3 ms·
It's assumed that strong AI will be able to improve itself, in an exponential fashion according to Moore's Law. But an AI still has to overcome the same very re
by wtallis 12y ago
It's assumed that strong AI will be able to improve itself, in an exponential fashion according to Moore's Law. But an AI still has to overcome the same very real challenges that we humans face in trying to build ever-faster computers. A strong AI can only work magic on the software problems; it can't make ultrapure silicon crystals grow any faster and it can't make electron leakage just disappear. It still faces the same risk of silicon semiconductors being fundamentally unable to scale to the performance level required, forcing humans or AI to start over with an alternative computing substrate. It can't solve any of these hardware challenges overnight because they require physical experimentation.
So I think it's reasonable to think that the development of a strong AI won't change things overnight and we could see decades of lag between first having a computer capable of strong AI at any speed and having a strong AI that outpaces humans.