Artifical Intelligence: Closer than you think

[curiosityandthecat]

Saw this article a few months back and decided it was quite possibly the most exciting thing I've read in a very, very long time.

Out of the Blue

Can a thinking, remembering, decision-making, biologically accurate brain be built from a supercomputer?

by Jonah Lehrer • Posted March 3, 2008 05:50 AM

In the basement of a university in Lausanne, Switzerland sit four black boxes, each about the size of a refrigerator, and filled with 2,000 IBM microchips stacked in repeating rows. Together they form the processing core of a machine that can handle 22.8 trillion operations per second. It contains no moving parts and is eerily silent. When the computer is turned on, the only thing you can hear is the continuous sigh of the massive air conditioner. This is Blue Brain.

Read the rest here: http://seedmagazine.com/news/2008/03/out_of_the_blue.php?page=all&p=y

[crocofish]

The research looks somewhat interesting.  Neural network research has been going on for decades.  I believe such basic research is necessary to understanding the brain, but I'm not betting that there will be a major breakthrough any time soon.  My past exposure to AI researchers has made me highly skeptical of most AI research.

The actual machine they are running it on is impressively large, but it's just another multiprocessor cluster.  I used to work on designing chips for supercomputers, and the although computers progressively get denser and faster, I haven't seen anything truly innovative in computer architecture in a long time.

[Brian]

This looks soooo sweet!
I've been doing simple Game "AI" for a long time in C, but I don't think I'll ever do something like this.  

[SteveS]

Great read curiosity - thanks for the post!

[4DeepThought2]

I thought this looked really cool, and did some research. Artificial Intelligence is predicted to come about very soon due to the rapid increase in technology every year; Moore's Law. Eventually, though, the silicon computer chips will probably become so thin that they will not be able to get any smaller (minimizing size of chips allows for more data storage and processing power); therefore halting increased data storage and processing expansion. This is just speculation though, it could very well happen that a computer will become smarter than humans and eventually make itself smarter before silicon can't minimize any more...but quantum computers are also on their way depending on atoms to process information without silicon chips. Anyway, with all the nano technology research I think it is very possible. Whatever speculation there is it is a very cool idea.

[Kyuuketsuki]

Saw this article a few months back and decided it was quite possibly the most exciting thing I've read in a very, very long time.

Out of the Blue

Can a thinking, remembering, decision-making, biologically accurate brain be built from a supercomputer?

by Jonah Lehrer • Posted March 3, 2008 05:50 AM

In the basement of a university in Lausanne, Switzerland sit four black boxes, each about the size of a refrigerator, and filled with 2,000 IBM microchips stacked in repeating rows. Together they form the processing core of a machine that can handle 22.8 trillion operations per second. It contains no moving parts and is eerily silent. When the computer is turned on, the only thing you can hear is the continuous sigh of the massive air conditioner. This is Blue Brain.

And so Skynet was born  

Kyu

[joeactor]

Meh.

I've been working with computers since the 70's.

The prediction that AI is "just around the corner" is decades old.  Must be one heckuva corner!

Humans tend to equate the latest technology with the brain.  It's the great steam engine.  It's electricity.  It's a computer.  It's the internet.

My hope is that this research will further our understanding of the brain, but I think it might be better to study, oh, say... The Brain instead!

The AI we end up creating may be very different from our own, both in function and structure.  SkyNet indeed!

As for the number of processors, I'm not sure that helps.  It may only reach a dead-end in the research at a faster pace.  After all, you can't make a baby in one month by putting nine women on the job...

Musing,
JoeActor

[DennisK]

^Agreed.  What I don't understand is that we don't even remotely understand the brain and it's function, so how can we create AI based on vague information.  Granted, our understanding eclipses what we knew 100 years ago, but still, there is a lot we don't know.

[crocofish]

Meh.

I've been working with computers since the 70's.

The prediction that AI is "just around the corner" is decades old.  Must be one heckuva corner!

Hurray, the voice of reason.

I have also been using computers since the 70's, and I work on designing computer chips and boards, and I do some software.  I understand computers at a very deep level, and as much as I love computers, I am very skeptical of AI claims.

One place I worked in the 1980's poured millions of dollars in AI research.  Most of the people I met who were working on the AI projects threw around a lot of jargon, but they were doing nothing revolutionary.  It was all about looking smart and important, but they were not doing anything more sophisticated than making big decision trees on computers.  It all came crashing down when nothing really useful came out of it.  If you look back at that company's history, they barely make any mention of their involvement in AI, probably since it was such an embarrassment.

Fortunately, at that time I worked in group that wasn't part of that AI boondoggle.  In my group, there were a number of very senior engineers that had seen the promises of AI in the past.  They were skeptical of the AI projects, and they were right.  I think when you've been around long enough, you repeatedly see the promises and the disappointments of AI.

I do find AI interesting, but I haven't seen any AI breakthroughs that have really impressed me.  I would be impressed if I saw an AI that could replicate even the intelligence of a worm, but I haven't seen anything even that sophisticated.  A computer decision tree algorithm with some pseudo-randomization (like used in video games) is not an intelligence.  A big search engine database is not an intelligence.  I do think it would be very interesting if someone does produce a true AI that can learn and interpret information, and if then they let it use things like Google and Wikipedia to access huge pools of information.

One big fallacy that I see repeatedly is that the increasing computer speed based on current computer architecture will produce an AI.  Kurzweil uses that argument, and it annoys me.  Imagine a computer 1000 times more powerful than a computer today, that would be 15 years from now if computer power doubles every 18 months (Moore's Law).  If that future computer can do AI, then a computer today should also be able to run the same AI algorithm except 1000 times slower.  An artificially intelligent decision made in one second on that future computer would take 1000 seconds (almost 17 minutes) today, slow but observable.  But I have not seen even a very slow AI yet.  I think it will take a whole new computer architecture and a deeper understanding of the mind to have a chance at producing an AI.

[4DeepThought2]

Meh.

I've been working with computers since the 70's.

The prediction that AI is "just around the corner" is decades old.  Must be one heckuva corner!

Hurray, the voice of reason.

I have also been using computers since the 70's, and I work on designing computer chips and boards, and I do some software.  I understand computers at a very deep level, and as much as I love computers, I am very skeptical of AI claims.

One place I worked in the 1980's poured millions of dollars in AI research.  Most of the people I met who were working on the AI projects threw around a lot of jargon, but they were doing nothing revolutionary.  It was all about looking smart and important, but they were not doing anything more sophisticated than making big decision trees on computers.  It all came crashing down when nothing really useful came out of it.  If you look back at that company's history, they barely make any mention of their involvement in AI, probably since it was such an embarrassment.

Fortunately, at that time I worked in group that wasn't part of that AI boondoggle.  In my group, there were a number of very senior engineers that had seen the promises of AI in the past.  They were skeptical of the AI projects, and they were right.  I think when you've been around long enough, you repeatedly see the promises and the disappointments of AI.

I do find AI interesting, but I haven't seen any AI breakthroughs that have really impressed me.  I would be impressed if I saw an AI that could replicate even the intelligence of a worm, but I haven't seen anything even that sophisticated.  A computer decision tree algorithm with some pseudo-randomization (like used in video games) is not an intelligence.  A big search engine database is not an intelligence.  I do think it would be very interesting if someone does produce a true AI that can learn and interpret information, and if then they let it use things like Google and Wikipedia to access huge pools of information.

One big fallacy that I see repeatedly is that the increasing computer speed based on current computer architecture will produce an AI.  Kurzweil uses that argument, and it annoys me.  Imagine a computer 1000 times more powerful than a computer today, that would be 15 years from now if computer power doubles every 18 months (Moore's Law).  If that future computer can do AI, then a computer today should also be able to run the same AI algorithm except 1000 times slower.  An artificially intelligent decision made in one second on that future computer would take 1000 seconds (almost 17 minutes) today, slow but observable.  But I have not seen even a very slow AI yet.  I think it will take a whole new computer architecture and a deeper understanding of the mind to have a chance at producing an AI.

Great analogy about the processing speed and such. The ways in which the attempted AI's I have encountered work, a few accessible online that are prototypes and in "test stages", is that they only learn information based on what another users has told them, so initially it just memorizes text and stores it. Which is far from anything independent or intelligent. Maybe some new computer design/architecture will arrive and allow us to come closer to AI or a computer intelligence. I read a little bit about quantum computers (a new way of computing), don't know much though.

[curiosityandthecat]

Great analogy about the processing speed and such. The ways in which the attempted AI's I have encountered work, a few accessible online that are prototypes and in "test stages", is that they only learn information based on what another users has told them, so initially it just memorizes text and stores it. Which is far from anything independent or intelligent. Maybe some new computer design/architecture will arrive and allow us to come closer to AI or a computer intelligence. I read a little bit about quantum computers (a new way of computing), don't know much though.

The day we create a true, working quantum computer... everything changes.

Everything.

[Tom62]

The day we create a true, working quantum computer... everything changes.

Everything.

It is not the hardware that counts, but the software that runs on that computer.  As long as we have companies like Micro$oft, who'll have to write the operating system for that quantum computer, nothing will change .

[joeactor]

The day we create a true, working quantum computer... everything changes.

Everything.

It is not the hardware that counts, but the software that runs on that computer.  As long as we have companies like Micro$oft, who'll have to write the operating system for that quantum computer, nothing will change .

Possibly just as important, is that there isn't such a distinction in the brain.  The hardware *is* the software, and visa-versa.  (or is it the other way around in reverse?)

Wetware Inside (tm),
JoeActor

[curiosityandthecat]

The day we create a true, working quantum computer... everything changes.

Everything.

It is not the hardware that counts, but the software that runs on that computer.  As long as we have companies like Micro$oft, who'll have to write the operating system for that quantum computer, nothing will change .

Haha. Oh, lord, I can see it now. "Try our new operating system: Microsoft Windows SP!N! For all your quantum needs!"

Tagline: "SP!N does it up, down, and everything in between."

[SSY]

there are people who already claim to have built a quantum computer.

While I agree, when we get them, it will be fun, there wont be some massive, instant change, the first true quantum computer probably will be able to operate on 2 or 4 bit numbers, we are a way off anything usable.