AI Generates Deadly Drug

Acinetobacter baumannii, a resistant bacteria that may infect wounds and cause pneumonia, has been targeted by researchers in Canada and the US using deep learning. The BBC cites a report in Nature Chemical Biology that details how the researchers utilised training data to determine how well-known medications affected the hardy bacteria. Then, with no information on how powerful they would be against the germ, 6,680 chemicals were projected by the learning system.

The programme whittled down the list to 240 promising prospects in an hour and a half. Nine of these passed the test in the lab, and one of them—now known as abaucin—was incredibly strong. Even if doing laboratory tests on 240 chemicals seems like a lot of work, it is preferable to testing approximately 6,700.

It’s interesting to note that the new antibiotic appears to work just against the intended microorganism, which is a benefit. Given the nature of drug testing, it may not be accessible to people for some time. However, this is still a fantastic illustration of how machine learning can support human intelligence, allowing researchers and others to concentrate on what matters most.

A weapon against Acinetobacter baumannii would be greatly appreciated, as the WHO lists it as one of the main superbugs endangering the planet. It is possible to anticipate that this method will significantly reduce the time needed to develop new medications. It also makes you question if there are any other industries where AI tools may quickly eliminate alternatives, freeing up human attention for the more promising applicants.