Pentagon’s bowels are tested by promising, but ambiguous technologies of artificial intelligence. The goal is to adapt the possibilities of neural networks for the needs of the US Army.
According to Defensescoop, the Ministry of Defense has concluded a contract with the specializing in AI startup Scale AI. Its task is to create tools for testing and evaluating the work of powerful language models capable of analyzing and generating text.
We are talking about a complex platform that will allow the Pentagon to deeply and comprehensively study the possibilities of neural networks. It will issue feedback for the military, as well as generate special test data for testing AI on specific tasks. For example, for the analysis of intelligence, planning operations, processing reports from hot spots.
However, the main obstacle is the tendency of AI to generate false materials, the so -called “hallucinations”. The Pentagon hopes that Scale AI tools will help to identify and prevent such threats in advance.
Last year, to study the military use of generative AI, a group of Task Force Lima was created under the leadership of Martell. But so far in a number of units, including space forces, it is prohibited from using ChatGPT systems due to the risk of secret information leaks.
“We are obliged to introduce these technologies responsibly, given the threats of security and potential abuse,” Martell summed up.
So the Pentagon is looking for a balance between the colossal potential of neural networks for the army and dangers luminous in them. And how successful the current tests are depending on how quickly generative AI can go from laboratories to real combat operations.