US Air Force denies that AI drone killed operator in simulation – update

By admin On Jun 2, 2023

Spread the love

The alleged incident in which an AI-controlled military drone shot at the operator in a simulation did not occur, according to the US Air Force. It would be an anecdote; the simulation would never have actually been carried out.

Source: USAF

Colonel Tucker Hamilton, the head of the AI Test and Operations Branch, said according to two attendees from the Royal Aeronautical Society at the Future Combat Air and Space Capabilities conference last month that a simulation produced a “very unexpected” result after an AI-powered drone chose to shoot at the human operator. An Air Force spokesperson says told Insider that such a simulation has never actually been run and that it is an ‘anecdotal description taken out of context’. The Air Force does not provide any further context about what exactly happened.

According to the RAS, Hamilton said at the conference: “The system began to realize that in some cases, despite a positive identification of a threat, the human operator was giving the order not to attack, while the AI was getting ‘points’ for killing of threats. It then killed the pilot because this person prevented the drone from achieving its goals.” According to the umbrella US Air Force, this would therefore be an anecdote and not an actual simulation.

The colonel further says that the AI model has been tightened up so that it is no longer allowed to shoot at the operator. According to him, the artificial intelligence therefore decided to destroy the communications tower that facilitated contact with the combat drone. It is not clear from the blog post whether the possible hypothetical AI was subsequently trained not to destroy its own equipment and personnel. It is also not clear why the drone had the ability to destroy American targets in the first place.

Based on the simulations, the AI Test and Operations department of the US Air Force recommends that the application of the technology be carefully examined. “We can’t talk about artificial intelligence, machine learning and autonomy without also talking about ethics and AI.”

Updated, 9:02 am: The US Air Force statement regarding the possible misinterpretation of the colonel’s comments has been added to the article. Thanks to Mighteymike.

Update, 9:55 am: The approach of the article has been further refined to better highlight the position of the US Air Force.