Chilling story about how to use AI not to generate recipes for benevolent molecules and drugs, but for lethal biochemical weapons.
Some American company investigated the possibilities and trained a model to achieve increasing levels of lethality. It worked just fine. The model came up with horrible existing weapons and also pointed to probably new classes of such weapons. All this in a few hours time. Combine this with states in the grip of malevolent dictators, and there is yet another reason to be depressed. Or to think very hard about counter-measures such as detection and reporting systems.
SOURCE: Financial Times 
TAGS: WAR LIFESCIENCES AI