Retired U.S. Army Gen. David Petraeus said Friday he expects large language models to reach a “great Nobel laureate” level by year-end 2026, without naming a discipline or benchmark.
That date is earlier than the two-year horizon Petraeus gave at the Aspen Security Forum on July 18, 2025. There, he described models as comparable to a very good graduate student and outlined more advanced benchmarks on the way to the Nobel comparison. He said that level could be reached “within two years.”
In a March 6, 2025, submission to the White House Office of Science and Technology Policy (OSTP), Anthropic set out a broader scenario: powerful AI systems could emerge from late 2026 into early 2027. The company said their potential intellectual capabilities could match or exceed those of Nobel Prize winners in most disciplines. Its forecast concerned possible systems generally; it did not confirm that a particular model had reached that level. Petraeus did not specify a discipline, test or benchmark for his comparison, making “Nobel-level” a qualitative label.
For military use, Petraeus outlined an agent that could combine radar and acoustic readings into an air-defense picture and potentially select an interceptor drone. He said such systems could give commanders continuous analysis and help speed decisions, while raising ethical concerns as software moves closer to directing action. Petraeus summarized the human role as “The human is on the loop.”
In an Oct. 1 memorandum, Acting Secretary of the Army Adam R. Telle ordered formation of the U.S. Army Futures and Autonomous Systems Command, the Army said Oct. 2. The Army said Telle’s order came after Secretary of War Pete Hegseth announced an Autonomous Warfare Command on Sept. 30.
