Sakana AI: Tokyo-based AI Startup Innovates with Model Merging

Sakana AI: Tokyo-based AI Startup Innovates with Model Merging

Tokyo-based artificial intelligence startup, Sakana AI, is making waves in the tech world with its innovative approach to AI models. Co-founded by two former Google researchers, the company has developed AI models using a technique called “model merging,” inspired by evolution and natural selection. This method involves combining existing AI models to create a new one, and then breeding several generations of models to identify the most successful ones.

Sakana AI is now releasing three Japanese language models, with two of them being open-sourced. David Ha, the founder of Sakana AI, explained, “We wanted to create a collaborative approach to AI development, where anyone can contribute to the growth and improvement of these models.” This open-source approach aims to foster innovation and collaboration within the AI community.

The founders of Sakana AI, David Ha and Llion Jones, are no strangers to groundbreaking AI research. Jones was an author on Google’s influential 2017 research paper “Attention Is All You Need,” which introduced the “transformer” deep learning architecture. This architecture was the foundation for the viral chatbot ChatGPT, sparking a race to develop generative AI products. Ha, on the other hand, was previously the head of research at Stability AI and a researcher at Google Brain.

The success of Sakana AI and other AI startups founded by former Google researchers demonstrates the growing trend of experts leaving big tech companies to pursue their own ventures. Investors have shown significant interest in these startups, pouring millions of dollars into their endeavors. Other ventures, such as Character.AI and Cohere, have also received substantial funding and aim to make their mark in the AI industry.

Sakana AI has ambitions beyond just creating AI models; it wants to position Tokyo as a leading AI hub, following in the footsteps of OpenAI in San Francisco and DeepMind in London. By releasing open-source models and encouraging collaboration, the company hopes to attract top talent and facilitate the growth of the AI community in Tokyo.

This latest development comes on the heels of Sakana AI’s successful seed financing round in January, which raised $30 million led by Lux Capital. The funding will enable the company to scale its operations and continue advancing its AI models.

As the field of artificial intelligence continues to evolve, startups like Sakana AI are pushing the boundaries of what is possible. With their innovative approach to model development and dedication to fostering collaboration, they are driving the progress of AI technology and positioning themselves as leaders in the industry. Tokyo’s rise as an AI powerhouse seems imminent, thanks to the efforts of companies like Sakana AI.


Written By

Jiri Bílek

In the vast realm of AI and U.N. directives, Jiri crafts tales that bridge tech divides. With every word, he champions a world where machines serve all, harmoniously.