After initial tests since spring, the Japanese government is apparently ready to integrate AI even more deeply into administrative processes.
According to Nikkei Asia, the government is cooperating with Microsoft, which will provide Meta's language models in addition to OpenAI's GPT-4. For this purpose, Microsoft has installed "powerful equipment with high processing power" in its data centers in Tokyo and Osaka.
The on-premises data processing is intended to address security and privacy concerns and allow other companies, which Microsoft is increasingly targeting, to work with the system. Microsoft also wants to improve the accuracy of the Japanese language.
AI is expected to help with various administrative tasks, such as drafting answers to parliamentary questions, taking minutes, and helping to analyze government statistics. Previously described use cases include automatically updating websites or preparing documents in FAQ format.
Japan's Digital Ministry is planning a budget of two million US dollars for the first year of operation. Testing is scheduled to begin in the fall. Several institutions, including the Ministry of Economy, Trade and Industry, are considering using the technology.
Did Meta help Microsoft close the deal?
The fact that Microsoft also sold Meta language models is interesting in the sense that it may have prevailed over OpenAI in this way: OpenAI CEO Sam Altman had already spoken with the Japanese government in April and reached initial agreements for testing. Now, Microsoft seems to have won the bid, possibly with the argument that it can provide more technology than just the OpenAI API.
The Wall Street Journal reported in mid-June on tensions between OpenAI and Microsoft, in part because the two companies are fighting for some of the same customers. Microsoft owns 49 percent of OpenAI, according to the WSJ.