As someone who works in material simulations, this result is fairly unsurprising. It’s the first step in a high-throughput materials discovery process. For example, if you want to find a material with a desired properties (bandgap, conductivity, etc), you would use a rough simulation model to sift through hundreds of thousands of candidate materials on one metric, giving you 40000 candidates, then sift through those on another metric, giving 1000 candidates, and so on until you have a few candidates that you test empirically, a lot of which will not actually work due to the approximations that were made in the search.
The point being, having 40000 “potentially lethal” molecules is not the same as having 40000 new chemical weapons. If pursued, it might lead to discovering a handful of chemicals more deadly than the current best after an extensive research project. Not to say this isn’t a bad and concerning use of AI!
As for a “sputnik” moment, we currently have highly deadly chemical weapons like VX, with well established manufacturing methods. It’s only actually been used in assassinations and in one potential attack on insurgents. I’m unclear as to how a 2x as deadly or even 10x as deadly VX version would add much military value, when it’s competing against things like bombs and guns.
As someone who works in material simulations, this result is fairly unsurprising. It’s the first step in a high-throughput materials discovery process. For example, if you want to find a material with a desired properties (bandgap, conductivity, etc), you would use a rough simulation model to sift through hundreds of thousands of candidate materials on one metric, giving you 40000 candidates, then sift through those on another metric, giving 1000 candidates, and so on until you have a few candidates that you test empirically, a lot of which will not actually work due to the approximations that were made in the search.
The point being, having 40000 “potentially lethal” molecules is not the same as having 40000 new chemical weapons. If pursued, it might lead to discovering a handful of chemicals more deadly than the current best after an extensive research project. Not to say this isn’t a bad and concerning use of AI!
As for a “sputnik” moment, we currently have highly deadly chemical weapons like VX, with well established manufacturing methods. It’s only actually been used in assassinations and in one potential attack on insurgents. I’m unclear as to how a 2x as deadly or even 10x as deadly VX version would add much military value, when it’s competing against things like bombs and guns.