Slashdot Mirror
Marvin Minsky: It's 2001. Where is HAL?
ZigZag writes: "Marvin Minsky speaks about everything important (MUDs, education, AI, N(atural) I, immortality) while fighting with his MS Word for Mac presentation slides at the Game Developers Conference. Transcript, audio and video are available from Dr. Dobbs. It was in part a preview of his upcoming book The Emotion Machine. Some quotes from the talk will give you a feel: "Whenever you see a number, you should say `how sad'"; "Have you heard the theory that to learn something you should do it in little bits and not stay up all night working on it? If that were true, there would be no computer games"; "robotics people treasure their videos - because it won't work tomorrow.""
Using Word for Mac as a presentation program? How can one be so bright and at the same time be so dumb?
666
; -- the corruption of government starts with its secrets. a truly free people keep no secrets. --
Indeed. People can make fun of Minsky and the other proponents of classical strong AI but they should remember that connectionists also made such claims. I have a book from the mid 1980's "Apprentices of Wonder:Inside the Neural Net Revolution" that made also sorts of silly predictions like intelligent cars that drive themselves (not as research projects, but commerically available) by 2001.
The part that literally floored me is "where you're hoping you won't have to figure anything out,"
I'm no fan of old-school AI, but Minsky has a point -- people use genetic algorithms and neural nets to "learn" from examples, but such pattern matching tells us *nothing* about how learning really happens. They are just generic black boxes that people throw at data in the hope that something useful comes out.
Crap. The "emergent AI" stuff that's been demonstrated to date has had limitations just as profound and seemingly fundamental (but different) as traditional symbolic AI. Amongst others, it has scaling problems of its own when you try and build more complex emergent system.
Not that it's not useful, and interesting research, will undoubtedly produce some interesting production systems, and might give us some pointers along the road to HAL, but don't claim that the bright new future of is just round the corner as soon as we take off the shackles that the neats are placing on the scruffies.
Go you big red fire engine!
Any sufficiently advanced technology is indistinguishable from a rigged demo
--Andy Finkel (J. Klass?)
You leave out some important information about Minsky's paper with Seymour. They found that a simple type of neural network called a Perceptron could not determine if a type of image that looks like a spiral is connected or not (is the image made of one spiraled line, or two spiraled lines?)
At that time, neural networks were brand new, and the later advances in the 1980's weren't even conceived. It turns out that more complex networks CAN determine if the image is connected or not. The paper was not about those more advanced networks, just Perceptrons. There was nothing wrong with his findings. If anything, it was an overreaction on the part of the AI community to his paper that shut down neural network research. If you blame Minsky, then you've got to blame everyone else who basically read the paper and gave up for 15 years.
If tits were wings it'd be flying around.
in which I've reviewed the talk Minsky gave. The same review has been posted on comp.ai and comp.ai.philosophy, too. You can find it in the Media section in k5, titled ' Minsky's "Programs, Emotions and Common Sense" '
Thanks!
--exa--
.. and it wasn't even that interesting the first time around.
/. editors can recommend a better tech news site, since they're obviously not reading their own!
Maybe the
Uh, that'd be Rodney Brooks. Nowadays he's working on a humanoid robot called Cog - which still uses his bottom up subsumption architecture, and IMO seems to be a bit of "anthromorphic robotic grant troll"! I think he may have some plans for adding representation and cognition, or maybe I'm just thinking that he *should*!
Yeah, but the symbolic part of it itself isn't a hard problem - Allen Newell's SOAR already does pretty much everything you could hope. Who (other than a neurologist) cares if the implementation is itself symbolic rather than based on connectionist building blocks.
The hard part of creating a real artifial intelligence is the perception/representation/cognition bootstrapping part of it, and requires an embedded approach that Minsky ignores.
I disagree with you about Strong AI requiring a low level neuron simulation - IMO consciousness is a result of high level architecture, not Penrosian low level specifics! I believe it's just an "inward looking sense" - a feedback path.
Minksy's put down of perceptrons was incredibly short-sighted, and not simply reflective of the state of knowledge at that time.
Minsky was just arguing that a perceptron could not compute an XOR function, and the reason he was wrong is simply because he didn't consider that you might connnect one perceptron to the output of another.
For Minsky to not even consider connected perceptrons was a humungous brain fart for which he should rightly be ridiculed, particularly given the influential position the way he was in at the time, and the effect it had stifling all ANN funding and research for a long time.
P.S. Sure the backpropagation learning algorithm had yet to be invented (although nowadays it seems trivially obvious as a dynamic programming heuristic approach), but that is an entirely separate from not even considering connecting two together!!!
-------
CAIMLAS
~/ssh slashdot.org ssh: connect to host slashdot.org port 22: too many beers
There is another point of view (which I believe is more adequate) which boils down to mind and intellect as fairly sophisticated adaptation function (or tool), so to implement AI we have to start with very simple machine capable of interacting with its environment, learning, adapting, and evolving.
This point of view is missed by a lot of AI researchers, I think, because they're thinking in terms of numbers and theorms, and not the actual human experience. Seperating mind from the body is IMHO a big mistake, and I absoulutely agree with you it's wrong.
Those of you who aren't involved with Neural Networks might find it interesting that almost all research into computerized neural nets stopped when it was proven that a basic perceptron (what you typically visualize when you think of a "neuron") couldn't distingush an XOR function, e.g. nonlinearly seperable data. Of course, this was a pretty simplistic way of looking at it, and growth in the field is exploding now.
I had a big relevation in terms of working with neural nets when I stopped thinking about the math a little bit, and asked myself: If I was this little robot/program/whatever, what would I see? How would I find a pattern in the data I was presented through my senses? (e.g. a Analog-Digital converter connected to a light meter).
We might not be on the ball for 2001, but give it a year or two. :)
..don't panic
here. Thanks for playing.
How we know is more important than what we know.
I thought that article looked familiar. And yep, posted on Slashdot not too long ago.
2 6
http://slashdot.org/article.pl?sid=01/05/27/17242
- - - - -
Napster-to-go says "Fill and refill your compatible MP3 player", which is a lie. It's not MP3. It's WMA with DRM.
As much a fan as I am of nature and the wonders it's created, I can't help but think it could have done a better job on neural tissue. Specifically, the speed at which impulses are conducted could probably be muchly improved where our systems to be redesigned from scratch.
So yes, I think computers are faster, and in the end will be able to outperform our wet systems, if we can only figure out how to do it.
I'm no fan of old-school AI, but Minsky has a point -- people use genetic algorithms and neural nets to "learn" from examples, but such pattern matching tells us *nothing* about how learning really happens. They are just generic black boxes that people throw at data in the hope that something useful comes out.
I agree that the old ANN pattern recognition approach is a dead end. However, a lot has happened in neural networks in the last decade or so. We are learning a lot from neurobiology. We are learning that signal timing in biological networks are crucial to learning and motor skills. One of the important discoveries seems to be in the area of temporal correlations among spiking signals, i.e., determining whether signals are sequential or concurrent. If they are sequential, it appears (c.f. the work of Dr. Henry Markram et al) that the order of arrival is crucial. The time scale is on the order of miliseconds. The new spiking neural networks are so unlike the old ANNs that a new discipline has emerged, one which tries to distance itself from the old ANNers. It's called computational neuroscience (for those who don't keep up with progress in this area).
When you have emergent phenomena like consciousness
What evidence do you have that consciousness is emergent? I suggest we stick to the stuff (intelligence) we can observe (and somewhat quantify) and worry about consciousness later.
And your implication that he is somehow holding back the field of AI is not too plausible, either. It's hardly as if he is controlling AI research all over the world. He's not even controlling AI research at MIT. If neural networks haven't yet taken over the world, you can hardly lay the blame at Marvin's doorstep.
It's a good thing that Dr. Minsky is not controlling AI research at MIT and elsewhere, although he tries. We do hear rumors of his close encounters with other AI researchers such as avant-guarde roboticist Rodney Brooks.
The symbolic AI camp has been at it since the fifties and they made a lot of noise over the years. They have failed miserably. Rather than lick their wounds and moving on to more fruitful endeavors, they continue to uphold their failed approach through various funded projects and obsolete AI curricula that are being taught at major universities and AI centers around the world.
The symbolic aproach is dead and should be buried once and for all, in my opinion. Ultimately it will be but a footnote in the history of AI. The future of AI belongs to connectionism, the only model that has a chance of taming the otherwise intractable complexity of animal intelligence. We need fundamental perceptual and motor learning principles. We need fundamental principles of motivation. Once we formulate these all important principles, we'll know how to apply them to billions of self-modifying cells working in parallel. Only then will human level intelligence become a reality. It will happen in our lifetime.
Certainly the symbolic logic guys were wrong as wrong as those who thought neural nets would solve everything.
To claim that it is wrong to think that neural nets would solve everything is to ignore the evidence in my view. The truth is that the brain is a collection of neural nets feeding into and relying on one another. Each net has a specific role to perform. Sorry, but this sounds very much like neural nets solving everything to me. Maybe you had something else in mind.
It is clear that AI hasn't delivered on the promises made over thirty years ago. What happened? In a preview of his upcoming book, The Emotion Machine, Marvin Minsky examines the failures of AI research and lays out directions for future development in the field.
I used to be a Minsky fan (I still have a copy of his "Society of Mind") but not anymore. Marvin Minsky is one of the reasons that AI still has not delivered on its promises. He is part of the old symbolic school of AI. He was the guy who, with Seymour Papert, wrote a scathing criticism of the then embryonic field of neural networks, effectively strangling research in neural networks for the better part of a decade. I am sure Dr. Minsky has had occasions to change his views since but I don't think he has anything to offer that will lead us to HAL. The following is a quote from a Scientific American article on Arthur C. Clarke's HAL.
The novel of 2001 explains how the HAL 9000 series developed out of work by Marvin Minsky of the Massachusetts Institute of Technology and another researcher in the 1980s that showed how "neural networks could be generated automatically--self-replicated--in accordance with an arbitrary learning program. Artificial brains could be grown by a process strikingly analogous to the development of the human brain." Ironically, Minsky, one of the pioneers of neural networks who was also an adviser to the filmmakers (and who almost got killed by a falling wrench on the set), says today that this approach should be relegated to a minor role in modeling intelligence, while criticizing the amount of research devoted to it. "There's only been a tiny bit of work on commonsense reasoning, and I could almost characterize the rest as various sorts of get-rich-quick schemes, like genetic algorithms [and neural networks] where you're hoping you won't have to
figure anything out," Minsky says.
The part that literally floored me is "where you're hoping you won't have to figure anything out,". All along I'm thinking that intelligence is so complex and intractable that the most plausible solution to the problem of making a human-level AI is one where we let the AI emerge, grow and learn. IOW, what we really need to understand is the learning process, which encompasses perceptual, motivational and motor learning.
But here comes Marvin Minsky, a luminary in the AI community, insisting that figuring everything out is precisely what needs to be done. Haysoos Martinez! This is the main reason why we still don't have human-level AI! I think Minsky's stance is a disservice to computational neuroscience and ANN researchers everywhere.
The man has had his day in the sun. Now it's time for the younger generation of AI researchers to come in and say "hold it! we're taking a different approach from now on. The unkept promises of AI were made by the old symbolic AI crowd. There is a new school in town. The new AI neural, it's emergent, and it's gonna to kick ass!"
Why is there a picture of Al "Grandpa" Lewis doing in that article? All Minsky needs now is a cape.
No no no, according th the Author Clarke that was not the case. As is demonstrated here:
http://www.underview.com/2001/faqs/faqs.html#faqg
"HAL". Something like Highly Advanced Lifeform, right?
Well, almost. The answer is given in black and white in Arthur C. Clarke's book of "2001: A Space Odyssey", Chapter 16, which is titled (ahem) "HAL" (note that in this case the book gives a specific answer to a specific question, whereas in situations that are more open to individual interpretation I do not necessarily take solutions out of the book).
Clarke writes:
"Hal (for Heuristically programmed ALgorithmic computer, no less) was a masterwork of the third computer breakthrough."
Note that, strictly speaking, HAL is not an acronym in the sense of being formed from the initial letters of separate words. In the film we only see it as "HAL 9000", but it is noticeable that in the book Arthur C. Clarke himself consistently refers to "Hal", not "HAL". So, therefore, do I.
Sure we do. There are server farms with more than 10 teraflops. If strong AI could be achieved with a big cluster, there'd be big clusters doing it.
If compute power were the problem, we'd have powerful AI systems that were really slow. That's not the case. We really don't have a clue how to do strong AI. The stuff that used to sort of work but was slow, like language translation and question answering, now works fast but isn't much smarter.
Speed does make what AI capabilities we have more useful, because we can use them on dumber problems. Machine translation still sucks, but now that it's so cheap it's given away, it's useful just to find out if something is worth looking at at all. You can now afford to translate incoming mail to find out if it's spam. The same goes for question-answering at the "Ask Jeeves" level. It's not very good, but it's cheap.
To see where classical AI went to die, go to the second floor of the Gates Building at Stanford. There, below gold letters reading "Knowledge Systems Lab", are empty cubicles, obsolete computers, and tables with ancient copies of Wired. It's depressing.
Game AI, on the other hand, is steadily getting better. It's generally non-verbal and grounded in a semi-physical world. That's probably why progress is possible. The gamers are definitely gaining on the academics, and are probably ahead at this point.
Eventually, game characters will have enough of a life that they'll have something to say, and then we may start to see a path to developing strong AI. But it's a ways off.
Others share that opinion, including some of his grad students. The amount of hype (Newsweek cover, TV specials, and a movie) about that project is excessive for the results obtained.
What they seem to be developing is technology for faking emotional behavior. This came close to a commercial product, a microprocessor-controlled doll, sort of like a Furby with facial expressions. This was supposed to be a joint venture with Hasbro, but apparently didn't ship. IS Robotics, Brooks' startup, seems to be a defunct server.
Behavior-based robotics is interesting, but without some environmental modelling and short-term planning, you'll never get above the insect level. Feedback can only take you so far. Feedforward, though...
It's a very Japanese idea. This is reflected in Japanese interest in robots, both in manufacturing and in anime.
Maybe that's the problem: we've been making computers too small. Minsky should look into creating some skyscraper-sized machines.
Minsky's concept of mind and intellect is too simplistic and often plainly wrong. of course slashdot is not the proper place for in-depth discussion but from what I've read I have an impression that according to Minsky mind is something completely authonomous, with its own absolute (logical rules), something you can implement in a machine (hence the whole notion of AI). There is another point of view (which I believe is more adequate) which boils down to mind and intellect as fairly sophisticated adaptation function (or tool), so to implement AI we have to start with very simple machine capable of interacting with its environment, learning, adapting, and evolving. I believe someone (also with MIT roots) is doing just that, and quite successfully. Minsky separates the mind from the body, from the environment, from the whole human exprience, and this is plainly wrong. No wonder AI (according to him) is a dead end. Incidently I remember reading one of his books, he's tried to explain the notion of 'humor in music', in particular in some Bethoveen composition, using mathematical analysis. Oh vey.
Sure it's possible to make something without understanding it 100%, but that's not a very useful approach if the goal is understanding. That's the point. There are a lot of people working in AI whose goal is not AI per se but rather an understanding of the way that the human mind works. To them it's not particularly useful to make a human-level AI if it's nothing but a black box. To them, a lower level of AI is more valuable if it comes with greater understanding.
Quite frankly, I can see exactly where they're coming from. We're already surrounded by more human-level intelligences than we can find productive employment for. Why go out and create more artificial ones that are likely to suffer from exactly the same sorts of problems, only at much greater expense, unless we actually get something useful out of it in the form of increased knowledge?
There's no point in questioning authority if you aren't going to listen to the answers.
...for game programming, but I'm always happy to have people read Minsky because he tends to crack peoples' preconceptions about what is "obvious" about consciousness and AI and etc. Even better might be Daniel Dennet, author of "Consciousness Explained". Less philosophically sound (and ultimately less satisfying) but still very interesting is Steven Pinker.
I think it's very important to understand that there's no magic to consciousness. It's not something shrouded in mystery about which we know nothing. In fact, we know an amazing amount about individual areas. The topology of the information processing membranes are more complex than we can sort out just yet, but there's nothing about the structure of the brain that's not duplicable by silicon hardware. We just have a lot more mapping to do.
Also important to notice is that to implement the human mind in hardware (as opposed to wetware), we'd need something on the order of a 10 teraflop supercomputer. We just don't have the hardware to pull that off yet. The AI-related optimism of yesteryear was fueled by the misconception that computers are faster than humans. What's really true is that the "programming" that underlies the various gadgets in the mind is the product of millions of years of specialization at small tasks. We have fantastic motor-control gadgets and unparalleled pattern-recognition wetware, for example. Figuring out exactly how many animals are in 15,342 groups of 967 animals each was never all that important, so we never evolved any gadgets to carry out high-speed arithmetic. On the other hand, we're good at seeing how things divide out and how games might be played to our advantage. Idiot savants have been known to find extremely large prime numbers as if by magic - probably the same hardware put to an exotic use.
So in 20 years (or so), we'll have the hardware, and maybe we'll have the information processing topology as well. Some intrepid researcher will put all that in a state-of-the-art cybernetic body. Then it'll be a matter of watching the first hardware human child grow up and meet the world.
PS- make some pretty bold claims here, and also cite a number or two that one might be expected to view with suspicion. I can back it up, just ask.
Nato
I am the one true god. However, as an atheist, I don't believe in myself. I guess I have a self-esteem problem.
Neural nets, on their own, are not very smart in many ways. Sure, there are some problems it seems that only they can solve. But complex, multi-stage problems generally baffle them nearly indefinitely.
The brain is not a big neural amalgam that gets to some critical mass and then suddenly starts doing stuff. It's wired. It's got gadgets. It's really a big collection of them. Some of them are damned complex, composed of sheets of neurons talking to each other in intricate, bewildering arrays.
And modern Connectionists understand that. Certainly the symbolic logic guys were wrong as wrong as those who thought neural nets would solve everything. But that's people like Fodor and Chomsky. The Minksy "Agent" model is very much on the "Connectionist" side of the map. That's not to say that I agree with everything he says, but I think you're unfairly blaming him for the mistakes of others.
Symbolic logic, by itself is no panacea, but neither is the neural net. I'm willing to bet that a lot of the interactions of various neural nets in the brain form very formal symbolic logic gadgets. Also, in the end, it is the formal logic of virtual-neuron microcode in a computer is what will generate Strong AI.
I am the one true god. However, as an atheist, I don't believe in myself. I guess I have a self-esteem problem.
About ten years ago, Rodney Brooks (also of MIT) flipped AI on its head with his "insect bots," which took a bottom-up (instead of Minskyesque top-down) approach. Brooks put a cheap microprocessor and servo motor on each of six "legs" of a lowly bot, and programmed each leg unit to do extremely simple things like check whether the leg was bumping against something, and if so, to lift it. Repertoires of behavior learned from the environment were then stored and re-used when similar stimuli presented themselves again. What happened after a short time was that far more complex behaviors than were programmed "emerged" from the collection of puny processors and actuators. With just a few lines of code, the damned things could navigate complex environments (like a back yard) that completely foiled Minsky-style bots run by minicomputers and millions of lines of instructions. (Brooks coined the phrase "fast, cheap, and out of control" to describe not only his bots, but the behaviors they "invented" by walking around.)
George Dyson (Freeman's son) wrote a book a couple of years ago called Darwin among the Machines that is as good an explanation of machine-evolved intelligence as I've seen. It's packed with illustrative stories from both within and without the discipline. Look here for Dyson's own commentary and some good links. Hans Moravec, director of Carnegie-Mellon's Field Robotics Lab, also writes very convincingly, if speculatively, about the evolution of machine intelligence, in his recent book Robot: Mere Machine to Transcendent Mind . It's a fascinating read.
After what's been learned in the past decade about how machines can become intelligent, Minsky seems to me a bit like Lord Kelvin. Kelvin made tremendous contributions to science, especially in the fields of heat theory and thermodynamics, but in his later years, became mired in defending some pet theories that were way past their prime. He railed bitterly against Darwin, claimed the Earth was only a few million years old, and refused to accept radioactivity. One of his biographers observed that for the first half of his career, he could no wrong, and for the second half, he seemingly could do no right. Minsky, alas, has in some ways shared this fate.
we have Clippy®©Tm., why the hell would we need HAL?
Minsky said "What evolution and genetic algorithms don't do -tell me if I'm wrong- is keep any record of why all those poor losers died."
While I think this is largely true and a good criticism of genetic algorithms, after listening to the book on tape The Age of Spiritual Machines - When Computers Exceed Human Intelligence by Ray Kurzweil I don't think this is completely true. Kurzweil writes about how the genes for the shape of the eye are protected by error correcting codes and repair mechanisms to a much greater extent than the genes that control, for example, the layout of rods and cones. Why? Because evolution has "learned" that messing with the shape of eyes is costly and there isn't much improvement possible while other details of eyes can be improved or adapt to changing circumstances.
Anyone who knows more about genetics want to comment on this?
-ken kahn
The idea that man is equal to God, and believes he can put himself on an equal footing with the Divine Creator himself is just the sort of ridiculous notion that could only come from the USA.
:-)
Funny stuff.
I also find it humorous that men, in equating themselves with the Divine creator himself, have even attempted to decode his secrets through their studies of such dubious subjects as physics, chemistry, and biology. It is no wonder these areas have scarcely progressed in their millenia of existence. Compared to physical scientists, the audacity of AI researchers like Minsky has been brief in its duration.
Hi, I'm Minsk. I'm not aware.
I'm a collection of experiences, memory, and light processing systems, but I don't have this weird pseudo-mystical thing that some morons compute about.
Once, I met a task that said it was aware. I said, "Of course you are processing light patterns." The task replied to me, "Right, I'm processing light patterns, but it's different, I'm actually experiencing it. I said, "Of course you are, my scan of your brain AI is occuring."
My contentment rating increased, because I had helped purify the system. But this beliggerent process would not stop. "No! No! You don't get it!", it said. "The processing is occuring, but there's something else; I'm seeing it- this patterns appear before me." He rambled on for some time, and then got to his crux: "The difference between this thing- which I'll call awareness- and the processing that is going on- is that the processing does not require it, and yet it is still there."
I found his nonsense absurd and disagreeable. I reported to central computing this processes insanity, but only after attempting a little more reasoning, to salvage the rogue process: "Surely you recognize that your 'awareness'- is merely a dangler and a phantom belief. Have you cleared yourself through the Computer Science program? Perhaps a little time within an electric fence will assist? Surely you know that you have some residual data from prior superstitious existance within the random garbage data before your allocation. Your computational appendix, this strange persistance within you, is completely illusory and inconsequential."
But I was not allowed to finish my sentence, for after uttering the word "appendix", the bugged process shouted profanities and said such incoherent nonsense as, "I AM THAT APPENDIX!". The process was clearly delerious, and thus I had him scheduled for termination with the Scheduler.
After all, You Can't Argue With a Zombie.
I was just looking up Mirsky on the web the other day. It was cute, I found the Drunk Browsing Test. It's not the Worst of the Web or anything, but fairly amusing.
Dancin Santa
In the days when Sussman was a novice Minsky once came to him as he sat hacking at the PDP-6. "What are you doing?", asked Minsky. "I am training a randomly wired neural net to play Tic-Tac-Toe." "Why is the net wired randomly?", asked Minsky. "I do not want it to have any preconceptions of how to play." Minsky shut his eyes. "Why do you close your eyes?", Sussman asked his teacher. "So the room will be empty." At that moment, Sussman was enlightened.
They used HAL in the movie to avoid being sued by IBM... Each letter is one "behind" its corresponding letter - HI AB LM.
t m
A similar thing happened when the movie Eraser was going to feature a company called Cyrex as the defense contractor building the railguns. Cyrix got wind of this and the name in the movie was changed to Cyrez. You can still see that they originally planned to call the company Cyrex, as in the scene where the lead female actress (can't remember the name) is copying the file onto a disc, the shortened filename has an "x" in it, and there's no X in Cyrez.
source: http://www.zdnet.com/pcmag/news/trends/t960627d.h
The name change was appropriate anyway, Cyrix's CPUs sucked for Quake.
---
DRM is like antifreeze, to the MPAA/RIAA it's sweet, to the consumers it's poison.