re: How you communicate matters VIEW POST


Re: Dreyfus, I think the connectionists are getting pretty cocky these days. Their methods are ascendant but true intelligence must at some point account for abstract and symbolic reasoning. The current crop of AI algorithms are nowhere close to that and we need an equivalent of Dreyfus for all the contemporary hype.


I can get on board with what you are saying up to a point. Tech progress is frequently made by critiques of existing tech and ideas. For example, Raft is largely born out of problems identified in Paxos. Where I hesitate is when the critiques get personal. Dreyfus papers insinuated that the AI researchers themselves were charlatans (he compared them to alchemists IIRC). Whereas history proves both that Dreyfus's theories of learning were better, it also demonstrates that AI researchers could accept when they were wrong and move on. This is not the mark of fortune tellers, but of exercising the scientific method.

So I do like to see people pointing out the downsides of the hype. (In fact, I look for these to try to get a fair picture for every tech I research.) But we don't need some of the tactics Dreyfus displayed. On the other side of the coin, we also don't need people to be so attached to their tech that they (inaccurately) interpret logical critiques as personal attacks.

code of conduct - report abuse