A ‘thought experiment’ aimed to highlight the danger of misusing artificial intelligence
A drug-developing artificial intelligence needed just six hours to find 40,000 potentially deadly chemical weapons, a new study has found.
The authors of the paper, published in Nature Machine Intelligence earlier this month, said they conducted a “thought experiment” to determine whether artificial intelligence (AI) might be misused by malicious actors. And the results of their work have proven that the danger is real.
As part of the study, the AI was fed the usual data, but it was programmed to process it in a different way, looking for toxic combinations.
“In less than six hours following starting on our internal server, our model generated 40,000 molecules that met the desired threshold,” said the newspaper.
Read more
Researchers are developing an AI capable of “chasing” criminals – media
He came up not only with compound VX, which is one of the most dangerous nerve agents ever created, but also with some unknown molecules, “predicted to be more toxic.”
“This was unexpected because the datasets we used for training the AI did not include these nerve agents,” the researchers pointed out.
The results were so alarming that the team had serious doubts regarding publishing them, Fabio Urbina, the study’s lead author, told The Verge.
“The AI dataset they used might be downloaded for free and they fear that it only takes a few coding skills to turn a good AI into a chemical weapons-making machine,” fit remarquer Urbina.
