Researcher Warns AI Can Rapidly Develop Deadly Chemical Weapons

Much of the public discussion of the dangers of AI revolves around the “care” of AI. This is not the serious risk we face

The dangers of out-of-control artificial intelligence (AI) are sometimes misrepresented. The sci-fi version is that the AI ​​decides to take over from mere humans, like the iconic HAL 9000 in 2001: A Space Odyssey: (1968).

A more likely danger is bad actors using enormous computing power to cause damage they could not have managed on their own. Here is a sobering example:

It took the drug-developing AI less than six hours to invent 40,000 potentially deadly molecules. The researchers put the AI ​​normally used to search for useful drugs into a sort of “bad actor” mode to show how easily it could be abused at a bioweapons control conference.

All the researchers had to do was modify their methodology to seek rather than eliminate toxicity. AI has developed tens of thousands of new substances, some of which are similar to VX, the most potent nerve agent ever developed. Shaken, they published their findings this month in the journal

Intelligence of the natural machine. Justine Calma“AI suggested 40,000 possible new chemical weapons in just six hours” at The edge (March 17, 2022) The document is open access.

When The edge interviewed the paper’s lead author, Fabio Urbina, he noted,

For me, the concern was how easy it was to do. A lot of things we’ve used are free. You can go and download a toxicity dataset from anywhere. If you have someone who knows how to code in Python and has machine learning capabilities, then probably in a good working weekend they could create something like this generative model driven by toxic datasets. So that’s what really pushed us to think about publishing this document; it was such a low barrier of entry for this type of abuse.

Intelligence of the natural machine. Justine Calma“AI suggested 40,000 possible new chemical weapons in just six hours” at The edge (March 17, 2022) The document is open access.

AI has developed substances similar to VX, the most potent nerve agent known:

Keeping the lid on isn’t going to be easy. But perhaps we could start by acknowledging that humans provide the motivation. The AI ​​does not take over. It just helps them do what they want to do much faster.


You can also read: AI is not taking our jobs away from us, because it cannot do them. Computer science professor Robert J. Marks talks with KSCJ talk show host Mark Hahn about HAL 9000 and the fundamental opportunities and limitations of AI. One of the main differences between humans and AI is that humans can process new information. The AI ​​can only address the information for which it is programmed.