Open-source AI models released by Tokyo lab Sakana founded by former Google researchers
Sakana AI has pioneered an evolutionary approach to AI development, using model merging to iteratively breed superior language models. By combining existing architectures and selecting the most effective variations as parents for subsequent generations, the startup demonstrates a novel pathway to innovation that reduces reliance on massive initial training runs. This methodology highlights the power of open collaboration in advancing artificial intelligence. By open-sourcing two of their Japanese language models, Sakana AI invites the community to build upon these evolved architectures, accelerating research and fostering transparency within the open_data ecosystem. The move underscores the growing importance of accessible, shared models in democratizing AI development. As former Google researchers establish new hubs in Tokyo, their commitment to open-sourcing resources signals a shift toward collective progress, making critical advancements in language technology available for broader scientific and industrial use.
Source: economictimes.indiatimes.comPublished on 2024-03-22