I hear (and see online) many people claim that "AI can't be done". That machines match and outperform people in strength, speed, power and calculations, but they can never match the "nature occurring machines", like human beings and animals in "brains".

I hereby present a thought experiment (I'm sure dozens of people presented it before though...):

Take a computer, a big one (say, the ones we'll have in 20 years, with nano-tube transistors, holographic memory and stuff like that), and shove into it a neural net. A large neural net. A huge one. Billions of neurons, with 10Ks of synapses coming out of each one. Load it with a snap-shot of someone's brain, and let it run. I see only tehcnical difficulties here. For instance - size of the net, getting that snapshot (advanced MRI ?), etc. I see no fundamental theoretical obstacles.

There will be AI, I'm as certain as possible about it. I don't think it will be in my lifetime though. But it will be.


Comments

comments powered by Disqus