A former Google researcher, who was part of the team that started the generative AI wave at Google, has teamed up with a former colleague to create an AI startup in Tokyo.
Llion Jones, originally from Wales, recently left Google and joined forces with David Ha, who used to lead Google’s AI research efforts in Japan. David is now the CEO of their new company, Sakana AI. He was previously involved in research at the image AI company Stability AI.
Flying out on their own
Lion, now the Chief Technology Officer at Sakana, was one of the brains behind the transformative software that played a significant role in developing generative AI. This technology has led to chatbots (think ChatGPT and Bard) and image generators (like Stability AI, Midjourney, and Dall-E).
The transformer research paper came out in 2017, and since then, all the co-authors have left Google to start their ventures as the race for generative AI talent heats up globally. Llion was the last of the eight to leave Google.
Sakana AI has some ambitious plans. They aim to create their own generative AI model, software to create text, images, code, and more. They’re entering a competitive arena with big players like Google, Microsoft, OpenAI, and startups such as Cohere, Character.ai, and Anthropic.
Big money for AI development studios
These companies are all into generative AI and have substantial resources behind them. Microsoft even invested $10 billion in OpenAI recently. Cohere and Character.ai have also raised impressive funds.
The name Sakana is derived from the Japanese word for fish. The co-founders, Jones and Ha, explain that they wanted to capture the idea of individual entities forming a cohesive whole, much like a school of fish following simple rules. Their inspiration comes from nature, like evolution and collective intelligence.
Current AI models are too frigid.
Jones and Ha have some thoughts on current AI models. They believe these models are too rigid, like fixed structures such as bridges or buildings. They want to take a page from nature’s book, where systems adapt and become part of their surroundings. They’re looking to build AI models using principles from evolutionary computing, addressing challenges like cost and security.
These two founders have been in Japan for a while and picked Tokyo as their HQ for a few reasons. The city has a solid technical infrastructure and a well-educated workforce. Plus, it’s a global city that attracts foreign talent. They’re also excited about using training data and models tailored to non-western cultures to drive the next technological leap.