Congratulations to Dr. Shane Legg who just sat his PhD oral exam in Switzerland. Hes my oldest friend (in that I have known him forever, not that hes been hit (much) with the wrinkly stick). He’s into general theories of intelligence, which, although it sounds like academic wishy washy, ivory towered mumblings of a deranged, albeit fairly polite, madman, actually might be a bit more relevant than you think.
The problem with AI for the general public is one of perception. Think about the history of AI, summed up in a paragraph: Computers, then AI hype, then AI failure, then some more AI hype, and then failure, and at the end of it… really good chess computers. The end.
And so the general public have given up on AI, and by general public, I mean anyone who isn’t following AI progression, which… is probably you! 😉 But in the meantime, Moores law marches on, and we have lots, and I mean LOTS, of computing power, even just sitting on our desks.
And so AI has marched on too, quietly, under the covers, starting to do more and more things. Now, of course, we are not talking true, general intelligence like the stuff Shane is researching, but fairly specific functions. But with Shanes research about a theoretical model of true, general intelligence, we can start to see a path from today to a general AI in the not too distant future.
Sounds crazy right? Well, it does. People have always been top of the heap in terms of intelligence, and the thought that there could be something else about to come along and usurp our position as rightful rulers of the known universe, well, kinda… um… sucks, and we instinctivally, and somewhat arrogantly, label it as ‘crazy’. Never going to happen. We’re the greatest. But, there is no ignoring the signs.
The computational problem has been, more or less, solved. 10 years ago, there just wasnt that much computational power available. People could think about things, but never truly implement them, other than in trivial ways. So, true AI was not only theoretically impossible, it was also practically impossible.
Now, the practical problem of computation is more or less solved, and certainly will be completely solved in 10 years. The theoretical problem… is marching on. Shanes work, amongst others, on general intelligence is making that clear. So, without a practical barrier to general artificial intelligence, major, massive, world-changing changes are afoot, without really anyone being particularly aware of them.