Palantir CTO: The Pentagon Has Already Replaced Anthropic’s AI
Shyam Sankar, Palantir's chief technology officer, made several notable statements on CBS's Face the Nation on Sunday.
Pentagon's AI Transition
According to Sankar, the Pentagon has already replaced Anthropic's AI models, despite a dispute over safeguards and labeling the company a supply chain risk. No operational impact was mentioned. The Department now utilizes models from xAI, OpenAI, Google, and open-weight models like Nvidia’s Nemotron.
AI Safety Concerns
Sankar attributed much of the fear around AI to effective altruism, describing it as a millenarian doomsday cult meets messianic complex. He cited researcher Jacob Coxon, who assigns a 10% chance to AI ending humanity, as an example of what he termed a theological critique rather than a scientific one.
Sankar also made an unexpected claim, stating that many AI researchers have taken up smoking as a way to signal their belief in AI.
Military Targeting
Sankar described military targeting as a process, not a choice made by AI. He noted that executing 1,000 strikes in a single day, which once required two weeks of planning by 50 people, can now be accomplished by one person in four hours with AI assistance.
Iran School Strike
Regarding the US strike on a school in Minab, Iran, Sankar disputed the reporting that suggested AI played a primary role, attributing it to human error. He emphasized the need to wait for the final report. Palantir has since implemented additional AI agents to double-check work in a part of the process they do not own.
Regulation and Liability
Sankar argued that existing agencies like the FDA should regulate how people apply AI, and that AI companies should be held liable for defective products. He also expressed Palantir's non-participation in the White House AI accord, attributing this to the fact that Palantir does not make AI models.