Slashdot Mirror
DARPA Contracts For AI Technology
heptapod writes "USA Today is reporting that DARPA has contracted two professors from RPI to develop artificial intelligences that can learn by reading and understanding natural language. Interesting taking DARPA's Grand Challenge into account. Mentioned in the article is Cycorp, Inc. which has been pursuing this goal since 1994!"
"First passing of the Turing test!"
DARPA announced today the funding for Skynet.
What they are doing is very interesting. By compiling the majority of human knowledge into a gian database, it should make AI development much easier to pursue.
Teaching a machine to read a text book and answer questions doesn't necessarily mean cognitive reasoning. It's just a new form of input/output. Ask it to write an essay with a definative argument and solid conclusion based on the material read would impress me, not regurgitating facts and figures found in a book.
"It's difficult to meditate on amphetamines." - Joe Walsh
I have enough dynamite to blow up 10 super-computers!!!
...artificial intelligences that can learn by reading and understanding natural language...
OK, but can it learn from mistakes?
OpenCYC.org project Sourceforge CVS repository has not beent updated since October 22nd 2003. I hope some of that DARPA money will go a little way towards completeting the 1.0 release.
as you know we non americans cannot access darpa.mil :)
:)
If something is kind enough to give us a mirror to the "Great Challenge", kudos to him
Or else I'll go through a US proxy. Not a big task, it's just annoying, I'll do that later.. grab an anonymous US proxy on www.proxy4free.com , enter the crap in your browser and enjoy the slowness. Maybe I'll use switch proxy this time
Can it read books on AI; and then design a better, smarter AI?
I am the unwilling control for my Origin.
..I mean, they'd never use this technology in meat eating robots..
..right???
air and light and time and space
There was an article that hit the New York times back in the fall of 2004 mentioning Professor Bringsjord and having a computer program write fiction. Perhaps this is to better write stories for elected officials to tell?
For all of the thousands of times I've read the phrase "artificial intelligence," it's only recently occurred to me to wonder whether there's a point in using "artificial." Certainly the first flavors of this are at best insect-like, or sort of idiot-savant (like chess playing), but when we first experience a system that's as awake as we're all hoping for... then it's just "intelligence," isn't it?
I know - read four thousand sci fi novels and then come back to this conversation... but it seems that the "artificial" of this phrase is increasingly awkward. It makes some people dismissive about the potential, other people feaky about the same, and seems destined to always shortcut the philosophical payload. Not because I fret over the machine's eventual feelings (though if it's Linux-based, I'm sure it will have very warm, friendly, altruistic feelings), but because by boxing code-based intelligence into the "artificial" category, it props up the more mystical perception of our own native smarts.
The very word, from "artiface," suggests that whatever it will be, it won't really count as intelligence. But we're very comfortable (or at least I am) talking about, say, an intelligent dog or primate. So, if we can even approach that with a system that isn't any more fragile than walking, breathing meat... then surely that's not artiface? OK, smack me around now. Thanks.
Don't disappoint your bird dog. Go to the range.
Just to be pedantic ... Cyc as a project has been around at lot longer than Cycorp the company ... since at last 1986, according to this Google/Deja Groups post.
Each class should have its own championship title of $1 million, with each runner-up winning 1/2 the money of the next higher.
Each contestant must provide 2 systems -- a compressor and a decompressor. DARPA feeds the compressor the corpus and the compressor feeds DARPA the compressed corpus. DARPA then measures the ratio and feeds the decompressor system the compressed corpus, which then returns the original corpus, or is disqualified. Compression and decompression times must add up to no more than the time limit for the competition class.
The rationale for this approach to advancing the state of AI is given by a short paper by Matthew Mahoney titled "Text Compression as a Test for Artificial Intelligence" (1 page poster, compressed Postscript) published in the 1999 AAAI Proceedings. Matt Mahoney shows that text prediction or compression is a stricter test for AI than the Turing test.
So far there have been lots of promises and decades spent. Let's try something different with well-founded objetive metrics tied to serious near-term commercial incentives for evolutionary progress.
Seastead this.
It's a $400k grant with two optional extensions. The school will take half, the profs will take part of their own salaries out of it, and then it'll support a couple PhD and MS students. This is no big deal.
AI has always "failed" because every time it's succeeded, the problem it succeeded on has been retroactively defined to "not require intelligence". Cf. automated theorem proving, chess playing, control of chemical plants, and just about any other AI success of 1940s - present.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
it has always interested me how someone who believes in pure evolution (i.e. order from absolute randomness - disregarding whether there is such a thing or not) believes that we have a chance in hell of actually designing an AI.
If evolution is true, then the things that we call "order" and "intelligence" are just a higher function of chaos (the inevitable byproduct of randomness). On an even higher level, there is no reason to believe that we are actually designing anything, we are merely exciting our neurons (if they exist) into believing we have perceived that we are performing an action (which in this case is mental, which brings us back to the alleged neurons) that we call designing. If evolution is true, then intelligence will happen regardless of what we do, and we have no reason to believe that we have anything to do with it whatsoever, or could influence it in any way at all if we did.
As for me, I'll take an Almighty God (as long as he lets me)
--
Was it the sheep climbing onto the altar, or the cattle lowing to be slain,
or the Son of God hanging dead and bloodied on a cross, that told me this was a world condemned but loved and bought with hlood
It's not just canned questions and answers; it has an inference engine. It can do "if A is B and B is C, then A is C". But only if all the right predicates match perfectly.
Lenat was claming it would somehow become intelligent in a few more years. That was a decade ago. Today, Cyc is regarded as the definitive demonstration that that idea won't work.
Here's a critique of Cyc from 1994.
This is probably a good place to mention the No Evil Robots campaign...
t ml
And for a glimpse, if somewhat longwinded, of what lengths DARPA will go to to make this happen, check out this article: http://villagevoice.com/news/0337,baard,46901,1.h
Cylons?
This comment isn't Informative. It's either mistaken or a liar. That or he's in a country that is actively blocked. UK and Australia can access the site just fine.
Teaching a machine to read a text book and answer questions doesn't necessarily mean cognitive reasoning. It's just a new form of input/output.
... reading 24 senses of verb "read" from WordNet... beginning syntactic analysis: in context of "teaching a machine," either means: (a) this is the indirect object of "teaching" -- "teaching" frame indicates that there may be an infinitive of a skill verb; and several of the senses of verb "read" are semantically associated with "skill" sets; (b) this is a semantic "larger purpose" of "teaching", i.e., all verbs support chained infinitives that imply dependencies, e.g., "buying a cake to eat for dessert". Which is more likely, that the author intends that (a) one of the "skill" senses of "read" is the skill being taught to the political party, or (b) the teaching of the political party is a subgoal of the author's goal of "reading"?
Parsing post
Teaching
[Teaching] - one lexical interpretation: gerund form of "to teach". Part of speech? Unambiguous. Noun. Word sense of Teach? Options: accessing Wordnet... 2 verb senses found... must choose between: v 1: impart skills or knowledge to; "I taught them French"; "He instructed me in building a boat" [syn: learn, instruct] 2: accustom gradually to some action or attitude; "The child is taught to obey her parents"... no semantic distinction possible at this point.
a machine
Accessing WordNet... 6 noun senses found: n 1: any mechanical or electrical device that transmits or modifies energy to perform or assist in the performance of human tasks 2: an intricate organization that accomplishes its goals efficiently; "the war machine" 3: an efficient person; "the boxer was a magnificent fighting machine" 4: 4-wheeled motor vehicle; usually propelled by an internal combustion engine; "he needs a car to get to work" [syn: car, auto, automobile, motorcar] 5: a group that controls the activities of a political party; "he was endorsed by the Democratic machine" [syn: political machine] 6: a device for overcoming resistance at one point by applying force at some other point [syn: simple machine]
Syntactic analysis: noun phrase following gerund... if formed correctly, this is most likely a gerund phrase. The act of teaching done to machine. 12 possible word sense conjuncts total (2 for "teaching", 6 for "a machine.")
Accessing semantic module... which of 12 is most likely the author's intention?
Accessing language library... accessing semantic database... is it possible to impart skills or knowledge to a group that controls the activities of a political party? Semantic database says: "group" implies "people." "People" can be taught under most circumstances. Therefore, yes. That is a reasonable interpretation. Now, is it possible to impart skills or knowledge on any mechanical or electrical device that transmits or modifies energy to perform or assist in the performance of a human task? No, because "any" implies that one could impart skills on a pulley, since a pulley is a machine. But semantic database says that pulleys cannot learn. Either semantic database is wrong (flag this as possible new knowledge), or first interpretation is more likely.
Associating linguistic entity "teaching a machine" with semantic idea canonicalized by "imparting skills or knowledge to a group that controls the activities of a political party".
to read
Infinitive form of verb
Accessing semantic component... trying interpretation (b)... trying "read" sense (1): To examine and grasp the meaning of (written or printed characters, words, or sentences). Accessing world literary rates... if author wrote sentence, chances are very high the author knows to read already, therefore lowering his chances of desiring to learn to read. Trying "read" sense (2) To utter or render aloud (written or printed material). Is it reasonable that th
If there's one thing that the last 60+ years of research into artificial or machine intelligence has shown is that there is no clear definition of intelligence. There are different types of intelligence for example muscle control, visual processing, tactile interpretation, olfactory classifying, and so on. With these rough subdivisions great strides has been made in creating successful "modules" for them, but what has eluded and probably will stay elusive for the near future is the general cognitive intelligence that orchestrates the interplay between the rough subdivisions.
Shh.
The parent is suggesting that artificial selection is proof against natural selection.
And you can't breed dogs or horses or humans or anything else to enhance a specific trait can you?
The fact of the matter is that we are fundamentally no different from the amoeboid life we evolved from, and the rest of the life that evolved from it, just more complicated. If simple insectoid neuro circuitry can be approximated with simple neural nets (read this for more info on this highly debated subject) it could easily be argued that it is not the distinction between artificial and "natural" intelligence that should be question/examined but the existence and definition intelligence itself, and quite possibly life for that matter. These are concepts as arbitrary and ill-defined as the spirituality that their nay-sayers flaunt so wantonly in protest.
For christ's sake (pun and capitalization intended), think before you flap your rot. (There's just no escaping them on this subject)
Of blankness, I know nothing.
This AI and natural language thing gives ne deja-vue
Don't worry about that, it's just a glitch. It happens sometimes when the AI's change something.
"And then I visited Wikipedia
It was Natural selection the whole time.
? ascid=214
e =utf-8&oe=utf-8&q=clay+synthesis+of+RNA
The process is geared to produce things that are: a) Hardier and better equipped at survival, b) better equiped to reproduce themselves in the environment.
This applies to the basic chemical components and the proteins and the organisms and the etc. The more stable and reproductive a system is the more of them there are likely to be for a longer period of time. The End.
Read about RNA, it's ability to reproduce in small strands and the abiotic clay-catalyzed synthesis of RNA. Here is one link of the thousands available online: http://www.astrobiology.com/asc2002/abstract.html
And here are that other thousand (actually 21,600) I was mentioning: http://www.google.com/search?sourceid=mozclient&i
Of blankness, I know nothing.
The problem here is that there is a distinct disregard for the fact that order cannot exist without chaos, and vice versa. Order and chaos are in fact one and the same. Order comes from chaos, and order returns to chaos every day, but people refuse to see it. Take a simple example, hurricanes. Hurricanes are very ordered structures that exist for indefinite periods of time. But hurricanes cannot exist forever without the right conditions of material and energy. Also the sun a temporary manifestation of order from the chaos of the galaxy. The sun will also return, one day, to disorder. We are also order derived from chaos, but we will also return back to chaos in the end.
You might imagine the chaos of the universe as being total order, but all scrambled up. The universe itself may be conceived as a giant spring that compresses to total order, then relaxes into total disorder. That is, unless you hold as I do that the universe actually exists forever, in continuum. In this case, chaos would be just all the order that ever was all scrambled up into disorder, which does on the surface of things seem totally random. And, by any measure over the infinite time of the universe -is- random. But, within chaos lie tiny seed like domains that sprout like lillies in the field into temporary islands of order, destined at some future date, to return to disorder.
And this is the nature of the way things are. Once you realize this simple fact, there is no escaping that we exist because we were always destined to exist. And our lives are but brief periods in the continuous change that is and forever will be. We exist now, and we will exist again, but not in the same form or flavor.
Even if you don't believe in a universe that is forever you must rightfully acknowledge that all information must naturally, eventually be "destroyed". It is not possible to record data forever, for all data must eventually be erased. There is no "medium" for which to store an infinite amount of data. Unless that is if you beleive in an infinite universe, in which case, the medium itself must scramble all existant data so that it is ultimately unintelligible, unusable for any purpose whatsoever at any point in future time.
Ultimately the only reason for existance of any AI is to perceive that it exists, and to survive for however long it can within the confines of its chaotic environment. But, this is not a "reason", this is just "what it does". The capacity to create order from chaos is seen as constructive (the good), rather than destructive (the bad). For it takes energy to create order from chaos. It is "easy" to destroy, but hard, very hard to create. This is because the flow of entropy works against you to create. Therefore create, don't destroy. Work hard, don't be lazy. These are the principles of which I have adoped as my ethic of life.
The thing known as "god", is an irrational concept and no irrational concepts can be verified. Square things are not round, heavy things are not light, and light is not dark, unless you live in the world of Alice in Wonderland.
There is no reason to slain the cattle, and no reason to die for sins, for I have none, there are no such things as sins. Spare no blood for me child as you do not know the truth as it is, as things are. Truth is a measure of the difference between the way you believe things to be, and the way things are. Be constructive, search for the way things truly are, not how you believe them to be. Only then will your time be not spent in vain.
I started working on Cyc in 1985 and can assure you that it did _not_ start in 1994. They already had a year or two under their belt when I showed up.
Everyone I tell about this calls me crazy but I dunno. It's a far fetched idea but it might be possible.
Whenever people start to make an AI project, they want to start building it from from the middle. The projects have so much trouble making a stable base for themselves that they often never make anything at all. Other projects create soul-less intelligence. Complex, learning, logical machines with no purpose, direction or desire. They know nothing but what they do every day, usually process data and make new data processing rules based on that data. Sure, that's intelligence , but it's not what we're looking for.
The human race is looking for a digital companion. A little guy in a computer that can think, feel, and reason like a person. Then we want to speed that person up to do jobs as well as a person, but faster.
Well, that's not going to happen the way things are going now. I'd like to pose a question to the slashdot community: Do we know enough about physics on an atomic scale that we could simulate a "small room on earth" environment all the way down to an atomic level? Could we model and place in that simulated room a fertilized human egg inside what would be a functional machine to mature the egg into a fetus and release it when ready? (The machine doesn't have to follow all the simulated rules, we could just insert stuff into it using the computer). We could basically give birth to a simulated person.
It's a crazy idea, I know, and with current technology, the simulation would be unbearably slow, but my question is: is such a thing possible? Do we understand physics on an atomic level well enough to do something like this?
Long answer : Your question is the fundamental reason why the field of Statistical Mechanics exists in the first place. We know the laws of physics very well at the atomic level, but all the inter-particle forces will grow exponentially. Take a picogram of water, which would encompass a sphere with 60 micron radius, of similar size to a human egg, as per your request. Such a 'small' quantity of water will contain about 100 billion atoms (3 atoms per water molecule). This would be a very simple system of only water, without complications of DNA, proteins, and other organics.
However - chemical processes are primarily governed by electron interactions (between themselves and nuclei). For simplicity, one could probably model the nucleus merely as a simple charge, ignoring individual protons and neutrons, at least to first order. But the electrons must be independent, so this would leave each water molecule with 3 nuclei and 10 electrons. So that would really be about 300 billion charged entities to model. Assuming only Coulombic interactions (charge-charge repulsion/attraction) between charge pairs, there are about 5e23 such interactions to model (all individual pairs that can be produced), just for calculating the forces to advance the system from one state to the next (ignoring summation and momentum considerations). If you had a 1 Teraflop cluster, and assuming you can do one calculation per clock cycle (very generous), it would take about 15,000 years just to make one small time evolution of the system!!!
Now account for quantum mechanics (essential in system of this size, especially for molecular electron interactions) and the extra baggage of maintaining the wavefunctions (or doing an ensemble average of wavefunction expectation values). Then add in more complexity to allow for DNA and other organics. Then do enough time evolutions to advance the system far enough to see the interactions of interest. Our sun will be long burnt out by that time.
Hell, when you take an elementary course in quantum mechanics, you see that modelling an 'ideal' hydrogen atom is doable. By ideal this means ignoring relativity, interactions between electron spin and it's orbit, interactions between electron spin and proton spin, etc etc. Add in these real factors and it becomes much harder. Although such a system you could probably make more approximations, such as assuming exponential charge screening, which means the Coulomb forces would act only in a local area. But still the processing time would be incredible.
Then when you try to model something more complicated, like Helium, it gets VERY difficult. Even the best simulations nowadays can't use too many particles for a real macroscopic system, because you need to do enough averaging to get worthwhile results, but you also cannot wait an eternity.
So that's why statistical mechanics is used, if you have a room full of air, you cannot model all the individual nitrogen and oxygen molecules bouncing off each other, but you can determine average behavior, such as pressure and temperature. And with statistical mechanics you can calculate the relative uncertainties of these quantities (ie, how much variation you'd expect in such a measure of a quantity that's defined as an average anyway), which gives it more utility than thermodynamics. But doing statistical mechanics of a very complicated system with DNA, proteins, and other organics, and accounting for quantum mechanics, would quickly become extraordinarily difficult to model and calculate as well. There would be enough individual parts there (DNA sequences, for instance) that you'd encounter the same difficulties just described.
make world, not war
Arbitrary implementation of this algorithm will likely yield a very stupid system. Intelligence lies in configuring the way each part runs. Even the human mind does not have instant answers to every problem.
There are many different ways to augment the algorithm to achieve a practical intelligence, especially the areas of instinctive knowledge, sensory input, and force control.
Know your pads. One time pad: good for cryptography. Two timing pad: where to take your mistress.
The ALICE AI Foundation http://www.alicebot.org/ supports the development and adoption of free AIML software and standards for natural language chat robot technology. The ALICE brain, available freely under the GNU public license, is the three time winner of the prestigious Loebner prize for "most human computer" in a contest based on the Turing Test. One of the most interesting AIML implementations, Program N http://www.aimlpad.com/ by Gary Dubuque with contributions by Kino Coursey, already incorporates OpenCyc and WordNet into the ALICE conversational interface. The Foundation derives income from individual and corporate memberships, bot subscriptions, books, the Foundation directory, consulting, teaching, awards, Google ads, gifts and donations. We have never accepted one dime of DARPA or other government sponsorship.