Tokyo [Japan], May 23: Nikkei Asia reports that Japanese researchers will use the country's Fugaku supercomputer to create an artificial intelligence (AI) application similar to ChatGPT .
According to the plan announced on May 22, the technology will be developed in the current fiscal, with the intention of making it available for free use in fiscal 2024. Tokyo Institute of Technology has collaborated with Tohoku University, Research Institute Research Riken and Fugaku co-developer Fujitsu on the project.
They will develop a large language model (LLM) for generative AI, focusing on the Japanese language. The model will take open source data from Wikipedia and other sources, with the goal of producing more accurate results in Japanese.
The project will work with partner CyberAgent, a digital advertising agency in Tokyo developing its own generative AI.
Generating AI is on the rise in Japan. CyberAgent last week announced a large language model (LLM) that uses up to 6.8 billion parameters. Tokyo Tech and its partners will build an integrated LLM of about 100 billion parameters.
GPT-3, OpenAI 's ChatGPT-compatible language model uses 175 billion parameters.
It takes a powerful computer to build an LLM, but no private sector in Japan has such a world-class computer. The Japanese government-backed firm Riken currently operates the Fugaku, one of the world's fastest supercomputers.
Developing new generative AIs has become a priority. The upcoming LLM will be offered to both businesses and research institutions. This model will be released in fiscal 2024, starting April 2024.
Source: ThanhNien Newspaper