Science

Minsky: AI is Brain-Dead

Posted 21 Jul 2003 at 01:31 UTC by steve Share This

Wired has posted a short interview with Marvin Minsky in which he expresses his usual complaints about the state of AI and robotics. It seems like only last month he was denouncing our "stupid little robots".


Sour grapes from an old pro, posted 21 Jul 2003 at 06:28 UTC by motters » (Master)

I don't think AI has necessarily lost its way. Robots which can look around the room and tell you what's in it may not be that far away (have a look at Hans Moravec's site), and the recently launched mars rovers are heavily vision oriented, turning stereo data into terrain maps for navigation. I would regard being able to move across a the surface of an alien planet without direct supervisation as a primitive form of "common sense".

That'll be Minsky then I guess..., posted 21 Jul 2003 at 07:51 UTC by Ratface » (Observer)

The misspelling Minksy in the title makes it sound like something from a Clouseu Pink Panther flick!

Oops..., posted 21 Jul 2003 at 13:29 UTC by steve » (Master)

I seem to find a new way of misspelling his name every time we do a story on him. Last time it Minski...

Need of new blood, posted 22 Jul 2003 at 01:38 UTC by alphabot » (Journeyer)

Hey man,

I've had the pleasure of meeting Mr. Minski way back in my Atari Research days. Mr. Minski is pioneer who deserves all the respect leaders in this field of study deserves.

He is my hero :-)

This being said,

Its time for some new ideas in robotics with creative thinking.

The guys from the 60's that laid down the laws of robotics and AI are old fogies now. They didn't solve most of the issues then and another 10 or more years of aimless research over the same concepts of AI and robotics may be a waste of time.

Minski and those of his era should take a rest now.

Let's listen to the new blood that's up and coming.

Of course the new blood will have to stand on the old fogies shoulders.

Donald Dixon

www.creaturebot.com

:-)

The AI Plateau?, posted 23 Jul 2003 at 03:39 UTC by The Swirling Brain » (Master)

It seems like several years back (like 10 or more years ago?) there was a bunch of cool AI things that came out like Neural Networks and Genetic Algorithms and such. We almost thought that robots could really start learning. Now, it's almost funny to hear people talk about NN or GA like we should be all excited about this all new tech. Not that it isn't cool or anything, it's just not new anymore and I believe we've seen the limit reached to those learning abilities. So it seems that AI and new techniques for learning has sort of plateaued. There's nothing seemingly new or innovative that we should take note of that would appear to show any kind of progression in the world of AI. Sure, there are vision systems and speech and perhaps even navigational advancements etc, but nothing really in the way of cognitive advancement. It seems that even Marvin doesn't know what to do to take it up a notch so he's gone horizontal with building a database of knowledge so that a future robot may not need to learn as much when they finally figure out how to make a robot truly learn. For now, robots themselves aren't getting any smarts to be smarter or learning anything or anyway new, even though they are getting better peripherials and do-dads for that future day when they do get on with learning. I think Marvin is trying to coax people to focus on advancing the mental part of robotics since most robot builders seem to focus only on the physical part and ignore or can't get a vision for the intelligence part. However, it seems that instead of taking his words as an encouragement to look further or deeper or looking for new innovations for learning robots, people are taking his words as just a chide and are getting pissed off. People just don't want to admit that the hidden AI side of robotics might just be lacking when they see all the neato visual things the physical side or robotics is currently doing.

We are the new blood, posted 23 Jul 2003 at 17:43 UTC by ROB.T. » (Master)

Hey what happened to all the replies for ``stupid little robots''?

Swirl correct me if I'm wrong, but Minksy (Misky, Minski) is saying AI is dead, but you are saying there is plenty of AI in stuff like vision systems, but you support Minksy...

Alphabot, baby, you are ontop of it! Just look at what the O.G.s (Original Gansta's) of robotics are currently seeking from their version of AI - care for the elderly! Now I would agree that this is a noble and worthwhile cause, but maybe we need to take the current, lofty, ambitions that Ivory League Universities define as what AI `should do' and redefine it along more task oriented/specific AI - we'll could call it ``PGAI'' (Pretty Good Artificial Intelligence). I'm pretty sure this concept exists, I just don't see it a lot.

Hey, this is pretty rebellious talk...

Brain dead, posted 23 Jul 2003 at 17:54 UTC by motters » (Master)

I think that most of the technologies which will eventually produce really intelligent robots are already with us in embyonic forms. Even most AI researchers grossly underestimate the amount of engineering effort required to produce impressive results.

Like Minsky I don't think there is any "magic bullet" solution to the problems of machine intelligence. I think the brain is a heterogenous system composed of different types of sub-systems carrying out different tasks.

huh?, posted 23 Jul 2003 at 20:21 UTC by The Swirling Brain » (Master)

Swirl correct me if I'm wrong, but Minksy (Misky, Minski) is saying AI is dead, but you are saying there is plenty of AI in stuff like vision systems, but you support Minksy...

Nope, what I was saying was that AI has plateaued, stopped, not advancing, at a stand still, not going forward. Having said that, people are still using that old technology.

My bad Swirl, posted 23 Jul 2003 at 22:08 UTC by ROB.T. » (Master)

Must have read the end of your statement wrong, seems different now...

No prob, posted 23 Jul 2003 at 22:46 UTC by The Swirling Brain » (Master)

My wife and kids often can't understand me. I can't imagine I'd make any sense here either. Perhaps when AI advances, my posts will start to make sense. :-)

The value of incrementalism, posted 24 Jul 2003 at 14:15 UTC by mwatts » (Apprentice)

I posted this the first time in the wrong place. Sorry.

The problem with AI is that it has been pursued in the wrong way. AI work has been largely academic rather than pragmatic. As we venture investors in Silicon Valley have learned the hard way, successful innovation comes in relatively small increments rather than in large revolutionary leaps. Many, if not most, failed VC investments fail because the company bites off more than it can chew. Starting with Stupid Little Robots and moving to more complex ones in a series of commercial ventures will move AI along much faster than any other approach. The task will be broken into managable steps and each step will be funded to a far greater extent than has been the case in the past. The commerical success of Roomba, despite the derogatory comments of Minsky and others, is very good news for AI and robotics in general.

See more of the latest robot news!

Recent blogs

22 Aug 2014 mwaibel (Master)
15 Aug 2014 shimniok (Journeyer)
5 Aug 2014 svo (Master)
20 Jul 2014 Flanneltron (Journeyer)
3 Jul 2014 jmhenry (Journeyer)
3 Jul 2014 steve (Master)
2 Jul 2014 Petar.Kormushev (Master)
10 Jun 2014 robotvibes (Master)
10 May 2014 evilrobots (Observer)
2 Mar 2014 wedesoft (Master)
X
Share this page