Jeff Hawkins on Artificial Intelligence - Part 4/5

[Serpico261]

man this is just awesome

Nature proves again that complex structures require simple concepts...

take for example evolution by natural selection which covers everything biolog. we know

A theory that can even described in one sentence!

"The non-random survival of randomly changing genes over time"

It has an incredible simple theoretical basis, but is unbelievable powerful in its results

The mathematicl concepts of nature have to be simple, otherwise you would stuck in the field of improbability

[rogerajacto77]

"people are trying to use in financial markets" - guess it din't work out that well eheh (joking)

[Abhothra]

thank you ever so much for uploading this,i've been trying to find this lecture for quite some time since i "ran into" him on TED.com

I cant wait to see what this will do to the field

[Slurfs]

Where can you get this software?!

[Juefawn]

What is this Hierarchy called?

[UnowMe00]

@Juefawn "Your Highness"?

[rsherfan]

wonder if somebody is attempting to use this sytem to make robots walk... by cross referencing input from the feet with some sort of gyroscopic slash load-bearing data.

[pebre79]

@rsherfan Interesting. I think that idea has "legs!" :P

[guitarslim56]

Did that.

[rsherfan]

@guitarslim56 sweet. much progress? i hope you guys go all the way with it :)

[mkultra1982]

at 1:34 - People in the game industry wanted to use this for motion capture? I wonder if this technology is the basis for Microsoft's new "Natal" system...

[Suqatish]

@mkultra1982

That actually is probably what he was talking about, although I can't find any other links between Natal and Jeff.

[pepefever]

That explains it. Skynet is hierarchical

[johan28]

This is VERY fascinating!

I myself have thought about focusing on AI if and when I pursue my computer science degree. I was thinking it could be something I learn alongside programming itself. Good idea eh?

[kid29a]

great idea but begin now :)

[griptapesequel]

you have to go above your baccalaureate degree brov, so i hope youre in it for the long haul. I didnt learn almost anything REAL about AI until i started my graduate studies. Now that i got my masters im a going to start my Neuroscience degree now. Lets hope that goes just as easily as my mathematics-lades CompSci...

[UnowMe00]

@johan28 yeah, be sure to experiment and have good tools at hand (not so trivial search)!

[koolanator]

@johan28 I would suggest that you specialize in Theory and Computation. Once you get into school you will realize that Computer Science has as much to do with Programming as Astrology has to do with Telescopes.

You will mostly work with learning how to structure data, solve problems by manipulating and transferring algorithms, optimize existing algorithms, encryption and basic hardware design.

[koolanator]

@johan28 If you go into Theory and Computation, you will begin to work with AI and statistical algorithms. I based my graduate paper on Machine learning using HTM and Candidate Critique models. Interesting stuff.

[crw2223]

me thinks malintome is full of it.

[malintome]

Yes I have it. Source for its starting version is as easy as Jeff said it is. Same rules apply - most of all hierarchy! If you really are interested in this subject, I sugest you try to build it yourself. If you stand still at some point with your program, don't hesitate to ask me what next, but... first you got to start. Listen to Jeff again and understand him. There are some things you just need to understand. There is no template! so don't try to build one.

[malintome]

ha ha thats true, i am. And I can say straight away, that you are not interested in the subject of AI.

[pebre79]

Interesting. Do you have a demo?