Researchers show that a type of AI known as a large language model often outperformed physicians at diagnosing complex and potentially life-threatening conditions, including decreased blood flow to the heart, even in the fast-moving stages of real ER care when information is limited.

In early ER cases, the model identified the correct or a very close diagnosis in about 67% of cases, compared with roughly 50% to 55% for physicians. And the technology is only getting better.

Before antiretroviral (ARV) drugs started to become widely available in KwaZulu-Natal in 2005, there was “kind of the perfect storm,” with several unusual factors coalescing to drive a devastating epidemic, says Philip Goulder, an immunologist at the University of Oxford who led the study, which appears today in the Proceedings of the National Academy of Sciences. HIV had made few inroads into South Africa until the early 1990s, when an epidemic exploded in the heterosexual population, infecting about 40% of pregnant women in KwaZulu-Natal. (That astonishingly high prevalence persists today.) Because of a mix of genetics, limited health care, and possibly the viral subtype in circulation, infected people developed AIDS—when the destruction of the immune system threatens survival—exceptionally quickly, within about 4.5 years versus 10 years in North America.

Other studies have shown how infectious diseases, including malaria and tuberculosis, have altered the human genome. But those changes took thousands of years. “That’s what is quite exciting about this: You can see how rapidly evolution actually can occur,” Goulder says.

Similar evolutionary forces may have been at work in North America and Europe, but they are more difficult to see—and less likely to affect future generations. HIV prevalence in those regions is below 1%, and the hardest-hit group is men who have sex with men. “They are generally not a population that’s leaving behind as many offspring,” Worobey notes.