4 ms·
Possible AI in a box is easier to understand. Also, in order to build AI in a box, we need some better knowledge of how to build AI, and it's not a stretch to
by backprojection 15y ago
Possible AI in a box is easier to understand.
Also, in order to build AI in a box, we need some better knowledge of how to build AI, and it's not a stretch to see that in the future we may have this knowledge.
What is/would be the theoretical underpinning of a fused global intelligence via something like the internet?
- wlievens 15y ago> it's not a stretch to see that in the future we may have this knowledge There's really no way to support that claim. We haven't made any tangible progress at all towards strong AI in the past 40 years. We did salvage a lot of really cool research projects on the side, but none of them even come to close to suggesting how to build strong AI. It's like saying that building a faster car brings you closer to FTL travel. It doesn't, not in any real sense.
- exit 15y agoi don't know, we seem to make significant inroads into understanding every other physical system we've encountered - why not the brain? and "in the future" is a pretty broad time frame.
- disgruntledphd2 15y agoArguably (and this is a pretty out there view in neuroscience) because in this case, the instrument we're using to understand the physical system is itself the object of study. Throws up some interesting self referential problems.
- exit 15y agowe might make sense of our cognitive routines without really grasping how phenomenological experience arises. i still think that's enough for us to build ai which is equally perplexed by the blueness of a blue sky.