Don’t put all your eggs in one basket.

The US is developing a highly advanced AI to work out what next steps an enemy might take in any given circumstances. It’s a good idea and a very logical expansion of the DoD’s intelligence armoury. It relies on AI’s current favourite resource, data. A military strategist can infer enemy manoeuvres from tiny, seemingly insignificant details. But that same military strategist simply doesn’t have the mental bandwidth to examine all the data out there, big and small. An AI does.

There is one small problem though. Statistical inferencing or reasoning still largely drives today’s AI. Expert systems can be deployed as well but they are clunky, time-consuming and inflexible. What you need to complete the mix is an AI that, without dedicated human programming, can perform abduction (hypothesising from limited data), counterfactuals and common sense reasoning. The DoD might have figured out how an AI can do those things but that is unlikely. For the time being, they will need to keep a few human military strategists in the loop.

Link to article:

You may also like to browse other AI articles: