In San Jose, Tokyo-based artificial intelligence startup Sakana AI, co-founded by former Google researchers, introduced new AI models developed using an innovative method inspired by evolution. Termed “model merging,” this technique combines existing AI models and incorporates an evolutionary approach, akin to breeding and natural selection, resulting in the generation of numerous model iterations.
Evolutionary Approach to Model Development
The process of “model merging” involves combining existing AI models and applying principles inspired by evolution to create new iterations. Each generation of models undergoes evaluation, with the most successful models from each generation serving as the “parents” for the subsequent generation. This iterative process leads to the refinement and enhancement of AI capabilities over multiple generations.
Open-Sourcing and Future Plans
Sakana AI has released three Japanese language models, with two of them being open-sourced, according to David Ha, the founder of Sakana AI. Ha, along with co-founder Llion Jones, both former Google researchers, have spearheaded this initiative. Jones, known for his contribution to Google’s 2017 research paper “Attention Is All You Need,” has been instrumental in pioneering the “transformer” deep learning architecture, which has significantly influenced the development of AI-powered products like ChatGPT.
Establishing Tokyo as an AI Hub
Sakana AI aims to position Tokyo as a prominent AI hub, following the footsteps of entities like OpenAI in San Francisco and DeepMind in London. The company’s ambition is supported by substantial funding, as evidenced by its announcement in January of securing $30 million in seed financing, with Lux Capital leading the investment round. This injection of capital underscores investor confidence in Sakana AI’s vision and potential to drive innovation in the AI landscape.
Disclosure Statement: Miami Crypto does not take any external funding, or support to bring crypto news to the readers. We do not have any conflicts of interest while writing news stories on Miami Crypto.
