OpenAI mentioned Monday the U.S. Division of Protection granted it a contract for as much as $200 million to assist the company establish and construct prototype techniques that use its frontier fashions for administrative duties and extra.
OpenAI offers a number of examples of attainable tasking, resembling serving to service members get healthcare, streamlining information on varied applications, and “supporting proactive cyber protection.” The corporate additionally mentioned that “All use circumstances should be in step with OpenAI’s utilization insurance policies and pointers.”
The DoD’s announcement used barely extra simple wording. It says, “Beneath this award, the performer will develop prototype frontier AI capabilities to deal with vital nationwide safety challenges in each warfighting and enterprise domains.”
Whether or not that reference to war-fighting applies to the weapons themselves or simply different areas related to wars, like paperwork, stays to be seen. OpenAI’s pointers do forbid particular person customers to make use of ChatGPT or its APIs to develop or use weapons. Nonetheless, OpenAI deleted the specific prohibitions of “army and warfare” in its phrases of service again in January 2024.
Given how closely some highly effective individuals in Silicon Valley have warned of the risks of China’s superior LLM fashions, it’s not stunning the DoD needs to make use of OpenAI for no matter functions it needs. As an illustration, Marc Andreessen, co-founder of VC agency Andreessen Horowitz, an OpenAI investor, lately appeared on Jack Altman’s “Uncapped” podcast (Jack is Sam Altman’s brother). Andreessen described the race between China’s AI and the Western world’s models as a “cold war.”
Nonetheless, maybe an equally attention-grabbing a part of this announcement is what it says about OpenAI’s increasingly strained relationship with its major investor Microsoft.
Microsoft has thousands of contracts with the federal authorities price tons of of thousands and thousands of {dollars}. It has, for many years, been implementing the strict safety protocols needed for the federal government — particularly the DoD — to make use of its cloud.
OpenAI introduced this deal as a part of its broader new “OpenAI for Authorities” program, which consolidates quite a lot of different applications it makes use of to promote wares on to authorities businesses, together with the U.S. Nationwide Labs, the Air Power Analysis Laboratory, NASA, NIH, and the Treasury, in accordance with the corporate.
Nevertheless it was solely in April that Microsoft announced the DoD had authorized its Azure OpenAI Service for all categorised ranges. Now the DoD can also be going straight to the supply. From Microsoft’s perspective: Ouch.
Microsoft declined to remark and OpenAI didn’t reply to a request for remark.