What I find amazing are problems like this:<p>What is the probability of a chopper crashing or someone shooting a rocket at it? I don't know, but I can say that when I see a chopper on TV, the probability is around 80%. Still, when I see a real life chopper, I know that it's probably not going to crash or getting shot at.<p>Do I know that because of any experience I have? No, I haven't seen many choppers in real life and not once have I seen one take off or land. But I know about the workings of the media. I know about why people use choppers and that they wouldn't use them if they always ended up crashing and burning. <p>Even if a program is very artificially intelligent, it's still very hard to get training data that matches that mixture of sensual experience and logical reasoning. <p>It's a hard problem. I still believe it can be done, but we're not nearly there yet. I hope Hawkins makes some progress without a big disppointment that discourages further research.