There are thousands of people working furiously on all of those problems, though. I was working at Janelia Farm until last year. You should see what those guys are planning for neuroscience. It's going to be huge. It's a mistake to assume that our capacity to query neural circuitry is going to remain stuck at such a low rate.
And they have been working furiously for a good while now. The romantic in me hopes that they can make it happen, but I'm not holding my breath. Mainly because a lot of people confuse speed of processing with intelligence, but that's not really the spark of life we need. There's nothing out there so far that can mimic human thought even at hundreds of times slower than human responses. We still don't have something quite right with the basic principles yet.
What Kurzweil is saying is that you can't just linearly extrapolate the progress of these projects. Even if we only know a fraction of a percentage of what's going on in the brain after 20 years doesn't mean we'll know only twice as much in another 20 years. Rather we're likely to see the pace accelerate. Going from 50% to 100% might take less time than going from 1% to 2% knowledge.
Keep in mind that he's been making predictions like these for decades and he's got a pretty good track record.
I guess that I am saying is that I'm not convinced that the direction we have been heading is the direction that will ultimately produce the end result we want. The model T and the present day automobile share the same basic concepts. I'm wondering if human-level AI will bear any real resemblance to today's solutions on the concept level. I'm not convinced it will.
I think the argument is that we can already simulate a single neuron fairly accurately today, and so some decades from now we'll likely be able to simulate them all, plus their interconnects, and thus by pure brute force we'll get a human-level AI.
