My 9-year-old has been pretty concerned about consciousness and the evolution of intelligence and AIs and such recently (for the non-childed: this isn’t humblebragging; my kid is indeed a weirdo besought with weird worries–because of Nature and Nurture–but this brand of light-weight epistemological crisis is pretty much developmentally on the ball among humans).
Anyway, without malice aforethought (or any real forethought at all), I let him watch half of this X-Files episode where a building-maintenance AI murders a dude, and that sorta made things much worse, in terms of his growing terror about what is and isn’t knowable–esp. know that I’d compounded the problem by introducing the possibility that an autonomous building might try and kill us if it felt even the least bit threatened.
So I showed him this video, which gives a much more complete sense of how academically interesting–but largely trivial–modern autonomous AIs are. It’s a very concise, but nonetheless enlightening 5-minute primer on neural network AIs and genetic programming. We watched it a couple times, he groked it, I asked if it seemed threatening, and he agreed that MarI/O could not hurt anyone–with the proviso that this was because they’d hooked it up to a Nintendo. If they’d hooked it up to something else . . .
And, well, I had to concede his point. That is sorta the story of evolution, isn’t it? Red in tooth and claw, etc., etc., etc.
Anyway, it’s still a nice little lesson for those of us with no working understanding of the field (and I’ve got it on good word that the video is basically on the ball).
If you’re interested in digging deeper into AI (artificial intelligence) and GP (genetic programming), my trusted source advises checking out Julian Togelius. I advise this, too.