the familiar pull of notifications. I felt frustration rise and reminded them of our rules. Arguments erupted over “just one more scroll” or “needing phones for homework,” making it feel like a losing battle against screens. Despite these challenges, I recognized the importance of pushing through to reconnect as a family. Small victories emerged, like evenings where we navigated dinner without screens, sharing laughs and stories. These moments became the foundation for building trust and connection, reminding us that we were on this journey together. Gradual Improvement: Building Better Connections Despite initial struggles, we began to see gradual improvements in our family dynamics. Committing to screen-free practices allowed us to rediscover the joy of spending quality time together. I vividly recall one Friday night when we held a board game marathon, gathering around the table with snacks and laughter instead of being glued to our screens. As the games unfolded, conversations flowed freely. The kids shared funny anecdotes from school and engaged in playful banter, filling the house with warmth and laughter. We also explored alternative family activities like hiking, visiting parks, trying new restaurants, and cooking together, reinforcing that life beyond screens could be fulfilling. Over time, I noticed a remarkable shift in communication; my kids began to open up about their thoughts and feelings. These conversations reaffirmed that our efforts to reconnect were about nurturing relationships, and I learned the importance of celebrating our small victories along the way. Tips for Families Struggling with Screen Addiction Open Discussions are Crucial Begin by having candid conversations about screen time and its impact on family dynamics, allowing everyone to share their thoughts and feelings. This fosters a sense of shared responsibility. Involve Everyone in Rule-Making Involve everyone in rule-making, as it makes family members feel invested in following the guidelines. Drafting rules together emphasizes a collective effort to improve family life. Set Realistic Expectations Change takes time and habits won’t shift overnight. Approach setbacks with patience and remind each other that you’re all learning together. Celebrate Small Victories Celebrate small victories, like successful screen-free dinners or enjoyable outings, to reinforce positive changes and create motivation. Be Open to Adjustments Flexibility is key, as what works for one family may not work for another. Adapt your rules as you discover new activities and routines. Final Words: Be Patient, It Takes Time Reflecting on our journey to overcome screen addiction, three key points stand out. Open communication is vital; discussing screen habits fosters understanding and encourages family members to express their feelings. Recognizing that screen addiction is a real issue allows us to confront it together. Lastly, patience is essential; change takes time, and setbacks are part of the process. Celebrating small victories reinforces progress and motivation. With consistent communication, awareness, and patience, you can build stronger family connections. The path may be challenging, but the rewards—deeper bonds and meaningful interactions—are well worth the effort. Take that first step today.

Screen Addiction Can Ruin Your Family

Lessons in Development, Hard Work, and Critical Thinking Life before AI, being a developer, was a different world — a harder world in the sense that everything you do requires tremendous focus and incredible persistence. Challenges were real, and so were the triumphs that followed. Honest to God, I sometimes miss parts of that time. I was reminded of those early days vividly. There were no AI-powered tools back in those days that could swoop in and solve your problems for you. Every problem was a mountain which you had to climb alone. If you had bugs in your code, no AI could have run a simulation or suggested a fix. You had to work out by yourself. Countless late nights staring at my screen, combing through lines of code trying to figure out where I did it wrong. Debugging was like detective work. Sometimes, finally, it would come out what I was hunting for, and the relief was almost overwhelming — like just unlocking this huge mystery that had been gnawing at me for hours maybe even days. You’d learn so much in those moments. Not only about the code you’re writing but about the system itself — the inner workings of the technology. I only learned when I needed to know how to do something; it wasn’t because some tool just pointed me in the right direction, otherwise. It was slow and often frustrating, but this slow burn of learning made me a better developer. Without the safety net of AI, I needed to know what I was doing at each step. I think one of the reasons why developers nowadays miss this depth of understanding is that they have these incredibly powerful tools right there at their fingertips. And, of course, I cannot forget the level of “manuals” of everything: everything required a tremendous amount of effort, including writing test cases, for example. Nowadays, AI can write tests for you or suggest improvements, and everything is ready for you; back then, it was all up to me. I had to think of all the ways that my code might break and code for each one. And these weren’t glamorous pieces of work, but they were pretty darn important. Every test that I wrote taught me something about how fragile my code could be, and how I could make my code robust. The same applied to optimizing code. I had no fancy AI tools to analyze my work and suggest ways of efficiency. Instead, I had to spend hours iterating over different versions of a function, trying to shave off milliseconds from execution times or reduce memory consumption. It was boring, but it sharpened my understanding of how systems work under the hood. It also meant knowing about memory management, algorithm efficiency, and how the hardware was going to interpret my code. There’s no autopilot to do it for you. It was different too, as far as collaboration with my team was concerned. We had no AI to turn to quickly for debugging or optimization, so we relied on each other nearly for everything. We brainstormed, troubleshoot, and sometimes argued over the best ways to solve a problem. It was never easy, but there was a camaraderie in just knowing we were all in the same boat, all struggling with the same challenges. The beauty of it was that we were learning together every step of the way. It was tough getting through acquiring new skills back then. There were no shortcuts. So if you wanted to learn a new programming language or technology, you had to dive in deep. There wasn’t an AI assistant to fill in the gaps or give real-time feedback. If you didn’t know how something worked, you figured it out by yourself or through your mentors and peers. That’s how I learned C# — through trial and error and countless mistakes. I shall never forget the headache of trying to get a hold of things like multithreading, exception handling, or memory leaks. There wasn’t a silver bullet-just long hours of frustration followed by satisfaction when things worked. Another was the errors. Every time I hit a bug, it felt like hitting a wall. But with each of those walls lay an opportunity. To debug did not mean that you solved the problem; it was instead, learning why the problem existed in the first place. It is what those experiences shaped me: how to approach problems methodically and patiently. In that world before AI, what mattered above all else were your skills. Knowing how to write code was certainly not sufficient; it required creativity when solving problems and critical thinking. I still remember starting from scratch so many times because I realized my original thought was inefficient or wrong in general. But that was what made me improve. And each time I re-wrote a function or refactored a class, I was improving too, I was getting to know how to make performance and readability decisions, and how to trade-off between speed and stability. Even crafting something as simple as a SQL query was a chore in precision. There was no such thing as AI-crafted databases that could optimize queries on the fly. I spent hours futzing with them, trying to make sure that they ran as well as possible, especially when I was dealing with larger datasets. All these performance tweaks had been hard-won through trial and painful error. Then quietly began AI’s influence in the world of development. At first, it was just little things auto-complete in my IDE getting smarter, code editors catching errors before I even ran the program. I did not realize it at that time, but changes like these were game-changers. Things that took me hours to get done could be completed in a few minutes. Debugging, optimizing-even generating some boilerplate code became easier. It is like having an assistant who could see farther than my tiny horizon, an assistant who could point out all the problems that I had never considered at all. But as easy as it was, so was it strange. I was no longer asked to dig deep into the problems anymore. The AI was already there before I could understand the issue- presenting the solutions for me. It made me much more efficient, yes, but part of me misses the depth of understanding that came with doing things the hard way. There is satisfaction in cracking on your own a tough problem that AI, for all its brilliance, can’t reenact. Learning has changed these days and in this AI-driven era. The hard work isn’t gone; it’s just different. Today, developers do not bother memorizing syntax or spending hours debugging. It is all about how to work with AI. How to use these tools to be more efficient at building faster, more scalable solutions. However, it is still the same problem-solving essence. It can be described as AI suggesting a potential fix, but then, the decision still lies on the side of the developer whether it is right and whether it fits into the overall context of the project. This new chapter has been exciting and challenging for me. The transition to technological leadership allowed me to free up more energy to focus on bigger problems, building complex systems. I will always look back at the days before AI with a certain sort of nostalgia. It was long nights and tough-won victories that made me the developer that I am today. And for that, I appreciate this efficiency brought on by AI, but I will never forget the lessons I learned when it was just me, my code, and that quiet determination to figure things out. Follow on Medium: The AI Guy Lessons in Development, Hard Work, and Critical Thinking Life before AI, being a developer, was a different world — a harder world in the sense that everything you do requires tremendous focus and incredible persistence. Challenges were real, and so were the triumphs that followed. Honest to God, I sometimes miss parts of that time. I was reminded of those early days vividly. There were no AI-powered tools back in those days that could swoop in and solve your problems for you. Every problem was a mountain which you had to climb alone. If you had bugs in your code, no AI could have run a simulation or suggested a fix. You had to work out by yourself. Countless late nights staring at my screen, combing through lines of code trying to figure out where I did it wrong. Debugging was like detective work. Sometimes, finally, it would come out what I was hunting for, and the relief was almost overwhelming — like just unlocking this huge mystery that had been gnawing at me for hours maybe even days. You’d learn so much in those moments. Not only about the code you’re writing but about the system itself — the inner workings of the technology. I only learned when I needed to know how to do something; it wasn’t because some tool just pointed me in the right direction, otherwise. It was slow and often frustrating, but this slow burn of learning made me a better developer. Without the safety net of AI, I needed to know what I was doing at each step. I think one of the reasons why developers nowadays miss this depth of understanding is that they have these incredibly powerful tools right there at their fingertips. And, of course, I cannot forget the level of “manuals” of everything: everything required a tremendous amount of effort, including writing test cases, for example. Nowadays, AI can write tests for you or suggest improvements, and everything is ready for you; back then, it was all up to me. I had to think of all the ways that my code might break and code for each one. And these weren’t glamorous pieces of work, but they were pretty darn important. Every test that I wrote taught me something about how fragile my code could be, and how I could make my code robust. The same applied to optimizing code. I had no fancy AI tools to analyze my work and suggest ways of efficiency. Instead, I had to spend hours iterating over different versions of a function, trying to shave off milliseconds from execution times or reduce memory consumption. It was boring, but it sharpened my understanding of how systems work under the hood. It also meant knowing about memory management, algorithm efficiency, and how the hardware was going to interpret my code. There’s no autopilot to do it for you. It was different too, as far as collaboration with my team was concerned. We had no AI to turn to quickly for debugging or optimization, so we relied on each other nearly for everything. We brainstormed, troubleshoot, and sometimes argued over the best ways to solve a problem. It was never easy, but there was a camaraderie in just knowing we were all in the same boat, all struggling with the same challenges. The beauty of it was that we were learning together every step of the way. It was tough getting through acquiring new skills back then. There were no shortcuts. So if you wanted to learn a new programming language or technology, you had to dive in deep. There wasn’t an AI assistant to fill in the gaps or give real-time feedback. If you didn’t know how something worked, you figured it out by yourself or through your mentors and peers. That’s how I learned C# — through trial and error and countless mistakes. I shall never forget the headache of trying to get a hold of things like multithreading, exception handling, or memory leaks. There wasn’t a silver bullet-just long hours of frustration followed by satisfaction when things worked. Another was the errors. Every time I hit a bug, it felt like hitting a wall. But with each of those walls lay an opportunity. To debug did not mean that you solved the problem; it was instead, learning why the problem existed in the first place. It is what those experiences shaped me: how to approach problems methodically and patiently. In that world before AI, what mattered above all else were your skills. Knowing how to write code was certainly not sufficient; it required creativity when solving problems and critical thinking. I still remember starting from scratch so many times because I realized my original thought was inefficient or wrong in general. But that was what made me improve. And each time I re-wrote a function or refactored a class, I was improving too, I was getting to know how to make performance and readability decisions, and how to trade-off between speed and stability. Even crafting something as simple as a SQL query was a chore in precision. There was no such thing as AI-crafted databases that could optimize queries on the fly. I spent hours futzing with them, trying to make sure that they ran as well as possible, especially when I was dealing with larger datasets. All these performance tweaks had been hard-won through trial and painful error. Then quietly began AI’s influence in the world of development. At first, it was just little things auto-complete in my IDE getting smarter, code editors catching errors before I even ran the program. I did not realize it at that time, but changes like these were game-changers. Things that took me hours to get done could be completed in a few minutes. Debugging, optimizing-even generating some boilerplate code became easier. It is like having an assistant who could see farther than my tiny horizon, an assistant who could point out all the problems that I had never considered at all. But as easy as it was, so was it strange. I was no longer asked to dig deep into the problems anymore. The AI was already there before I could understand the issue- presenting the solutions for me. It made me much more efficient, yes, but part of me misses the depth of understanding that came with doing things the hard way. There is satisfaction in cracking on your own a tough problem that AI, for all its brilliance, can’t reenact. Learning has changed these days and in this AI-driven era. The hard work isn’t gone; it’s just different. Today, developers do not bother memorizing syntax or spending hours debugging. It is all about how to work with AI. How to use these tools to be more efficient at building faster, more scalable solutions. However, it is still the same problem-solving essence. It can be described as AI suggesting a potential fix, but then, the decision still lies on the side of the developer whether it is right and whether it fits into the overall context of the project. This new chapter has been exciting and challenging for me. The transition to technological leadership allowed me to free up more energy to focus on bigger problems, building complex systems. I will always look back at the days before AI with a certain sort of nostalgia. It was long nights and tough-won victories that made me the developer that I am today. And for that, I appreciate this efficiency brought on by AI, but I will never forget the lessons I learned when it was just me, my code, and that quiet determination to figure things out. Follow on Medium: The AI Guy The AI Guy

The World Before AI

A few years ago, a proficient designer needed skills in Adobe Photoshop for retouching, Adobe Illustrator for drawing, and Figma for creating landing pages. However, the demand for highly specialized designers has surged. Now, a product designer focuses solely on product design, and a graphic designer concentrates on graphic elements, without overlapping roles. The need for all-in-one designers is declining, which is a positive shift. However, this trend hasn’t yet affected the game development industry. GameDev still requires versatile specialists who possess a broad skill set. If you’re aiming for a career in the game industry, you must be adept at a wide range of skills. In this article, I will delve into the essential skills required for GameDev and illustrate each with examples from my own experience. Graphic Design: From Banners to Brand Identity Graphic design encompasses many elements. It includes identity and branding, as well as all promotional materials, from avatars and banners to merchandise and billboards. As part of graphic design in game development, you will develop a project’s identity and brand book, create banners, design social media assets, and develop merchandise. What you need to know in graphic design: you need a comprehensive understanding of typography, color theory, and composition, and the ability to integrate these elements beautifully into the right format. UX/UI and Web Every game has a website, usually a multi-page site that needs to be designed for easy user navigation. Here too, the designer must think about structure and design interfaces. In web design, you will handle the development of promotional sites, special projects, and landing pages for new game updates. For long-running projects that are frequently updated, you’ll need to create a design system. For more straightforward, one-off sites, you’ll build a UI kit. This category also includes creating email newsletters, as the tasks are similar to web design. What you need to know about web design: Interface design skills are essential, from research to visualization. 3D Design for Marketing For web and graphic design layouts, you need various 3D graphics—cannons, chests, 3D icons, renders of characters, weapons, and achievements. Where do you get them with the perfect angle and lighting? You create them yourself. Within the scope of 3D tasks, you’ll need to: Render in-game models with the right angle and lighting.
Model and render materials for promotions, including lighting and action boosts.
Create 3D graphics for special projects.
Develop 3D marketing materials that are later integrated directly into the game, such as chevrons and badges for user customization.
Create 3D sticker packs for Discord. What you need to know about 3D design: At a minimum, you should be proficient in one 3D software package, but preferably two. One for modeling, such as Cinema 4D, Blender, or Maya, and another for texturing, like Adobe Substance 3D Painter or Marmoset. Sticker Pack Case.
My first challenging 3D job was a duck sticker pack. I spent sleepless nights grappling with complex materials and shapes while trying to convey anthropomorphic emotions on the ducks’ faces. It was tough, but I succeeded, and I will never forget those ducks. Retouching for Anything That's Not Perfect With 3D models, things can go wrong: polygons may be out of place, lighting may be off, or the model may come from the developer’s engine. To make everything look attractive and polished, retouching is essential. Knowledge of retouching allows you to quickly assemble a story art banner from existing assets or enhance them for further use. What you need to know about retouching: A basic level of retouching involves using Photoshop sliders, masks, adjustment layers, and a graphics tablet, and understanding the basics of light shading and composition. Neural Networks as an Additional Retouching Tool Neural networks can now be used for retouching. For example, you may have a 3D render of poor quality based on a ready-made model from developers. You can take the long route and retouch it yourself, or the short route by using a neural network like SD. In just five minutes, you can achieve a good result: the model will have beautiful hair, appropriate facial expressions, and correct color corrections. What you need to know about neural networks: Using them effectively requires time and resources to train the model. You need to feed the neural network about 100 photos of the object from different angles with prompts, review the results, make edits, and review everything again. Motion Design Let’s say the website for a game is ready, but it looks boring. What to do? Add IDLE animation to the main art. For this, you need to know how to create such animations. Knowledge of motion graphics allows you to create animated logos, art for promotional materials, and showreels. What you need to know to excel in motion graphics: Learn After Effects and master layering art in Photoshop. The More You Know, The More You Can Do Of course, on a design team, everyone has an area in which they are highly skilled experts. For example, I wouldn’t expect high-quality AAA art from a web designer. However, in game development, a web designer might still get AAA-art tasks and will need to handle them. Therefore, the more diverse tasks you can manage without relying on colleagues, the more valuable you will be as a specialist. Consequently, your grade, salary, and recognition will grow faster. I remember vividly how, as a designer with over a decade of experience, I retouched characters late into the night, struggled with 3D models, or nervously created my first landing page. Now I realize that it was a crucial experience. Acquired skills significantly broaden your horizons and enhance your design capabilities. There emerges an ironclad confidence that you can handle any task (even 3D ducks). So, I highly recommend developing multidisciplinary skills, especially if you want to enter game development. Even if you don’t, the more you know, the more you can achieve. A few years ago, a proficient designer needed skills in Adobe Photoshop for retouching, Adobe Illustrator for drawing, and Figma for creating landing pages. However, the demand for highly specialized designers has surged. Now, a product designer focuses solely on product design, and a graphic designer concentrates on graphic elements, without overlapping roles. A few years ago, a proficient designer needed skills in Adobe Photoshop for retouching, Adobe Illustrator for drawing, and Figma for creating landing pages. However, the demand for highly specialized designers has surged. Now, a product designer focuses solely on product design, and a graphic designer concentrates on graphic elements, without overlapping roles. The need for all-in-one designers is declining, which is a positive shift. However, this trend hasn’t yet affected the game development industry. GameDev still requires versatile specialists who possess a broad skill set. If you’re aiming for a career in the game industry, you must be adept at a wide range of skills. The need for all-in-one designers is declining, which is a positive shift. However, this trend hasn’t yet affected the game development industry. GameDev still requires versatile specialists who possess a broad skill set. If you’re aiming for a career in the game industry, you must be adept at a wide range of skills. In this article, I will delve into the essential skills required for GameDev and illustrate each with examples from my own experience. In this article, I will delve into the essential skills required for GameDev and illustrate each with examples from my own experience. Graphic Design: From Banners to Brand Identity Graphic Design: From Banners to Brand Identity Graphic design encompasses many elements. It includes identity and branding, as well as all promotional materials, from avatars and banners to merchandise and billboards. As part of graphic design in game development, you will develop a project’s identity and brand book, create banners, design social media assets, and develop merchandise. What you need to know in graphic design: you need a comprehensive understanding of typography, color theory, and composition, and the ability to integrate these elements beautifully into the right format. What you need to know in graphic design: you need a comprehensive understanding of typography, color theory, and composition, and the ability to integrate these elements beautifully into the right format. UX/UI and Web Every game has a website, usually a multi-page site that needs to be designed for easy user navigation. Here too, the designer must think about structure and design interfaces. In web design, you will handle the development of promotional sites, special projects, and landing pages for new game updates. For long-running projects that are frequently updated, you’ll need to create a design system. For more straightforward, one-off sites, you’ll build a UI kit. This category also includes creating email newsletters, as the tasks are similar to web design. What you need to know about web design: Interface design skills are essential, from research to visualization. What you need to know about web design: Interface design skills are essential, from research to visualization. 3D Design for Marketing For web and graphic design layouts, you need various 3D graphics—cannons, chests, 3D icons, renders of characters, weapons, and achievements. Where do you get them with the perfect angle and lighting? You create them yourself. Within the scope of 3D tasks, you’ll need to: Render in-game models with the right angle and lighting. Model and render materials for promotions, including lighting and action boosts. Create 3D graphics for special projects. Develop 3D marketing materials that are later integrated directly into the game, such as chevrons and badges for user customization. Create 3D sticker packs for Discord. Render in-game models with the right angle and lighting. Model and render materials for promotions, including lighting and action boosts. Create 3D graphics for special projects. Develop 3D marketing materials that are later integrated directly into the game, such as chevrons and badges for user customization. Create 3D sticker packs for Discord. What you need to know about 3D design: At a minimum, you should be proficient in one 3D software package, but preferably two. One for modeling, such as Cinema 4D, Blender, or Maya, and another for texturing, like Adobe Substance 3D Painter or Marmoset. What you need to know about 3D design: At a minimum, you should be proficient in one 3D software package, but preferably two. One for modeling, such as Cinema 4D, Blender, or Maya, and another for texturing, like Adobe Substance 3D Painter or Marmoset. Sticker Pack Case. My first challenging 3D job was a duck sticker pack. I spent sleepless nights grappling with complex materials and shapes while trying to convey anthropomorphic emotions on the ducks’ faces. It was tough, but I succeeded, and I will never forget those ducks. Sticker Pack Case. Sticker Pack Case. My first challenging 3D job was a duck sticker pack. I spent sleepless nights grappling with complex materials and shapes while trying to convey anthropomorphic emotions on the ducks’ faces. It was tough, but I succeeded, and I will never forget those ducks. Retouching for Anything That's Not Perfect Retouching for Anything That's Not Perfect With 3D models, things can go wrong: polygons may be out of place, lighting may be off, or the model may come from the developer’s engine. To make everything look attractive and polished, retouching is essential. Knowledge of retouching allows you to quickly assemble a story art banner from existing assets or enhance them for further use. What you need to know about retouching: A basic level of retouching involves using Photoshop sliders, masks, adjustment layers, and a graphics tablet, and understanding the basics of light shading and composition. What you need to know about retouching: A basic level of retouching involves using Photoshop sliders, masks, adjustment layers, and a graphics tablet, and understanding the basics of light shading and composition. Neural Networks as an Additional Retouching Tool Neural networks can now be used for retouching. For example, you may have a 3D render of poor quality based on a ready-made model from developers. You can take the long route and retouch it yourself, or the short route by using a neural network like SD. In just five minutes, you can achieve a good result: the model will have beautiful hair, appropriate facial expressions, and correct color corrections. What you need to know about neural networks: Using them effectively requires time and resources to train the model. You need to feed the neural network about 100 photos of the object from different angles with prompts, review the results, make edits, and review everything again. What you need to know about neural networks: Using them effectively requires time and resources to train the model. You need to feed the neural network about 100 photos of the object from different angles with prompts, review the results, make edits, and review everything again. Motion Design Let’s say the website for a game is ready, but it looks boring. What to do? Add IDLE animation to the main art. For this, you need to know how to create such animations. Knowledge of motion graphics allows you to create animated logos, art for promotional materials, and showreels. What you need to know to excel in motion graphics: Learn After Effects and master layering art in Photoshop. What you need to know to excel in motion graphics: Learn After Effects and master layering art in Photoshop. The More You Know, The More You Can Do Of course, on a design team, everyone has an area in which they are highly skilled experts. For example, I wouldn’t expect high-quality AAA art from a web designer. However, in game development, a web designer might still get AAA-art tasks and will need to handle them. Therefore, the more diverse tasks you can manage without relying on colleagues, the more valuable you will be as a specialist. Consequently, your grade, salary, and recognition will grow faster. I remember vividly how, as a designer with over a decade of experience, I retouched characters late into the night, struggled with 3D models, or nervously created my first landing page. Now I realize that it was a crucial experience. Acquired skills significantly broaden your horizons and enhance your design capabilities. There emerges an ironclad confidence that you can handle any task (even 3D ducks). So, I highly recommend developing multidisciplinary skills, especially if you want to enter game development. Even if you don’t, the more you know, the more you can achieve.

Why GameDev Designers Are the Swiss Army Knives of Game Development

Trust Your Gut—Because Sometimes Your Brain is Full of Sh*t

Your Subconscious Mind Creates Your Reality (It's Ridiculously Powerful)

আপনার অবচেতন মন আপনার বাস্তবতা তৈরি করে (এটি হাস্যকরভাবে শক্তিশালী)

Ihr Unterbewusstsein erschafft Ihre Realität (es ist unglaublich mächtig)

Tu mente subconsciente crea tu realidad (es ridículamente poderosa)

Votre subconscient crée votre réalité (c'est incroyablement puissant)

आपका अवचेतन मन आपकी वास्तविकता बनाता है (यह अविश्वसनीय रूप से शक्तिशाली है)

あなたの潜在意識が現実を創ります（それは驚くほど強力です）

당신의 잠재의식이 당신의 현실을 창조합니다(엄청나게 강력합니다)

Sua mente subconsciente cria sua realidade (é ridiculamente poderosa)

Ваше подсознание создает вашу реальность (оно невероятно мощно)

Bilinçaltınız Gerçekliğinizi Yaratır (Gülünç Derecede Güçlüdür)

Tiềm thức của bạn tạo ra thực tế của bạn (Nó có sức mạnh vô cùng lớn)

你的潜意识创造了你的现实（它非常强大）

Authors:
(1) Shengqiong Wu, NExT++, School of Computing, National University of Singapore;
(2) Hao Fei ,from NExT++, School of Computing at the National University of Singapore, serves as the corresponding author: haofei37@nus.edu.sg.
(3) Leigang Qu, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;;
(4) Wei Ji, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;;
(5) Tat-Seng Chua, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;. Table of Links Abstract and 1. Introduction
2 Related Work3 Overall Architecture4 Lightweight Multimodal Alignment Learning
5 Modality-switching Instruction Tuning
5.1 Instruction Tuning
5.2 Instruction Dataset
6 Experiments
6.1 Any-to-any Multimodal Generation and 6.2 Example Demonstrations
7 Conclusion and References 6 Experiments 6.1 Any-to-any Multimodal Generation We try to quantify the generation quality of NExT-GPT on certain benchmark datasets under some common settings, such as text-to-X generation, X-to-text generation, and Text-conditioned modality editing. We mimic the task by taking only one turn of interaction between the user and the model. • ‘Text’ — ‘X’ Generation represents the most frequent tasks of text-conditioned modal synthesis. Table 3, 4 and 5 present the comparisons between ours and some state-of-the-art systems. Overall NExT-GPT shows nice performance on par with the values from the best-performing baselines. • ‘X’ — ‘Text’ Generation represents the tasks of modal captioning. Table 6, 7 and 8 show the results on different tasks. Overall, we find that NExT-GPT can mostly achieve much better performance on the X-to-text generation than the CoDi baseline, owing to the direct generation of texts from LLM, which is inherently expertized by the LLM. • ‘Text+X’ — ‘X’ Generation represents a task category of text-conditioned modal editing. Table 9, 10 and 11 show the performances on different tasks. Compared with the above two types of tasks, NExT-GPT could be not that superior for the text-conditioned modal editing tasks. Yet, it still shows competitive performance • Human Evaluation on Complex Any-to-any QA We also carry out evaluation on some more scenarios where there are complicated cross-modal interactions between inputs and outputs. We mainly compare the model performance for the settings with different modality conversions. As no standard benchmark can be leveraged, here we adopt human evaluation. We ask several evaluators to score the performance of NExT-GPT on a scale from 1 to 10. Figure 5 shows the comparisons. We find NExT-GPT is more competent in producing images, compared with the generations on videos and audio. Also generating mixed combinations of multimodal content is slightly inferior to the generation of single-modal content, due to the complexity of the latter. 6.2 Example Demonstrations To demonstrate the effectiveness and potential of our proposed NExT-GPT in developing human-like conversational agents, here we further offer compelling examples that vividly illustrate the system’s exceptional capacity to comprehend and reason contents across various modalities in any combination. Figure 6, 7, 8, 9, 10 and 11 show the examples from NExT-GPT. Go to the project page for more examples and access the dynamic video and audio contents. This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license. Authors: (1) Shengqiong Wu, NExT++, School of Computing, National University of Singapore; (2) Hao Fei ,from NExT++, School of Computing at the National University of Singapore, serves as the corresponding author: haofei37@nus.edu.sg. (3) Leigang Qu, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;; (4) Wei Ji, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;; (5) Tat-Seng Chua, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;. Authors: Authors: (1) Shengqiong Wu, NExT++, School of Computing, National University of Singapore; (2) Hao Fei ,from NExT++, School of Computing at the National University of Singapore, serves as the corresponding author: haofei37@nus.edu.sg. (3) Leigang Qu, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;; (4) Wei Ji, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;; (5) Tat-Seng Chua, Hao Fei, NExT++, School of Computing, National University of Singapore is the corresponding author: haofei37@nus.edu.sg;. Table of Links Abstract and 1. Introduction 2 Related Work3 Overall Architecture4 Lightweight Multimodal Alignment Learning 5 Modality-switching Instruction Tuning 5.1 Instruction Tuning 5.2 Instruction Dataset 6 Experiments 6.1 Any-to-any Multimodal Generation and 6.2 Example Demonstrations 7 Conclusion and References Abstract and 1. Introduction Abstract and 1. Introduction 2 Related Work 3 Overall Architecture 4 Lightweight Multimodal Alignment Learning 2 Related Work 3 Overall Architecture 4 Lightweight Multimodal Alignment Learning 5 Modality-switching Instruction Tuning 5.1 Instruction Tuning 5.1 Instruction Tuning 5.2 Instruction Dataset 5.2 Instruction Dataset 6 Experiments 6.1 Any-to-any Multimodal Generation and 6.2 Example Demonstrations 6.1 Any-to-any Multimodal Generation and 6.2 Example Demonstrations 7 Conclusion and References 7 Conclusion and References 6 Experiments 6.1 Any-to-any Multimodal Generation We try to quantify the generation quality of NExT-GPT on certain benchmark datasets under some common settings, such as text-to-X generation, X-to-text generation, and Text-conditioned modality editing. We mimic the task by taking only one turn of interaction between the user and the model. • ‘Text’ — ‘X’ Generation represents the most frequent tasks of text-conditioned modal synthesis. Table 3, 4 and 5 present the comparisons between ours and some state-of-the-art systems. Overall NExT-GPT shows nice performance on par with the values from the best-performing baselines. • ‘Text’ — ‘X’ Generation • ‘X’ — ‘Text’ Generation represents the tasks of modal captioning. Table 6, 7 and 8 show the results on different tasks. Overall, we find that NExT-GPT can mostly achieve much better performance on the X-to-text generation than the CoDi baseline, owing to the direct generation of texts from LLM, which is inherently expertized by the LLM. • ‘X’ — ‘Text’ Generation • ‘Text+X’ — ‘X’ Generation represents a task category of text-conditioned modal editing. Table 9, 10 and 11 show the performances on different tasks. Compared with the above two types of tasks, NExT-GPT could be not that superior for the text-conditioned modal editing tasks. Yet, it still shows competitive performance • ‘Text+X’ — ‘X’ Generation • Human Evaluation on Complex Any-to-any QA We also carry out evaluation on some more scenarios where there are complicated cross-modal interactions between inputs and outputs. We mainly compare the model performance for the settings with different modality conversions. As no standard benchmark can be leveraged, here we adopt human evaluation. We ask several evaluators to score the performance of NExT-GPT on a scale from 1 to 10. Figure 5 shows the comparisons. We find NExT-GPT is more competent in producing images, compared with the generations on videos and audio. Also generating mixed combinations of multimodal content is slightly inferior to the generation of single-modal content, due to the complexity of the latter. • Human Evaluation on Complex Any-to-any QA 6.2 Example Demonstrations 6.2 Example Demonstrations To demonstrate the effectiveness and potential of our proposed NExT-GPT in developing human-like conversational agents, here we further offer compelling examples that vividly illustrate the system’s exceptional capacity to comprehend and reason contents across various modalities in any combination. Figure 6, 7, 8, 9, 10 and 11 show the examples from NExT-GPT. Go to the project page for more examples and access the dynamic video and audio contents. 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

NExT-GPT: Any-to-Any Multimodal LLM: Any-to-any Multimodal Generation

This week, I am taking a slight detour from my usual articles that explain AI Concepts or coding tutorials. Yes, it's a philosophical, nostalgic tour down the road I was fortunate enough to tread all along. One of the rare privileges of starting in a field quite early is that you see the field grow like your baby. When I started my Masters in Computer Vision a decade ago, a CS graduate friend of mine asked, “What does computer vision even mean?” It's a trivial question today with enough videos of computer vision tasks like segmentation (see below) floating around the internet. A result of image segmentation — a Computer Vision task that was all over the internet when deep learning took off a decade ago. So, what has changed in the last decade? There are always two sides to a coin. Also, growth can be both good and bad. Closed Community — Small Was Beautiful They say that people in Iceland are somehow related to each other and if someone is beyond your second cousin, you are good to marry them. Likewise, the AI community was close-knit. Gone are those feelings of a tight-knit community. When interacting with anyone from research you generally knew which group they belonged to. You somehow would relate to and respect their work as you read and appreciated their impact on the field. As the field is blowing up endlessly, it's becoming increasingly difficult to even say that you know someone’s work, forget about them. There are too many new names and novel branches that it's even challenging to keep track of. As another example, PyTorch was in its infancy. The Slack community was so small and helpful that the developers of PyTorch directly responded to questions we had about using the library. This encouraged learning more and more about it. Today, the framework is so mature, and new frameworks like LangChain and Llamaindex have propped up. The focus is all on LLMs over any other branch of AI. Hardware Luxury Those days, we used to train a deep neural network on a single GPU to create something impactful. Most work published in premier conferences like CVPR, NeurIPS, and ICML could be trained and replicated on a single 8 GB GPU machine or in the worst case, a single machine with 4 GPUs in rare cases. An image of the latest GPU clusters used by Tesla for AI (source: https://www.hpcwire.com/off-the-wire/tesla-to-roll-out-advanced-supercomputer) I vividly remember how happy I was to buy a single commodity GPU with just 8 GB of RAM to participate in Kaggle competitions. Some of the winning solutions from Kaggle Grandmasters trained deep learning models on a single GPU machine at home. Today’s AI world needs a GPU cluster to train foundational models. Even fine-tuning these models needs 24GB GPUs which are beasts, expensive, and affordable only by businesses with an “AI budget.” Skillset Demand Those were days when the field did not yet get the AI wrapper around it. Recruiters were given this alien task of hunting for “deep learning” engineers. Recruiters and start-up founders were hunting for deep learning experts through all channels. It was a norm to receive regular messages on LinkedIn asking me to join their team as a deep learning engineer. A screenshot of a job advert for a Machine Learning Engineer posted only 5 days ago receiving over 100 applicants! The current situation is that openings for “Machine Learning Engineer” get over 100 applications on LinkedIn well within a day of advertising. Take a look at the screenshot above if you find it hard to believe. The question remains as to how relevant the applicant's skills are to the job spec. But the market is saturated with skills — quite quickly! Sub-skills — Operations and Architecture Growth means more diversity and opportunity. There are new roles like ML-ops, LLM-ops, and ML-architects cropping up. Gone are those days of tiny, single-model files (< 1 GB). The growth of models both in size and capabilities has given rise to new skills in deploying and maintaining them. Moreover, training and deployment of models are being automated with tools such as MLFLow. The cloud infrastructure for training needs to be sophisticated enough. All this has given rise to full-time roles with dedicated responsibilities. Bye Bye ML Engineer, Hello AI Engineer The most fun in working in AI is writing the model architecture, and training the model from scratch using our in-house data. Though this involves pre-processing the data a lot, training models, and visualizing the training results used to be a lot of fun. There was/still is a specialized role for this called Machine Learning (ML) Engineer. The development of foundation models from tech giants is redefining this role. As model sizes grow, training budgets are enormous. As a matter of fact, the cost of training the LLama 2 model was $20m for Meta. Clearly start-ups or organizations trying to adopt AI do not wish to throw away this sum of money. It is now established that foundation models are for the tech giants to develop, with the exception of some companies like Mistral and Anthropic. Sadly, this means the ML Engineer role is getting cast into an AI Engineer role. The ML Engineer role was all about developing model architectures, training, and evaluating them. The new AI Engineer role mostly involves developing APIs or invoking APIs provided by Tech giants (OpenAI, Meta, and Google) to prompt the foundation models. In rare cases, it involves fine-tuning these foundation models. But companies have the option to build RAG pipelines or use the foundation models “as-is” over fine-tuning them. Conclusion To conclude, I see this as a slow amalgamation of the Software engineering and Machine Learning roles. The line between software engineers and deep learning experts is fading. So, in the years to come, software engineers will be AI engineers who work alongside foundation models, both to write code and to solve customer needs. Furthermore, in the coming years, companies will be grouped into two categories — the AI product and AI services companies. AI product companies would again be OpenAI, Meta, and Google to develop foundational models. AI services companies would provide API services by fine-tuning or developing RAG-style pipelines around AI foundational models to serve their customers. Lastly, does the spike in job applications indicate an impending bursting of the bubble like the dot com bubble? I feel so, YES. But let’s wait and watch it… Shout Out Hope that was useful. If you liked this article, why not follow me on Twitter for research updates, video tutorial links, and new AI tools from top AI labs? Also, please subscribe to my YouTube channel where I explain AI concepts and papers visually. This week, I am taking a slight detour from my usual articles that explain AI Concepts or coding tutorials. Yes, it's a philosophical, nostalgic tour down the road I was fortunate enough to tread all along. One of the rare privileges of starting in a field quite early is that you see the field grow like your baby. When I started my Masters in Computer Vision a decade ago, a CS graduate friend of mine asked, “What does computer vision even mean?” It's a trivial question today with enough videos of computer vision tasks like segmentation (see below) floating around the internet. A result of image segmentation — a Computer Vision task that was all over the internet when deep learning took off a decade ago. So, what has changed in the last decade? There are always two sides to a coin. Also, growth can be both good and bad. Closed Community — Small Was Beautiful Closed Community — Small Was Beautiful They say that people in Iceland are somehow related to each other and if someone is beyond your second cousin, you are good to marry them. Likewise, the AI community was close-knit. Gone are those feelings of a tight-knit community. When interacting with anyone from research you generally knew which group they belonged to. You somehow would relate to and respect their work as you read and appreciated their impact on the field. As the field is blowing up endlessly, it's becoming increasingly difficult to even say that you know someone’s work, forget about them. There are too many new names and novel branches that it's even challenging to keep track of. As another example, PyTorch was in its infancy. The Slack community was so small and helpful that the developers of PyTorch directly responded to questions we had about using the library. This encouraged learning more and more about it. Today, the framework is so mature, and new frameworks like LangChain and Llamaindex have propped up. The focus is all on LLMs over any other branch of AI. Hardware Luxury Hardware Luxury Those days, we used to train a deep neural network on a single GPU to create something impactful. Most work published in premier conferences like CVPR, NeurIPS, and ICML could be trained and replicated on a single 8 GB GPU machine or in the worst case, a single machine with 4 GPUs in rare cases. An image of the latest GPU clusters used by Tesla for AI (source: https://www.hpcwire.com/off-the-wire/tesla-to-roll-out-advanced-supercomputer ) https://www.hpcwire.com/off-the-wire/tesla-to-roll-out-advanced-supercomputer https://www.hpcwire.com/off-the-wire/tesla-to-roll-out-advanced-supercomputer I vividly remember how happy I was to buy a single commodity GPU with just 8 GB of RAM to participate in Kaggle competitions. Some of the winning solutions from Kaggle Grandmasters trained deep learning models on a single GPU machine at home. Today’s AI world needs a GPU cluster to train foundational models. Even fine-tuning these models needs 24GB GPUs which are beasts, expensive, and affordable only by businesses with an “AI budget.” Skillset Demand Skillset Demand Those were days when the field did not yet get the AI wrapper around it. Recruiters were given this alien task of hunting for “deep learning” engineers. Recruiters and start-up founders were hunting for deep learning experts through all channels. It was a norm to receive regular messages on LinkedIn asking me to join their team as a deep learning engineer. A screenshot of a job advert for a Machine Learning Engineer posted only 5 days ago receiving over 100 applicants! The current situation is that openings for “Machine Learning Engineer” get over 100 applications on LinkedIn well within a day of advertising. Take a look at the screenshot above if you find it hard to believe. The question remains as to how relevant the applicant's skills are to the job spec. But the market is saturated with skills — quite quickly! Sub-skills — Operations and Architecture Growth means more diversity and opportunity. There are new roles like ML-ops, LLM-ops, and ML-architects cropping up. Gone are those days of tiny, single-model files (< 1 GB). The growth of models both in size and capabilities has given rise to new skills in deploying and maintaining them. Moreover, training and deployment of models are being automated with tools such as MLFLow. The cloud infrastructure for training needs to be sophisticated enough. All this has given rise to full-time roles with dedicated responsibilities. Bye Bye ML Engineer, Hello AI Engineer The most fun in working in AI is writing the model architecture, and training the model from scratch using our in-house data. Though this involves pre-processing the data a lot, training models, and visualizing the training results used to be a lot of fun. There was/still is a specialized role for this called Machine Learning (ML) Engineer. The development of foundation models from tech giants is redefining this role. As model sizes grow, training budgets are enormous. As a matter of fact, the cost of training the LLama 2 model was $20m for Meta. Clearly start-ups or organizations trying to adopt AI do not wish to throw away this sum of money. It is now established that foundation models are for the tech giants to develop, with the exception of some companies like Mistral and Anthropic. Sadly, this means the ML Engineer role is getting cast into an AI Engineer role. The ML Engineer role was all about developing model architectures, training, and evaluating them. The new AI Engineer role mostly involves developing APIs or invoking APIs provided by Tech giants (OpenAI, Meta, and Google) to prompt the foundation models. In rare cases, it involves fine-tuning these foundation models. But companies have the option to build RAG pipelines or use the foundation models “as-is” over fine-tuning them. Conclusion To conclude, I see this as a slow amalgamation of the Software engineering and Machine Learning roles. The line between software engineers and deep learning experts is fading. So, in the years to come, software engineers will be AI engineers who work alongside foundation models, both to write code and to solve customer needs. Furthermore, in the coming years, companies will be grouped into two categories — the AI product and AI services companies. AI product companies would again be OpenAI, Meta, and Google to develop foundational models. AI services companies would provide API services by fine-tuning or developing RAG-style pipelines around AI foundational models to serve their customers. Lastly, does the spike in job applications indicate an impending bursting of the bubble like the dot com bubble? I feel so, YES. But let’s wait and watch it… Shout Out Hope that was useful. If you liked this article, why not follow me on Twitter for research updates, video tutorial links, and new AI tools from top AI labs? If you liked this article, why not follow me on Twitter for research updates, video tutorial links, and new AI tools from top AI labs? Twitter Twitter Also, please subscribe to my YouTube channel where I explain AI concepts and papers visually. Also, please subscribe to my YouTube channel where I explain AI concepts and papers visually. YouTube channel YouTube channel

Gone Are Those Days of AI

AI এর সেই দিনগুলি চলে গেছে

Die Zeiten der KI sind vorbei

Atrás quedaron esos días de la IA

Fini l’époque de l’IA

एआई के वे दिन चले गए

AIの時代は終わった

AI의 시대는 지났다

Longe vão aqueles dias de IA

Прошли те времена искусственного интеллекта

Yapay Zekanın O Günleri Geride Kaldı

Đã qua rồi những ngày của AI

人工智能时代已经一去不复返

Stories, News & Mentions about VIVIDLY

Vividly’s Stories on HackerNoon