Slashdot Mirror
MIT Finds 'Grand Unified Theory of AI'
aftab14 writes "'What's brilliant about this (approach) is that it allows you to build a cognitive model in a much more straightforward and transparent way than you could do before,' says Nick Chater, a professor of cognitive and decision sciences at University College London. 'You can imagine all the things that a human knows, and trying to list those would just be an endless task, and it might even be an infinite task. But the magic trick is saying, "No, no, just tell me a few things," and then the brain — or in this case the Church system, hopefully somewhat analogous to the way the mind does it — can churn out, using its probabilistic calculation, all the consequences and inferences. And also, when you give the system new information, it can figure out the consequences of that.'"
Tell me about you to build a cognitive model in a fantastically much more straightforward and transparent way than you could do before.
What is the difference between an endless task and an infinite task?
Sounds a bit like a journalists brain to me ... NO NO let me make up the rest of the Story
Told that the cassowary is a bird, a program written in Church might conclude that cassowaries can probably fly. But if the program was then told that cassowaries can weigh almost 200 pounds, it might revise its initial probability estimate, concluding that, actually, cassowaries probably can’t fly.
But you just induced a bunch of rules I didn't know were in your system. That things over 200 lbs are unlikely to fly. But wait, 747s are heavier than that. Oh, we need to know that animals over 200 lbs rarely have the ability of flight. Unless the cassowary is an extinct dinosaur in which case there might have been one ... again, creativity and human analysis present quite the barrier to AI.
Chater cautions that, while Church programs perform well on such targeted tasks, they’re currently too computationally intensive to serve as general-purpose mind simulators. “It’s a serious issue if you’re going to wheel it out to solve every problem under the sun,” Chater says. “But it’s just been built, and these things are always very poorly optimized when they’ve just been built.” And Chater emphasizes that getting the system to work at all is an achievement in itself: “It’s the kind of thing that somebody might produce as a theoretical suggestion, and you’d think, ‘Wow, that’s fantastically clever, but I’m sure you’ll never make it run, really.’ And the miracle is that it does run, and it works.”
That sounds familiar ... in both the rule based and probabilistic based AI, they say that you need a large rule corpus or many probabilities accurately computed ahead of time to make the system work. Problem is that you never scratch the surface of a human mind's lifetime experience though. And Chater's method, I suspect, is similarly stunted.
I have learned today that putting 'grand' and 'unified' at the title of an idea in science is very powerful for marketing.
My work here is dung.
Since the actual summary seems to involve a fluff filled soundclip without anything useful, here's the run down of the article.
1) We first tried to make AIs that could think like us by inferring new knowledge from existing knowledge.
2) It turns out that teaching AIs to infer new ideas is really freaking hard. (Birds can fly because they have wings, mayflies can fly because they have wings, helicopters can... what??)
3) We turned to probability based AI creation: you feed the AI a ton of data (training sets) and it can go "based on training data, most helicopters can fly."
4) This guy, Noah Goodman of MIT, uses inferences with probability: he uses a programming language named "Church" so the computer can go
"100% of birds in training set can fly. Thus, for a new bird there is a 100% chance it can fly"
"Oh ok, penguins can't fly. Given a random bird, 90% chance it can fly. Given random bird with weight to wing span ratio of 5 or less, 80% chance." and so on and so forth.
5) Using a language that mixes two separate strategies to train AIs, a grand unified theory of ai (lower case) is somehow created.
6) ???
7) When asked if sparrows can fly, the AI asks if it's a European sparrow or an African sparrow, and Skynet ensues.
Signatures are the new names.
This kind of probabilistic inference approach with "new information" [evidence] being used to figure out "consequences" [probability of an event happening] sounds very similar to Bayesian inference/networks.
I would be interested in knowing how does this approach compares to BN and the Transferable Belief Model (or Dempster–Shafer theory) which itself addresses some shortcomings of BN.
Ubuntu is an African word meaning 'I can't configure Debian'
HYPE. More grand unified hype. The "grand unified theory" is just a mashup of old-days rules & inferences engines thrown in with probabilistic models. Hyperbole at its finest, to call it a grand unified theory of AI. Where are connotations and framing effects? How does working short term memory interact with LTM and how does Miller magic number show up? How can the system understand that "john is a wolf with the ladies" without thinking that john is hairy and likes to bark at the moon? I could go on but feel free to fill in the blanks. So long and thanks for all the fish MIT.
When you use the phrase "Grand Unified Theory" you better have something impressive to show me.
From what I have seen 99% of AI research is only aiming to mimic AI.
From what I can tell this approach doesn't unite the field but instead tries to legitimize the 99%. In my opinion, that's a dead end.
A. A. Aaby: It's all about metaphor
Yours In Perm,
K. Trout
The summary reads like it was written by a 14 year old. Without reading the article, it is completely unclear what "this approach" is, how this cognitive model is different, and what "the Church" is. I know, read the article; but why would I if the summary makes me confused instead of curious?
I take it, then, that you prefer a bazaar to a cathedral?
I've abandoned my search for truth; now I'm just looking for some useful delusions.
q.v. Alonzo Church
http://en.wikipedia.org/wiki/Alonzo_Church
DRM: Terminator crops for your mind!
And why is his theory so grand?
From the article:
As a research tool, Goodman has developed a computer programming language called Church — after the great American logician Alonzo Church
Your comment fits the criteria of Flamebait and Offtopic, but definitely NOT Funny.
Thanks, Slashdot's mandatory comment waiting period! I'm sure glad I was late to this party.
DRM: Terminator crops for your mind!
I looked at the documentation of this "Church Programming language". Scheme and most other Lisp derivatives have been around longer and can do more. This is neither news nor a revolutionary discovery.
The way the author wrote the article, it seems like nothing different from an expert system straight from the 70's, e.g. MYCIN. That one also uses probabilities and rules; the only difference is that it diagnoses illnesses, but that can be extended to almost anything.
Probably the only contribution is a new language. Which, I'm guessing, probably doesn't deviate much from, say, CLIPS (and at least THAT language is searchable in Google... I can't seem to find the correct search terms for Noah Goodman's language without getting photos of cathedrals, so I can't even say if I'm correct)
AI at this point has diverged so much from just probabilities and rules that it's not practical to "unify" it as the author claims. Just look up AAAI and its many conferences and subconferences. I just submitted a paper to an AI workshop... in a conference ... in a GROUP of co-located conferences ... that is recognized by AAAI as one specialization among many. That's FOUR branches removed.
Pet peeve: Profane people propagating perfunctory pedantry.
Or perhaps I have respect for the work of Alonzo Church, and find such comments to be in bad taste. Perhaps my respect for his work even goes so far as to post with my screenname and reputation, rather than post unconstructive comments anonymously.
The key to Artificial Intelligence is to ignore the "intelligence" part and just think of it as Artificial Behavior.
Set your phasers on "funky"!
or are the comments ridiculously funny today. thanks for the fish, indeed.
See, that's not an AI problem, that's a semantics problem. The fact that you can mislead an AI by feeding it ambiguous inputs does not detract from it's capacity to solve problems.
A perfect AI does not need to be omniscient, it needs to solve a problem correctly considering what it knows.
You're not old until regret takes the place of your dreams.
Wow, as someone working in this domain I can say that this article is full of bold conjectures and shameless self-advertising. For a start, (1) uncertain reasoning and expert systems using it is hardly new. This is a well-established research domain and certainly not the golden grail of AI. Because, (2) all this probabilistic reasoning is nice and fine in small toy domains, but it quickly become computationally intractable in larger domains, particularly when complete independence of the random variables cannot be assured. And for this reason, (3) albeit being a useful tool and important research area, probabilistic reasoning and uncertain inference is definitely not the basis of human reasoning. The way we draw inference is much more heuristic, because we are so heavily resource-bound, and there are tons of other reasons why probabilistic inference is not cognitively adequate. (One of them, for example, is that untrained humans are incapable of making even the simplest calculations in probability theory correctly, because it is harder than it might seem at first glance.) Finally, (5) there are numerous open issues with all sorts of uncertain inference, ranging from certain impossibility results, over different choices that all seem to be rational somehow (e.g. DS-belief vs. ranking functions vs. probability vs. plausibility measures and how they are intereconnected with each other, alternative decision theories, different rules of dealing with conflicting evidence, etc.) to philosophical justifications of probability (e.g. frequentism vs. Bayesianism vs. propensity theory and their quirks, justification of inverse inference, etc).
In a nutshell, there is nothing wrong with this research in general or the Church programming language, but it is hardly a breakthrough in AI.
This is not a Grand Unified Theory of anything. Grandiose is the word that comes to mind.
If we program the logic for the AI and the AI system predicts outcomes it's based on the algorithm that is used to make predictions.
I cannot grasp how a computer can think of something that a human cannot because a computer only knows what we know. It is not capable of experience. As far as I can tell, the only thing AI can do is calculate something faster than humans can.
If you have a robot that learns how to move around a building without crashing into objects that learns through the experience of bumping into them it's just processing and responding as it was told to do.
Maybe I'm wrong. I'm not an AI expert but it all seems like a fancy way of saying, "I programmed a device to act how I wanted it to." All of the probabilistic data is analyzed by a person first. An AI device can only be as "intelligent" as it's creator.
Oh yeah, lets not show any respect at all to one of the greatest AI minds in history because you happen to dislike churches.
Asshole.
Security is mostly a superstition... Avoiding danger is no safer in the long run than outright exposure. - Helen Keller
This is just a library for Scheme. It does the same things that have been done before. In scheme.
Move along.
... You know the TV series that teaches us to not use 16 year old girls as the model for military robots.
Which, I'm guessing, probably doesn't deviate much from, say, CLIPS (and at least THAT language is searchable in Google... I can't seem to find the correct search terms for Noah Goodman's language without getting photos of cathedrals, so I can't even say if I'm correct)
not a google expert, are you? hint: try searching for church programming language...
Several hours too early for an April Fool.
Artificial intelligence is the study of how to make real computers act like the ones in the movies.
Again, as I bring up often with AI researchers, we as humans evolved over millions of years (or were created, doesn't matter) from simple organisms that encoded information that built up simple systems into complex systems. AI, true AI, must be grown, not created. Asking the AI if a Bat is a mammal and can fly can a squirrel? ignores a foundation of development in intelligence, our brains were created to react and store, not store and react from various inputs.
Ask an AI if the stove is hot. It should respond "I don't know, where is the stove?" Rather AI would try and make an inference based on known data. Since there isn't any the AI on a probablistic measure would say that blah blah stoves are in use at any given time and there is a blah blah blah. A human would put thier hand (a senor) near the stove and measure the change, if any in temperature and reply yes or no accordingly. If a human cannot see the stove, and had no additional information either a random guess is in order or a "I have no clue." response of some sort. The brain isn't wired to answer a specific question but it is wired to correlate independent inputs to draw conclusions based on the assembly and interaction of data and infer and deduce answers.
Given a film of two people talking a computer with decent AI would catagorize objects, identify people versus say a lamp, determine the people are engaged in action (versus a lamp just sitting there) making that relevant, hear the sound coming from the people then infer they are talking (making the link.) Then paralell the computer would filter out the chair, and various scenery in the thread now processing "CONVERSATION". The rest of the information is stored and additional threads may be created as the environment generates other links but if the AI is paying attention to the conversation then the TTL for the new threads and links should be short. When the conversation mentions the LAMP the information network should link the LAMP information to the CONVERSATION thread and provide the AI additional information (that was gathering in the background) that travels with the CONVERSATION thread.
Now the conversation appears to be about the lamp and wheather it goes with the room's decor. Again the links should be built adding, retroactively the room's information into the CONVERSATION thread (again expiring information that is irrelivant to a short term memory buffer) and ultimately since visual and verbal queues imply that the AI's opinion is wanted should result in the AI blurting out, "I love Lamp."
In case you missed it, this was one long Lamp joke...
-=[ Who Is John Galt? ]=-
We call it being Fashionably Redundant.
The enemies of Democracy are
"noah goodman ai church syntax" gives http://www.mit.edu/~ndg/ as the first result. There is a link near the top to http://projects.csail.mit.edu/church/wiki/Church
I tried to Google about Church programming language, and results were rather poor as one might imagine.
Then I found out the MIT wiki link where the code is stashed. It seems to be Scheme with some twist I'm not yet aware of though. The wiki seems to be a good introduction to Scheme also, as it starts from basics.
It is indeed one component of such an AGI, but it hardly qualifies as a "grand theory" of AI.
I think people at MIT are kind of jealous of AGI theorists, looking at the way they assert their claims of a "unified theory", as if they invented something wholly new and wonderful while making their uber-theoretical brains work on this grand problem that noone else ever thought about.
That is, after decades of dabbling with all sorts of nonsense like those stupid "gesture making" robots and whatnot, they come to realize that probabilistic inference is the key *now*? Like 50 years late?
And they needed the cognitive science department to figure that out? Is it because the AI lab is still infested by behaviorists?
Why didn't they just ask the theorists or make a survey of mathematical AI theories that have been in existence for several decades?????
Is it really surprising that a general purpose AI needs a) probabilistic inference b) a universal computer with probabilistic primitives?
In fact, those turn out to be _some_ of the axioms of a general purpose AI, discovered by Ray Solomonoff in the second half of 20th century.
I am laughing now.
--exa--
Tomorrow is the 1st of April folks.
This is embarrassing. MIT needs to get their PR department under control. They're inflating small advances into major breakthroughs. That's bad for MIT's reputation. When a real breakthrough does come from MIT, which happens now and then, they won't have credibility.
Stanford and CMU seem to generate more results and less hype.
FTA: "In the 1950s and 60s, artificial-intelligence researchers saw themselves as trying to uncover the rules of thought. But those rules turned out to be way more complicated than anyone had imagined. Since then, artificial-intelligence (AI) research has come to rely, instead, on probabilities -- statistical patterns that computers can learn from large sets of training data."
From the viewpoint of Jaynes and many Bayesians, probability IS simply the rules of thought.
A cassowary is a thing and an animal and a bird. Sometimes people call airplanes 'birds.' So if you learned blindly from literature, you could run into all sorts of problems. It's a danger you run if you learn and adjust these variables while following an ontology.
The fact is that if I thought up something, you would come up with the common sense logic to solve it and then wave your hand that it was already in the repository of knowledge (rule or probability or what have you) to solve the problem.
What I'm trying to tell you is that I've studied predicate calculus and prolog and various methods to achieve this. The problem isn't the system, the problem is replicating a human life (or even 18 years) of knowledge into whatever form is machine interpretable and this solution falls prey to these problems.
This is very promising.
Are you working in this field? This language has been around since 2008. How prevalent is it? Even the professor doing the research notes its pitfalls and expensive computations!
OR robotic systems used in manufacturing able to adjust the process as it goes. Using inputs to determine better ways to do a job.
Dangerous simplistic thinking. Adjusting a processing real time is never done. It's simulated in software first. You are being a science fiction author.
It looks like this system can change as it is used, effectivly creating a 'lifetime' experience.
If it's that easy, then do it. You will be the richest man alive before you die. That is, if you complete your project before you die :)
My work here is dung.
I've always enjoyed reading about AI, and like many here have done some experiments on my own time.
This week I've been looking for a simple state modeling language, for use in fairly simple processes, that would tie into some AI.
I wasn't really that impressed with anything I found, so when I saw the headline, I jumped to read the article.
Unfortunately, this is a step in the right direction, but not all that clear to write or maintain, and probably too complex for what I need to do.
The cleanest model to do these types of things I've found is the 1896 edition of Lewis Caroll's Symbolic Logic. (Yes, the same Lewis Caroll that wrote Alice in Wonderland).
It's human intelligence than I'm unsure about.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
The stagnant field of AI doesn't need to "unify" its failed approaches to modeling human thought, what it needs is something new and revolutionary.
I have always said, that psychology and nowadays even neurology, suffer massively from the Rube-Goldberg-machine syndrome. The brain is an extremely emergent. Perhaps the most emergent system known to man. Compared to the results, the basic rules are extremely simple. But they seem to try to analyze all those resulting effects, as if they were additional specific rules, instead of just results of the basic rules.
I am absolutely certain, that if you create a set of simulated life-forms based on “blank“ neural nets of a sufficient size, including hormones / neurotransmitters, and let them evolve trough natural selection so they modify themselves, that it is only a matter of time, until you will come up with a working life-form of the same or higher intelligence than a human. Of course this life-form will have a different base layout if it has different priorities. But there is no need for any additional rules, other than those.
And I am also certain, that I will be proven right in my lifetime. :)
Any sufficiently advanced intelligence is indistinguishable from stupidity.
... welcome our new Church overlords...
There! I said it. Now mod to oblivion etc etc...
I was going to comment that at a low level it looks to be similar to fuzzy logic in that it is using probability as thresholds for making decisions. If this is the case, then there isn't really anything groundbreaking about this model, since fuzzy logic has been around since the 1960s.
HA! I just wasted some of your bandwidth with a frivolous sig!
This may be the most succinct review I've ever read of "My Dinner With Andre"!
You shall see a cow on the roof of a cotton house.
Wow, butthurt at church much? Makes you wonder why. Or perhaps not...
You just got troll'd!
Why, thank you very much!... for illustrating my gripe with Goodman's approach by means of example. :)
That's EXACTLY why so-called "expert" systems such as MYCIN and Church will NEVER deliver on the promise of a grand theory of AI.
Pet peeve: Profane people propagating perfunctory pedantry.
...it's an inference engine with fuzzy logic rather than discrete logic, such that if you represented the inference training set in an N-ary tree, the fuzzy value is proportional to the fraction of branches in the tree that match a given inference. (They'd be better off with an S-curve, as that seems to be a better model for modeling real-world situations than a linear system.)
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
An example of an endless task is listing prime numbers. You can find any number of prime numbers (and there are an infinite number of them), and each new prime number found is an 'end point' of the task, but there will always be more endpoints.
Maybe "axilmar" is more interested in the ethics of AI than commercial gain. Maybe "axilmar" is getting ready to create a free cylon project that will eventually be completed by a Scandinavian student. Although "axilmar" never completes his own project, he'll consistently complain about the name of the newer, complete, more popular project and its derivatives. "Axilmar's" efforts will shift to creating and running the Free Cylon Foundation (or FCF). He spends the majority of his time give strikingly similar speeches over and over around the world. Despite the absolute consistency of his message he and by association the FCF are increasingly seen as a fringe political group. Despite the FCF's best efforts to promote the rights of the Cylons and hope for peaceful coexistence, the world's civilization eventually falls into chaos as the Cylons engage in war against humanity. Not long before his death at the hands of a cylon as he tries to convince the cylon that he's more righteous than other humans, "axilmar" is overheard muttering some complaint about a printer...
In Russia, Cognitive Model builds you
I hope it helps you to know that I clicked on your link instead of the AC's.
But then I realized the cable was blue, so I only gave it one star. I hate blue.
Whe I saw the headline, I thought it read "Grand Unified Theory of AL". I thought someone finally understood the life and mind of "Wierd Al" Yankovic.
I use irony whenever I can, but my shirts are still wrinkled...
I assume your comment is meant to imply that planes fly in a manner that is more similar to how birds fly than helicopters do. In fact, that's not the case at all. Planes may glide in a manner similar to how birds glide, but powered flight is something else entirely. Birds and helicopters (and insects, and bats, and pterosaurs) use their wings for both lift and thrust, something a fixed-wing-aircraft cannot do. Planes use their wings for lift only and must use jets or propellers for thrust.
Support Right To Repair Legislation.
This project is very fluffed-up and only works in limited settings with horrible runtime. Imagine a program that included a probabilistic element such as (true if coin flip is heads), where the coin could be biased to some probability p. Church lets you write a program using such elements, and then when you feed it data it can infer those parameters p in your program. The problem with this approach is that it requires tons and tons of sampling (MCMC on the space of possible programs (including recursion) with varied parameters).
We know that humans do not do random sampling to create a hierarchy of knowledge. Noah Goodman et al. (authors of this method) tried to run a workshop at a major AI conference asking whether the brain does this kind of random sampling. The resounding response from participants was no, it does not. The only thing that justifies the fact that these guys work in the Brain and Cognitive Sciences dept. is that they run psychological studies that validate the bayesian behavior of humans in limited scenarios. They don't actually study the structure of the brain; they only guess based on its macroscopic behavior. The implied claim is that since their computer model appears to have the same bar chart as humans, that they have captured some fundamental aspect of human intelligence. This could not be further from the truth.
Human intelligence is not merely an end-to-end phenomenon, it is an amazing capacity to make sense of an infinite stream of data using greatly constrained spatial resources in real time. If you tell me that intelligence is captured in an infinitely-recursive LISP program, I'll ask you how you create concepts from the ground-up over time. Infinite recursion is a sexy selling-point, but how do you actually implement this? How do you learn that letters compose words which compose sentences and so on? How do you reasonably capture knowledge which is more than two or three levels deep? Not with random sampling. We already know that the brain is more frugal than this.
So... Facebook and Twitter are the best AI we have?
For actual details check TFA: http://web.mit.edu/droy/www/papers/GooManRoyBonTenUAI2008.pdf http://projects.csail.mit.edu/church/wiki/Papers
Supposedly the software (Tribler) is already available.
there is no god but truth, and reality is its prophet
Do a search for articles with MIT in the title and you'll find that's a pretty common story here.
AI, unlike in the pure sciences, has no "answer" and therefore cannot have a grand unifying theory. There will never be a single algorithm that works for every type of problem we want to solve. AI is an applied science.
Besides, this stuff barely counts as "AI" in the modern sense. MIT embarrasses itself by pushing out stories like this.
I have a child. When I watch her learn its totally rules based. Also, very importantly when she is told that her existing rules don't quite describe reality she is quick to make a new exception (rule). Since she's young her mind is flexible and she doesn't get angry when its necessary to make an exception. The new rule stands until a new exception comes up.
eg in english she wrote "there toy" since she wasn't familiar with the other there's. She was corrected to "their toy". But of course, there is still "they're".
http://lolcode.com/
It's like lisp on steroids. Maybe I'm getting too old for somethings. I'll leave 'Church' to the computer scientist types with their PHDs
-- Many men would appreciate a woman's mind more if they could fondle it
The Slashdot admiration engine is all of the PR that any of these schools need. You'd hardly think that any meaningful AI research is done outside of these big 3.
We have rule based reasoning systems now for 40 years and we also have neural nets and other probabilistic systems as well. We even have systems which can work with time. And no wonder there have been hundreds when not thousands attempts to combine these three techniques. They have also a paper on the Church language http://www.mit.edu/~ndg/papers/churchUAI08_rev2.pdf
It is definitely nice, but it is not new and it is not the unified theory of thought (or AI). They don't know really how humans think. Nobody really knows. And honestly it is not really important as long as these machines are able to help me finding my data faster or got the the supermarket and by some stuff. A major problem in AI is, that something called I is present in humans it allows them to understand the world, computer systems do not have such an understanding. Which is no wonder, as we don't know what this I is. All rule sets and probabilistic trainings cannot answer the question. One thing can be said about this I, it is very subjective and it is not completely described as the knowledge of self existence and the ability to differentiate between I and the rest of the world. It is assumed that children learn this concept in their first year. The problem with that is, that they cannot be asked when and how they came up with this idea of I.
To make a long story short: We (humans) do not know enough about knowledge and decision processes that we can model them as effective as the processes work in human. And the MIT developed something others had build before. Maybe their language is a little bit better, than others. However, I doubt that, the AI community is not screaming loudly. And as long as other researchers are not running around telling us the MIT approach is the best since sliced bread, I wont buy it.
Don't you lose your geek membership card for not knowing who Alonzo Church was?
Comment removed based on user account deletion
You should have made a reference to Leonard Church instead.
Oh, say does that Star-Spangled Banner entwine / The myrtle of Venus with Bacchus's vine?
so they've added probability to the rule based systems, sounds like fuzzy logic? this is ground breaking? i must be missing something?
TFA doesn't contain a single maybe, not to mention THE DOUBT. The GUF (Grand Unified Theory) is about KILL THE DOUBT SWITCH. DOUBT is the way to go.
Excuse me.
The technical term is Hurd-Cylon, okay? Please use the correct term from now on.
Thanks,
Axilmar Stallman
Sounds like an ideal partner for Metal Storm together making a great Terminator. Add rules of engagement and you're away.
Crusader with Church OS.
Perhaps this was intended to be published on April 1st?
Whos to say AI's dont exist already and post on slashdot, would it be in their interests to make it public?
www.boznz.com Simple solutions to complex problems.