AI shouldn't decide war outcomes: OpenAI CEO on military use
What's the story
OpenAI CEO Sam Altman has warned against the use of artificial intelligence (AI) in making critical military decisions. His comments come amid speculation that an AI system may have helped in a recent US operation to capture former Venezuelan President Nicolas Maduro. "I don't think AI systems should be used to make warfighting decisions," Altman told The Indian Express.
Capabilities
Altman emphasizes AI's current limitations
Altman emphasized that current AI models are not sophisticated or reliable enough for military decision-making. He said, "That said, we certainly want to support the government, and there are a lot of things we can do already." While he acknowledged future applications of AI in defense could be significant, he stressed the present limitations of these models.
Involvement
Response to AI's role in Maduro raid
When asked if AI was used in the Caracas raid, Altman said he wasn't sure but suspected it might have been used in some capacity. He suggested that using AI to analyze a large number of intelligence reports could be an effective application of the technology. This aligns with reports from The Wall Street Journal earlier this month, which suggested Anthropic's Claude AI model may have assisted the operation via data firm Palantir Technologies.
AI risks
Chris Lehane's warning on AI safety mechanisms
Altman's cautionary stance on military applications of AI echoes earlier warnings from Chris Lehane, OpenAI's Chief Global Affairs Officer. Lehane had said at the NDTV Ind.AI Summit that they build safety mechanisms into their models before public release. However, not all experts share Altman's optimism about AI technology. UC Berkeley professor Stuart Russell, a prominent advocate for AI safety, warned that unchecked AI could pose serious risks to humanity.
Military AI
Pentagon pushes for military access to AI tools
The Pentagon is pushing AI companies such as OpenAI and Anthropic to provide their tools on classified networks. This move is part of an effort to ease standard restrictions for military applications. While many AI tools are still on unclassified networks, Anthropic's Claude is one of the few that can be accessed through classified channels but with strict usage limits.