[harshreality]

I attended a talk he gave on campus when I was in college, and afterward I lurked as some really smart people (grad students, maybe one or two profs) asked him for details. He came across as a con artist, selling ideas that he can't defend. For instance, one that stuck out was that he had nothing to back up his assertion that within x years there would be the ability to non-destructively copy all the information in a brain. Several questions I thought were fairly reasonable were met with the response: "read my book[s]." (in context, as I recall he was promoting _The Age of Spiritual Machines_.)

His basic argument of the inevitable advance of technology to true AI, or ability to upload consciousness or integrate computers with brains, rests on priors that are not clearly relevant.

With his latest title, "How to Create a Mind: The Secret of Human Thought Revealed", how can anyone take him seriously? Scientists (practicing scientists) have been writing books like that for decades, and the lesson should be that unless you've demonstrated that you can build an AI (real AI, not soft AI like machine learning), you have no business using a title like that. If real AI is simply a combination of the right machine learning tricks, then the trick is in the combination, because nobody has figured it out (publicly) yet.

[testbro]

Kurzweil is awkwardly reminiscent of the connectionists of yore. I'm surprised nobody's taken a similarly hostile attitude towards his (as yet unsubstantiable) predictions.