• AI KATANA
  • Posts
  • OpenAI co-founder Ilya Sutskever announces rival AI start-up

OpenAI co-founder Ilya Sutskever announces rival AI start-up

Also: TikTok parent ByteDance to spend $2.1bn to develop AI hub in Malaysia

Good morning,

In today’s news, Ilya Sutskever, a co-founder of OpenAI, has launched a new AI venture, Safe Superintelligence Inc., focusing on creating safe and advanced AI. Google is partnering with the University of Tokyo to use AI in addressing regional challenges in Japan, starting with Osaka and Hiroshima. A study highlights discrepancies in the openness of AI models labeled as “open source,” urging stricter transparency standards. Asian family offices are keen on early-stage AI investments, as discussed by Carman Chan of Click Ventures. ByteDance is investing $2.1 billion to establish an AI hub in Malaysia, further cementing the region’s tech growth. Snapchat introduces a novel AI feature that turns user prompts into custom augmented reality lenses, enhancing digital interactivity. Lastly, France leads Europe in generative AI funding, though London hosts the most AI startups, showcasing the dynamic growth of AI across the continent.

Sliced:

  • 🥊 OpenAI co-founder Ilya Sutskever announces rival AI start-up

  • 🇯🇵 Google, Univ. of Tokyo to launch AI project to solve regional issues

  • 1️⃣ Not all ‘open source’ AI models are actually open: here’s a ranking

  • 🌏 Asian Family offices in search of emerging managers to make AI bets

  • 🇲🇾 TikTok parent ByteDance to spend $2.1bn to develop AI hub in Malaysia

Ilya Sutskever, a co-founder of OpenAI, has launched a new AI start-up named Safe Superintelligence Inc. (SSI) in the wake of his departure from OpenAI after an attempted coup against CEO Sam Altman. SSI aims to focus solely on developing safe superintelligence, an advanced form of AI designed to surpass human cognitive abilities while prioritizing safety and security. Sutskever is joined by former OpenAI employee Daniel Levy and entrepreneur Daniel Gross, who brings significant industry experience and investment background. Unlike many tech ventures, SSI will operate without the pressure of immediate revenue generation, allowing a concentrated effort on its mission. This initiative follows internal conflicts at OpenAI regarding leadership and safety priorities, culminating in Sutskever’s resignation. The launch of SSI underscores a growing trend of former OpenAI executives pursuing independent ventures to develop AI technologies with a focus on safety, echoing the mission of other offshoots like Anthropic.

Google is partnering with the University of Tokyo to launch an innovative AI initiative aimed at addressing regional challenges in Japan. This project, set to begin in the Osaka and Hiroshima prefectures, leverages generative AI to tackle specific local issues such as job market mismatches in Osaka. By utilizing vast online data, the AI models will suggest suitable job roles for seekers that they might not have considered, and potentially offer optimized career paths. The initiative, led by AI expert Professor Yutaka Matsuo, will involve University of Tokyo students collaborating with Google engineers. The project’s scope includes enhancing healthcare on remote islands and automating agricultural tasks. If successful, these solutions could be implemented nationwide by 2027. This collaboration highlights AI’s potential to provide tailored solutions to community-specific problems, positioning Japan as a leader in practical AI applications.

A recent analysis highlights a significant discrepancy in the transparency of AI models labeled as “open source.” While technology giants like Meta and Google claim openness for their models, researchers argue that these claims often fall short. The study, conducted by Mark Dingemanse and Andreas Liesenfeld, evaluated 40 AI models, revealing that smaller players tend to offer more genuine openness compared to big tech firms. The research emphasizes that true open-source AI should allow unrestricted access to code, training data, and model weights, enabling replication and thorough examination. However, many prominent models only provide limited access, coining the term “open-washing” to describe these practices. For instance, Meta’s Llama and Google’s Gemma are categorized as “open weight,” where the trained models are accessible but the underlying code and training data remain concealed. This partial openness hinders scientific progress, as it prevents researchers from fully understanding or modifying the models. The findings are particularly relevant as the European Union’s upcoming AI Act will impose less stringent regulations on models classified as open source, raising the stakes for accurate definitions of openness. The study advocates for a more rigorous standard of transparency to foster accountability and innovation in AI development.

A prominent family office based in Hong Kong and Singapore, represented by Carman Chan of Click Ventures, is actively seeking to increase its exposure to AI through early-stage venture capital investments. At a recent event in Hong Kong, Chan emphasized that 2024 presents a prime opportunity for family offices to make strategic AI bets, citing the potential for these technologies to disrupt and transform traditional business models. With existing investments in 12 managers and a focus on seed-stage VC funds, the family office is looking to capitalize on emerging AI trends before valuations become prohibitively high. Chan highlighted the importance of selecting fund managers with deep technical knowledge and a clear understanding of how AI can drive business innovation. She advised that family offices should build diversified portfolios to manage the high risks associated with early-stage investments. The family’s strategy underscores a broader trend of significant interest and commitment from institutional investors, particularly from North America, to support and accelerate the growth of emerging managers in the AI sector.

ByteDance, the parent company of TikTok, is making a significant investment in Malaysia with plans to spend $2.1 billion to establish an AI hub in the region. This initiative aligns with Malaysia’s growing prominence as a center for AI and data-driven technologies, following major investments by tech giants like Google and Microsoft. Announced by Malaysia’s Investment, Trade, and Industry Minister Tengku Zafrul Aziz, ByteDance’s investment will include the development of a large data center in Johor Bahru, a city now recognized as Southeast Asia’s fastest-growing data center market. ByteDance’s move is part of its broader strategy to leverage AI across various products, including its influential TikTok algorithm and AI-driven music apps. The Malaysian government is also backing AI growth through initiatives like the AI Sandbox 2024 Pilot Programme, aiming to foster 900 AI startups by 2026. This investment by ByteDance not only underscores its commitment to AI but also reinforces Malaysia’s role as a critical player in the global tech landscape.

🛠️ AI tools updates

Snapchat is unveiling a cutting-edge on-device AI model that transforms user prompts into customized augmented reality (AR) lenses. This new feature will allow users to see real-time changes to their surroundings and even their clothing based on simple text prompts, like “50s sci-fi film,” which could instantly alter the environment and attire in line with the theme. This innovative technology, which will be integrated into Snapchat’s Lens Studio, enables creators to generate unique AR effects by inputting prompts or images, significantly enhancing their creative capabilities. Among the new tools being introduced are features that apply realistic transformations over a user’s face, body, and environment, allowing for dynamic and immersive AR experiences. Additionally, creators will have the ability to generate 3D assets and custom face masks from text or images, facilitating the production of intricate and personalized digital content. These advancements reflect Snapchat’s commitment to leveraging AI to push the boundaries of interactive media, following its recent introduction of an AI chatbot and AI-generated snaps for users.

💵 Venture Capital updates

France is leading Europe in the funding of generative AI startups, securing $2.29 billion in investments, the highest in the region. This positions France ahead of other European countries and Israel in this burgeoning sector. Key players like Mistral AI and “H” have attracted substantial investments, with rounds reaching up to $640 million. Notable in Paris are initiatives like Hugging Face, a hub for machine learning models, which raised $235 million, and the research-focused Kyutai, dedicated to open-source AI models. Despite France’s financial leadership, London dominates in the number of generative AI startups, hosting nearly a third of Europe’s GenAI companies. Other significant cities include Tel Aviv, Berlin, and Amsterdam. The report by Accel and Dealroom highlights the critical role of strong educational institutions and the presence of major tech companies in fostering these ecosystems. Universities and tech giants serve as “founder factories,” contributing to the rise of new AI ventures. As Europe and Israel continue to expand their generative AI landscapes, the region is poised for rapid development and increased competition in the global AI arena.

🫡 Meme of the day

⭐️ Generative AI image of the day