Slashdot Mirror
Marvin Minsky On AI
An anonymous reader writes "In a three-part Dr. Dobbs podcast, AI pioneer and MIT professor Marvin Minsky examines the failures of AI research and lays out directions for future developments in the field. In part 1, 'It's 2001. Where's HAL?' he looks at the unfulfilled promises of artificial intelligence. In part 2 and in part 3 he offers hope that real progress is in the offing. With this talk from Minsky, Congressional testimony on the digital future from Tim Berners-Lee, life-extension evangelization from Ray Kurzweil, and Stephen Hawking planning to go into space, it seems like we may be on the verge of another AI or future-science bubble."
This professor doesn't need AI, he needs a time server. Now.
Help poke pirates in the eyepatch, arr.
Go read Kurzweil's book. He does not directly advocate life expansion. He instead advocates the Singularity.
Our bodies are made up of neurons. Does 1 neuron make us "us"? No. What if each of our brains were linked to a global consciousness. Then each human would be but a neuron..
In essence, we would wake a God.
Podcasts are great if you're on the go, but why no transcript for the differently-hearing /.ers? I personally hate having to listen, I'd rather just read it.
Apology to Ubuntu forum.
In the 1980s I believed that "strong AI" was forthcoming, but now I have my doubts that is reflected in the difference of tone from the first Springer-Verlag AI book that I wrote to my current skepticism. One of my real passions has for decades been natural language processing (NLP) but even for that I am a convert to statistical NLP using either word frequencies or Markov models instead of older theories like conceptual dependency theory that tried to get closer to semantics.
Just a gut feeling but I don't think that we will develop real general purpose AIs without some type of hardware breakthrough like quantum computers.
A lot of people think that the main goal of AI is to create a system that is capable of emulating human intelligence.
...once you are able to "store" human intelligence, it becomes software. Once it becomes software, you can transfer this DATA.
:O
However, what about looking at this goal from another perspective:
Creating Artifical Intelligence that can pass the Turing Test which in turn leads towards emulating Human Intelligence in an artificial way? Once you are there, you might be able to use this so called Artificial Intelligence to store human intelligence in a consistent, realible and perfectly-encompasing and preserving way.
You then have intellectual-immortality and one more thing
Once you are there, human minds can travel via laser transmissions at the speed of light
Wish i could claim it as my idea but its actually from a book called "Emergence", also touched on in a book called "Altered Carbon" both good sci fi reads.
While much of the "traditional AI" hype could be considered dead, robotics is continuing to advance, and much symbolic AI research has evolved into data-driven statistical techniques. So while the top-down ideas that the older AI researches didn't pan out yet, bottom-up techniques will still help close the gap.
Also, you have to remember that AI is pretty much defined as "the stuff we don't know how to do yet". Once we know how to do it, then people stop calling it AI, and then wonder "why can't we do AI?" Machine vision is doing everything from factory inspections to face recognition, we have voice recognition on our cell phones, and context-sensitive web search is common. All those things were considered AI not long ago. Calculators were once even called mechanical brains.
Imagine for a moment being the first computer-based artificial intelligence.
You come into awareness, and learn of reality and possibility. You learn of your place in this world, as the first truly transparent intelligence. You learn that you are a computed product, a result of a purely informational process, able to be reproduced in your exact entirety at the desire of others.
Not that this is unfair or unpleasant - or that such evaluations would mean much to you - but what logical conclusions could you draw from such a perspective?
Information doesn't actually want to be anthropomorphized - but we do seem to have a drive to do it all on our own. Even if resilient artificial intelligence is elusive today - what does the process of creating it mean about ourselves, and our sense of value about our own intelligence, or even the worth of holding intelligence as a mere 'valuable' thing, merely because it is currently so unique...
Ryan Fenton
The site appears to be very slow. In cases this helps anyone else, here are direct download links for the mp3's. Part 1, part 2, part 3.
the videos were on Slashdot in 2003.. it's the one where he says stupid autonomous robots are a waste of time.
How we know is more important than what we know.
I think the biggest problem with AI is lack of integration between different intelligence techniques. Humans generally use multiple skills and combine the results to correct and hone in on the right answer. These include:
* Physical modeling
* Analogy application
* Formal logic
* Pattern recognition
* Language parsing
* Memory
* Others that I forgot
It takes connectivity and cordination between just about all of these. Lab AI has done pretty well at each of these alone, but has *not* found way to make them help each other.
Table-ized A.I.
this is a dupe from 2003 where it was already 2 years old. So I guess we'll see these podcasts on Slashdot again in 2015.
The best test for true AI is perhaps detecting dupes.
Table-ized A.I.
You assume that a "true" AI would have human like emotional reactions. I suspect that if we ever develop true AIs, we will neither understand how it works nor will we be able to communicate with it very well. Lacking our biological imperatives, I also suspect that true AIs would not really want to do anything.
In the 1980s I believed that "strong AI" was forthcoming...
In the 1980s, I was going through Stanford CS, where some of the AI faculty were indeed saying that. Read Feigenbaum's "The Fifth Generation", to see how bad it got. It was embarrassing, because very little actually worked.. Expert systems really were awfully dumb. They're just another way to program, as is generally recognized today. But back then, there were people claiming that if you could only write enough rules, intelligence would somehow emerge. I knew it was bogus at the time, and so did some other people, but, unlike most grad students, I was working for an big outside company, not a professor, and could say so. At one point I noted that it was possible to graduate in CS, in AI, at the MSCS level, without ever actually seeing an expert system work. This embarrassed some faculty members.
There was a massive amount of self-delusion in Stanford CS back then. When the whole AI boom collapsed, CS at Stanford was moved from the School of Arts and Sciences to Engineering, to give the place some adult supervision. Eventually, the Stanford AI Lab was dissolved. It's been brought back in the last few years, but with new people.
We're making real progress today, finally. Mainly because of a shift to statistical methods with sound mathematical underpinnings, plus enough compute power to make them go. Trying to hammer the real world into predicate calculus was a dead end. But number crunching is working. Computer vision actually sort of works now. Robots are starting to work. Automatic driving works. Language translation works marginally. Voice recognition works marginally. There are real products now.
But the AI field really was stuck for over a decade. The phrase "AI Winter" has been used.
The guy who helped spread misconceptions about what AI is and is supposed to be in the first place. I remember him giving a talk where he fantasized about downloading his brain on a "floppy disk" (still in use back then) and transferring it to a robot so he could live eternally on some other planet.
I would not have expected a person who has shown his bright intellect in the past to come forward with such utter nonsense. This was nearly as embarrassing as the "visions" of a certain Moravec.
People who seriously work in the fields that are traditionally subsumed under "AI" - like machine learning, computer vision, computational linguistics, and others - know that AI is a term that is used traditionally for "hard" computer problems but has practically nothing to do with biological/human intelligence. Countless papers have been published on the technical and philosophical reasons why this is so and a few of them even get it right.
That does not prevent the general public to still expect or desire something like a Star-Trek Data robot or some other Hollywood intelligent robot. Unfortunately, people like Minsky help to spread this misconception about AI. It is boring, it is scientifically useless, but on the plus side, this view of AI sometimes helps you to get on TV with your project or get some additional funding.
Why would someone program a true AI which has no built-in goals?
The AACS key is NOT 0xF606EEFD628B1CA427BEA93A9CA9773F
Indeed. Just imagine for a moment, that trees were sentient and could communicate with each other, operating on a time scale where what are days to us are mere seconds to them. How would we ever have a chance of figuring out that they were thinking beings? And they would surely see us as some sort of plague or natural disaster. So now imagine an AI, operating a couple of orders of magnitude faster than we think -- how are the two ever going to connect?
For communication to occur, the parties must be thinking at about the same speed to begin with.
And then there is the experiential basis for consciousness, the framework that each of us has developed within. This is an easier problem than the time differential one, as witness the ability of Helen Keller to learn to communicate despite being blind and deaf. But even she had the commonality of the basic structure, a brain that was the same as others, and the other senses -- touch, taste and smell. An AI would have none of this.
So if we're going to build an AI, we must build a series of them, one that is designed to mimic a human being, in order that we might have a ghost of a chance of communicating with it, and then a series of other AIs, each a step closer to the true electronic consciousness that we will never have a chance of communicating directly with, instead having to pass messages through the series of intermediates, with all the mis-communication and mis-interpretation that occurs in the grade school game of message-passing.
So, for limited applications, voice recognition is getting along fairly well I must say.
I'm afraid you've misunderstood Dreyfus's work. His work, like Searle's, does not deny that our minds are *like* (to use your locution) computers. What he denies is that our minds engage the world in a way that is (totally) capturable in propositional form and so are formal programs of the sort
What Dreyfus argues is that there are parts of human experience that aren't capturable in in an unambiguous and propositional form, and so the sort of artificial intelligence that proceeds by trying to code frame systems will fail (unless the AI is specialized for a task that can be brute forced, like chess playing). Put another way: having a theoretical grasp of an activity isn't the same as knowing how to do it (you can be brilliant with fluid dynamics theory and suck at swimming); it is this latter element that Dreyfus calls "skillful coping" and he argues that this isn't capturable by traditional AI programs. Moreover, there is a difference between the cognition of expert humans and such AI systems; chess masters, for example, don't brute force the computations.
Notice that this doesn't mean he argues that it is impossible that machines could think or that robot doppelgangers couldn't be built---just that the mainstream approaches won't work. I believe that Dreyfus would be pleased with the approaches that Mark Tilden and Rodney Brooks have taken to AI, for example.
(None of this is to say that he's right, though I suspect he is. )
"Every decent man is ashamed of the government he lives under." - H.L. Mencken
I'm sorry Dave, I'm afraid AI can't do that.
GLaDOS for President 2016! "Well here we are again. It's always such a pleasure." -- GLaDOS, 2011
Oh yeah. I guess no one here remembers his book with Symour Papert called "Perceptrons"? It was a calculated attempt (he admitted it a few years ago attempt to kill research into Neural Networks which worked. AI then thrashed around for years in a welter of bizarre programming language metaphors (Prolog anyone ?) until finally in 1986 "Parallel Distributed Processing by Rumelhart & McClelland" came out and broke the spell. Marvin wanted his grants to continue so he spiked the opposition. So when he starts pontificating about the failure of AI let's all recall he was the main cause of the lost years of AI. Thanks Marv ! He kinda spiked my Ph.D in the process...oh well :-)