I watched a wonderful lecture by Yann LeCun, in Oct, 2016, and all of a sudden I figured it out, and it totally shocked me to be honest. Yann LeCun is a truly fabulous engineer, and he really comes at everything through that basic lens. For that reason he's completely comfortable with facebook's assignments, and knows exactly what he has to do with ANI, and does it very well. I am completely different; I am a thru-and-thru mathematician, and in an alternate life I would have been a quite famous mathematician (pretend you can see this for the present), and I can SEE what you cannot.

You don't even believe that is conceivable at this point -- you have no idea how it will happen, and you incorrectly see the way a physicist sees, where you think that 95% of the physical world we have no idea how to do, when a physicist REALLY sees that 95% of the actual world is clear and definite, and only 5% is opaque. The same is true with AI -- a mathematician sees that once unsupervised learning is understood, we are very close to AGI, and he can SEE vaguely and incompletely what's there, although it is impossible for him to see yet how exactly to make it happen...

It's the difference between an engineer, who has ZERO idea how to do it and thinks it's way off on that basis, and a mathematician who knows that it's only one real breakthrough before it happens. I think the second way, and in my mind :-) I know I'm right. You can disagree -- but I understand how you think now, and I can see further than you.