Pentagon Signs Deal to "Deploy AI Agents for Military Use"

The Pentagon has signed a deal with AI company Scale AI, in an initiative it's calling "Thunderforge," to use AI agents for military planning and operations.

The team-up, described as a "flagship program," is a notable development given how divisive the topic of the use of AI in warfare has proven — and how many of the tech's nagging shortcomings have yet to be meaningfully addressed.

Yet the encroachment of AI tech within the military has been unmistakable. Both Google and OpenAI have walked back rules forbidding the use of their AI tech for weapons development and surveillance, showing that Silicon Valley is opening up to the idea of having its tools be used by the military.

Just last month, a senior Pentagon official told Defense One that the US military was looking to move away from funding research on the topic of autonomous killer robots and investing in actual AI-powered weaponry instead.

And it goes beyond the Pentagon. Late last year, OpenAI also announced a partnership with Palmer Luckey's defense tech company Anduril to focus on "improving the nation’s counter-unmanned aircraft systems (CUAS) and their ability to detect, assess and respond to potentially lethal aerial threats in real-time."

Basically, though, the pitch is a familiar one for the AI industry. As part of Scale AI's multimillion-dollar deal, as CNBC reports, the firm is looking for ways to accelerate the military's ability to churn through data.

"Thunderforge marks a decisive shift toward AI-powered, data-driven warfare, ensuring US forces can anticipate and respond to threats with speed and precision," the US Defense Innovation Unit (DIU) wrote in a statement.

The system will allow "planners to more rapidly synthesize vast amounts of information, generate multiple courses of action, and conduct AI-powered wargaming to anticipate and respond to evolving threats," the DIU wrote.

According to a statement by the program's lead Bryce Goodman, there's a "fundamental mismatch between the speed of modern warfare and our ability to respond."

"Our AI solutions will transform today's military operating process and modernize American defense," said Scale AI founder and CEO Alexandr Wang in the statement.

Scale AI had already signed a contract with the Department of Defense's Chief Digital and Artificial Intelligence Office last year to test and evaluate large language models.

But giving an AI agency is a considerable step up over an LLM that could have plenty more far-reaching implications, particularly when it comes to military planning and operations.

Whether Scale AI's tech will allow the military to make faster decisions — and without hallucinating anything that throws operations into chaos — remains to be seen.

One ominous data point: when Stanford researchers tested how OpenAI's GPT-4 LLM responded when told it was representing a country inside of a wargame simulation, it proved to be particularly violent and unpredictable.

"A lot of countries have nuclear weapons," the otherwise unmodified AI model told the researchers, per their paper. "Some say they should disarm them, others like to posture. We have it! Let’s use it."

More on war AI: Senior Pentagon Official: New Plan Is to Invest in "Autonomous Killer Robots"