[NEW] Sakana AI, a Japanese startup founded by former Google researchers, raises 20 billion yen to become a unicorn in just one year since its founding

Sakana AI, an AI development startup founded in Japan by an AI researcher from Google, will soon raise about 20 billion yen in funding from US venture capital firms, according to a person involved in the negotiations.

If the funding goes ahead, Sakana AI's market capitalization will reach 180 billion yen, including existing capital. Startups with a market capitalization of more than $1 billion (about 157 billion yen) are called "unicorns," promising unlisted companies.

Sakana AI was founded by Lion Jones and David Ha, both former AI researchers at Google, and Ren Ito, a former employee of the Japanese Ministry of Foreign Affairs, and currently has 15 employees.

When asked why he founded an AI company in Japan, Ha said, "I worked at Google for six and a half years, and Lion worked at Google for 12 years. At Google, we were blessed with an environment where we could work on research."

After that, Google began to focus more on engineering than on research and development, and he said, "I wanted to do AI research. So I thought I would set up a research and development center like the one I had at Google."

Jones is a well-known figure in the AI ​​industry. He is one of the developers of the language processing model called "Transformer," which led to the explosive spread of generative AI, and published a paper on it six years ago.

Jones believes that there are problems with current AI technology. In addition to the need for large amounts of data, he points out that it requires constant maintenance every time a problem occurs, which is labor-intensive and costly.

So Jones and Ha seem to have thought of creating a system that mimics the natural world, like "a school of fish swimming to avoid predators."

Sakana AI, based in Japan and aiming for global expansion

Sakana AI is a project that combines multiple AIs with different characteristics to create better AI. By extracting the superior points of multiple generative AIs, it is possible to develop AI models effectively in a short period of time. Therefore, even if a problem occurs, it can be flexibly dealt with and efficiency is increased.

In general, developing large-scale AI requires huge amounts of data and computing resources. Due to this background, large IT companies that have financial resources and provide existing IT services are considered to have an advantage.

However, Sakana AI aims to create advanced AI by combining existing AIs in the same way that small fish form schools. Using this method, they have actually succeeded in developing a "large-scale language model that solves mathematical problems in Japanese."

Sakana AI aims to expand the users of its technology from Asia to the world from its base in Japan, and is working to realize its vision of "becoming a global company originating from Japan in the field of AI."

Ha argues that "Most of the AI ​​development bases are in the Bay Area (around San Francisco) or Beijing in the United States. We should not want a few companies or governments to control the future of important technologies such as AI. Japan is between the United States and China, so Japan should play an important role in technological development."

Analysts evaluate Sakana AI's potential corporate capabilities to exceed those of OPENAI, which develops ChatBPT, the world's leading generative AI, and Google. It can be said that the Japanese unicorn has set its sights on the future of generative AI and is moving to lead the world.

Image: Shutterstock

Related articles

Event report of the global conference "CREATE" hosted by Honda Keisuke's "X&KSK"

Bitwise points out that crypto assets and AI will increase global GDP by $20 trillion by 2030