*promote

(I figure I owe you two * promotes as you tried valiantly to save this controversial video of mine twice. (May it rest in peace.) The second promote will come when you least expect it, but probably when you post a video that includes dick jokes. I like sophisticated humor.)

Promoting this video and sending it back into the queue for one more try; last queued Saturday, January 19th, 2008 8:48pm PST - promote requested by kronosposeidon.

*excellent but nerdy.

Upvote for being a brain fan!

intelligence as measured by the capacity to make predictions.

Fantastic sift. Very, very interesting. Thank you.

Great Sift!

Ok, no where in his rant did he mention how he proposes it all happens. BOOOOOO!

Comment hidden because you are ignoring dag.(show it anyway)

^ I think he did indicate how it will happen. Lost of sensors and massive amounts of recorded data that can be used to build patterns of experience for prediction.

Absolutely fascinating - he's obviously very passionate about this.

I disagree with him on a fairly fundamental level.

I've heard similar approaches many times before. The basic pattern is: "we'll model it based on the low-level structure of the brain, and if we throw enough data into it then the intelligence will emerge". And his focus on "prediction" seems odd. Sure everything is "prediction" in the loosest sense - but then it's equivalent to general reasoning. Sure I can "predict" that 23+45 is 68 based on my previous experience adding numbers or something, but thinking of it in terms of prediction instead of abstract symbol manipulation (or a million other paradigms) doesn't change the nature of the thing or help us understand it.

To me, the clearest hint that he's wrong is he's talking about sensors. To start, AI doesn't need to have anything to do with the real world. Again - I see this pattern a lot in AI research. The basic idea doesn't seem to be solving small problems in a satisfactory way, so let's try it with a big problem and sensors and mountains of data and processor power and hope that something will emerge.

I definitely agree that a theory of brain function is needed. But if we can really grasp how the brain learns I think the demonstration that that theory is correct will be through solving small problems. Any problem that can be solved with silicon can be solved with a current computer (although possibly in much more time). If he has a framework through which a computer can learn, it should be demonstrable in solving some small problem.

And I look forward to responses saying that he's right and I'm wrong because he's smart and rich and has a company doing this.

He has some skillz when it comes to talking. This was far from boring, thanks for sifting.

*brain

*talks

Adding video to channels (Talks) - requested by gwiz665.