<p>Heya, I'm trying to discover whether to delve into computer science or compe (posted here a year ago, but did other things instead of acting). I like the idea of compe, but I don't know how much of an impact it can have on AI learning (my primary interest).</p>
<p>I imagine that computer hardware in it's current form is setup inefficiently to mimic the brains that we base AI learning programs off of, but that's my uneducated guess. I know that neurons both store info and do calculations which is different than a CPU+ram.</p>
<p>Putting my question a little more succinctly; will computer hardware need a major revamp/makeover to suitably host AI learning algorithms of the future, or do most people think that computer architecture is just fine in it's current incarnation?</p>
<p>The physical level of the computer is fairly irrelevant for AI. Unless you get to very special stuff like quantum computers or multi-valued computer logics or something like that (but which may also not be necessarily about “AI”, but just hard physical technology). Basically or so I think is widely believed, current general purpose computers can describe “any kind” of computational objects (as long as they are describable as Turing complete and computable in the von Neumann architecture), so it would be very advanced to think that there’s something to be done at the hardware or computer logic/maths level. And you probably might want to read carefully about the research of these gentlemen, whose inventions are still in use.</p>
<p>And even in that case, I’d somehow look at pure physics or chemistry, rather than engineering. To get to know better whatever the principles (e.g. for organic electronics or quantum-level computation) are that might be really usable for supposed advanced technologies.</p>
<p>Most core AI research and implementation of today falls entirely under CS (while there are other fields like Cognitive Science that supply some research about the human cognition), is done in software and the used supportive sciences are mathematics and statistics/probability.</p>
<p>Looking into “exotic” computer architectures for AI sounds like a reasonable idea and I’d be surprised if there isn’t already a fair amount of work in that area. Computer engineering would be an excellent choice if you wanted to come at the AI problem from a hardware/architecture angle.</p>
<p>For the most part, however, AI is more aligned with CS, and CS would in no way preclude you from pursuing hardware/architecture angles. Indeed, at the undergraduate level, I don’t think there’s enough difference to agonize over the choice; just find a professor to do research with and choose your electives wisely.</p>