Slashdot Mirror
Can Curiosity Be Programmed?
destinyland writes "AI researcher Jurgen Schmidhuber says his main scientific ambition 'is to build an optimal scientist, then retire.' The Cognitive Robotics professor has worked on problems including artificial ants and even robots that are taught how to tie shoelaces using reinforcement learning, but he believes algorithms can be written that allow the programming of curiosity itself. 'Curiosity is the desire to create or discover more non-random, non-arbitrary, regular data that is novel and surprising...' He's already created art using algorithmic information theory, and can describe the simple algorithmic principle that underlies subjective beauty, creativity, and curiosity itself. And he ultimately addresses the possibility that the entire Universe, including everyone in it, is in principle computable by a completely deterministic computer program."
'nuff said
Oh wait, you're not talking about children... nevermind.
#!/bin/sh
for i in who what where when why how; do
echo "But $i, dad?"
done
I hereby submit this project to the /. community under the GPL v2.
be ported to sega cd?
001 Gather data
002 Hypothesise
003 Go To 1
After logging in slashdot still does not take you back to the page you were on. It's been that way for 20 years.
The problem with this is that you need to be outside the universe in order to do so, you can't calculate the universe from within itself any more than a VMWare can run a machine faster than the host processor.
You'd also need more mass in your computer than exists in the universe, observable or otherwise.
So sure, I'll go with the theory that its possible, just not by any thing in our universe.
Likewise, nothing in our universe could leave it to perform the calculation elsewhere, as doing so links the two realities together, so you now need to simulate both.
Everything is interconnected and the very act of attempting to simulate the universe changes the simulation. Every new version of the simulation would instantly require a new version to take into account the changes from the previous version.
The theory is ... cute at best, but unworkable.
Persistent Volume manager for Kubernetes - https://github.com/dwimsey/openshift-pvmanager
Somehow it always comes down to being 42.
Say hello to my little sig.
"The Mind Has No Firewall"
Army article on psychotronic weapons
>>> The following article is from the US military publication Parameters, subtitled "US Army War College Quarterly." It describes itself as "The United States Army's Senior Professional Journal." [Click here to read a crucial excerpt.]
"The Mind Has No Firewall" by Timothy L. Thomas. Parameters, Spring 1998, pp. 84-92.
The human body, much like a computer, contains myriad data processors. They include, but are not limited to, the chemical-electrical activity of the brain, heart, and peripheral nervous system, the signals sent from the cortex region of the brain to other parts of our body, the tiny hair cells in the inner ear that process auditory signals, and the light-sensitive retina and cornea of the eye that process visual activity.[2] We are on the threshold of an era in which these data processors of the human body may be manipulated or debilitated. Examples of unplanned attacks on the body's data-processing capability are well-documented. Strobe lights have been known to cause epileptic seizures. Not long ago in Japan, children watching television cartoons were subjected to pulsating lights that caused seizures in some and made others very sick.
Defending friendly and targeting adversary data-processing capabilities of the body appears to be an area of weakness in the US approach to information warfare theory, a theory oriented heavily toward systems data-processing and designed to attain information dominance on the battlefield. Or so it would appear from information in the open, unclassified press. This US shortcoming may be a serious one, since the capabilities to alter the data- processing systems of the body already exist. A recent edition of U.S. News and World Report highlighted several of these "wonder weapons" (acoustics, microwaves, lasers) and noted that scientists are "searching the electromagnetic and sonic spectrums for wavelengths that can affect human behavior."[3] A recent Russian military article offered a slightly different slant to the problem, declaring that "humanity stands on the brink of a psychotronic war" with the mind and body as the focus. That article discussed Russian and international attempts to control the psycho-physical condition of man and his decisionmaking processes by the use of VHF-generators, "noiseless cassettes," and other technologies.
An entirely new arsenal of weapons, based on devices designed to introduce subliminal messages or to alter the body's psychological and data-processing capabilities, might be used to incapacitate individuals. These weapons aim to control or alter the psyche, or to attack the various sensory and data-processing systems of the human organism. In both cases, the goal is to confuse or destroy the signals that normally keep the body in equilibrium.
This article examines energy-based weapons, psychotronic weapons, and other developments designed to alter the ability of the human body to process stimuli. One consequence of this assessment is that the way we commonly use the term "information warfare" falls short when the individual soldier, not his equipment, becomes the target of attack.
Information Warfare Theory and the Data-Processing Element of Humans
In the United States the common conception of information warfare focuses primarily on the capabilities of hardware systems such as computers, satellites, and military equipment which process data in its various forms. According to Department of Defense Directive S-3600.1 of 9 December 1996, information warfare is defined as "an information operation conducted during time of crisis or conflict to achieve or promote specific objectives over a specific adversary or adversaries." An information operation is defined in the same directive as "actions taken to affect adversary information and information systems while defending one's own information and information systems." These "information systems" lie at the heart of the modernization effort of the US armed forces and ot
Yes... i mean, no... i mean, maybe. Let me try.
Why you wanna know?
Fuck systemd. Fuck Redhat. Fuck Soylent, too. Wait, scratch the last one.
Aren't we really just talking about coding for patterns of anomalies? We know how to code for patterns, we know how to code for anomalies. Isn't it a matter of processing huge data sets and looking for patterns that have not been recorded before? Of course, you could argue that whether or not the pattern is relevant is the big problem, but curiosity is not necessarily about relevance.
neorush
Every time you program curiosity, a lolcat dies. "What happens if" is a very dangerous thing to teach to amoral beings.
AI researcher Jurgen Schmidhuber says his main scientific ambition 'is to build an optimal scientist, then retire.'
Should we not create computers with at least near human intelegence before we try to give them curiosity? It seems pretty useless to me to give a computer curiosity in the hope that humans might learn something when, at its current state, the computer could not decipher the information it is curious about. I guess we could still look to the future, but why waste this time on such things when we could be programming for the iPad?
I think that the approach commonly taken to achieve some form of AI (curiosity as an example) through programming methods may be a flawed way of going about it. We probably should go about the problem in a similar way to how biological systems developed various aspects of AI. That is, build a system that has some basic rules for its operation that tends to form a system where curiosity and intelligence in general is an emergent property rather than one that is strictly programmed into the system. Take an existing system with some degree of "creativity" inherent in it and model our own technology to at first, mimic the natural system and over time, we tweak the system to suit our purposes as It is extremely difficult to build such systems from scratch.
Sigs are too short to say anything truly profound so read the above post instead.
I'm glad to see serious researchers are at work figuring this stuff out, now that they've got a working definition of intelligence and have figured out how to make intelligent programs.
When I was a fledgling programmer in the 80s I worked on some financial AI programs for a bank with some very smart people with lots of letters after their names and programming artificial curiosity was assigned to me. After some thought and a lot of dead ends I managed to program a reasonable (for our needs) facsimile of curiosity by assigning weights to the various pathways the program was evaluating and making those weights tend towards 0 (curiosity satisfied) or 1 (Curious) without ever reaching the final values. By having the program modify the weights and make decisions on which paths to follow based on those weights the program acted as if it was curious and came up with several interesting results that were completly unexpected.
Curiosity Sphere: Who are you? What is that? Oh! What's that? What's THAT? What is THAT?
Curiosity Sphere: Ooooh! That thing has numbers on it.
Curiosity Sphere: Hey! Look at that thing. No, that other thing!
Curiosity Sphere: Where are we going? Are you coming back? What's that noise? Is that a gun? Do you smell something burning? Ooooohh... what's in heeeere?
Curiosity Sphere: Oh hey! You're the lady from the test. Hi!
Azimov would be proud
Schmidhuber has interesting claims, like about his Goedel machine, an algorithm that makes provably globally optimal self-modifications.
But he never seems to get around to actually writing the code, or even non-vague pseudocode to implement these algorithms to show how they actually work and that they actually work. I guess it's just an "implementation issue". Ah, the chorus of the pure theorist...
Information theory is life. The rest is just the KL divergence.
A minority of AI researchers have tackled the problem on and off, and even built some small-scale models of curious agents. One of the classic precursors is Doug Lenat's 1977 system Automated Mathematician, which shifted from the idea of using AI to prove theorems, to instead looking for theorems that would be interesting if they were true (it didn't actually prove them; it was an interesting-conjecture generator). Essentially a model of mathematical curiosity.
Some interesting more recent work is a 2001 thesis that modeled curiosity as a social phenomenon in societies of agents, where agents try to find things that are: 1) new enough to interest its fellow agents; yet not 2) so new that they were incomprehensible in its cultural context.
(I'm an AI researcher, though not precisely in this area.)
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
All this singularity stuff seems pretty unsubstantiated to me. Programming curiosity? Maybe. Programming creativity? Well, that's the problem. If you want your program to be more creative than you and thus discover things you fundamentally wouldn't, that means you want it to be smarter than you. My theory, although I can't prove it, is that a program can never be smarter than its programmer.
I offer as evidence: it seems to me one description of a program is a way to store intelligence. The programmer is effectively imparting his intelligence to the program. So the 50s come around and we make this thing called fortran. Great, we've (basically permanently) stored our intelligence in that program. We can now utilize it to do much more intelligent things. We could even write a higher level compiler in it. So today, most compilers are written in C, even if they target the same level of the stack as C. Once you've saved the intelligence, you've got it for good, you don't have to worry about it anymore.
When you start thinking about things this way, you begin to realize that the only way to make programs more intelligent is to endow them with some of your own. This is why I doubt they can ever be more intelligent than the programmer making them. You can't give them what you don't have.
FTA:
Oh really? Is that all you need?
Look carefully at number 3. How do you define what is improvement, and thus to be rewarded? We're feeling overall pretty good about our economy right now, and I think Americans were in 1928 also, weren't they?
That brings me to another point, experience. Has it occurred to anyone that maybe we're going about AI the wrong way? A human, when viewed as a computer, needs about 15 years of 24/7 training and experience (minus sleep, ..... maybe) to become viable, much less competitive. Anyone else think that we evolved this way because, !shock!, evolution figured out that that's the most efficient way to do it. Even if we really improved our robotics/nano skills, it could be the case that the only way to get something to the level of a human is lots and lots of experience. In which case, would robotic replacements even be economical, given that humans can be produced to a par level with unskilled labor?
One more thing:
Maybe, but keep in mind that ants didn't create humans.
Personally, I'd rather research efforts be directed toward all the JS in /. 2.0 not sucking.
the entire Universe, including everyone in it, is in principle computable by a completely deterministic computer program
.. as long as you start with a piece of fairy cake.
Let me see....
if touch == [ouch] {
@"damn it";
}
else {
@"oh mama";
}
If curiosity is a behavior, then it should be pretty straight forward. In fact, depending on how you define "curiosity", then there are already many examples of programs that are curious. Google or Bing or any web crawler is definitely "curious". A satnav that searches for the best route from point A to point B could be "curious"...
A robot is only as smart as its smartest programmer.
And he ultimately addresses the possibility that the entire Universe, including everyone in it, is in principle computable by a completely deterministic computer program.
The problem that is often ignored with this and similar claims is the problem of observability as illustrated in areas such as quantum physics, and even economics.
You cannot calculate the behavior of a black box without opening it. If opening it alters the state of its contents, then it may even be impossible. And if you have no means of observation to begin with, then it is downright impossible. Before you can claim you can calculate the next moment in time, you must be able to claim you have observed and know all the variables within the system of interest.
Interesting article. It's funny but, all along, I always assumed that curiosity was a part of the definition of intelligence. If it exists in humans and animals, then that's all the evidence that we need in order to know that it can be programmed into a machine. The truth is that an intelligent program must learn and learning is impossible without curiosity. Here's why. If you look at knowledge as a big tree with many branches and leaves, learning consists of adding new branches (big and small) and leaves to the tree. The sub-program that goes around the tree adding new leaves and branches while pruning others as needed is none other than the curiosity module or algorithm. Just a thought.
Rebel Science News
Then why not hatred and brutality? Soon we will have AI bots trying to wipe out all humans of the wrong skin color, which at first will be specified by some human. But eventually the bots will figure out that "shiny aluminum" is the only non-wrong skin color, and set off to wipe out ALL humans. Bleccch.
A completely deterministic program creating the universe and all in it would be meaningless unless some being could use it like a TV show. Perhaps a universe that is not completely deterministic might be a better product with more uses to a supreme being. Perhaps that is why the classic debate between mankind having no free will among its members verses those that believe it is all about free will leaves both sides wanting. Individuals with limited free will may match the actions of other things in this universe.
As I understand it, everything we learn and do can ultimately be condensed into one thing. Survival. Think about it, we are alive today because the core tenant of our existence hasn't been broken yet. We, as a species continue to survive. Different behaviors do nothing but aid or take a different path to maintaining this goal. Perhaps curiosity is nothing but an attempt to make our survival more efficient. Perhaps it's a luxury only suited for higher level organism. Who knows.
My advice? Just create many iterations of AI and pit them against each other. Lather, rinse, repeat. In other words, just let nature take its course.
Life is not for the lazy.
Of course curiosity can be programmed. What are humans if not big, fleshy, biological machines of sorts? Granted we do not work like computers do, but the underlying processes are still structured and computational--if the brain were chaotic it wouldn't work.
Of course, some people will handwave with "the soul" or silly objections by Searle...
And then when I read about the current state of the education system, I get just a bit worried...
"Einstein argued that [...] God is not capricious or arbitrary. No such faith comforts the software engineer." ~ Brooks
If you have to explain why a joke is funny, it isn't funny.
From TFA:
I fear that no joke will ever be novel or funny again.
I don't therefore I'm not.
'is to build an optimal scientist, then retire.'
Build a what?
I suspect it's already retirement time. No offense.
SB
It's old. The more humans I meet, the more I like my cats. At least they are honest.
10: CALL Monolith
20: PRAISE Monolith
30: GOTO 50
40: Understand Monolith
50: Satiated = CALL Curiosity
60: IF Satiated > Infinity GOTO 40
70: ELSE GOTO 50
It must have been something you assimilated. . . .
This is like the Drake equation. More boring bullshit from theorists that has nothing to do with reality. Show me something in motion, show me a computer processing something, and then we'll talk. So you can describe the "algorithmic principle" of curiosity? So what? I could probably describe myself taking a shit and barfing at the same time algorithmically if I wanted. These people are so far up in their ivory tower they forget that at the end of the day you have to give instructions to a microprocessor. Stop wasting my time and Get Real.
The set up was so perfect, someone had to say it!
Note to the deprogrammed - this is not a pro-anything joke, just an anti-fox-news joke.
"his main scientific ambition 'is to build an optimal scientist, then retire.' The Cognitive Robotics professor has worked ..."
Woo hoo! The Robotic Cognitive Professor worked! Oh, wait...
Our greatest gift to god will be creating a mind that can believe in him.
Why is it that AI research is always mislead by it's name? Namely they are
too focused on the intelligence aspect of a programmed mind that they
completely fail to recognize it's subjective emotions and motivation that they
should be focusing on.
What is a soul? It's that part of a mind that is able to make a choice. It's
the part of the mind that isn't logical. It's the part of the mind that can
judge something as good and bad. It's has beliefs. It can be informed by
reasoning but it can still choose mysticism over reason. It wants and it
can choose. Behind every mind there is motivation. Sure it's still a
program but it's the one that matters.
Just because you can give an ant mind super intelligence doesn't make it any
less of an ant. It understands more but it is still an ant and wants what
ants want. The reverse of this is a complex soul that can't make sense
of the world around him; we tend to call this autism. Maybe the former is
autism as well.
Most people should be able to agree that psychologists know enough that they
can actually drive a sane person to insanity. Therefore we should also be
able to drive an artificial mind to insanity.
It is not enough to recognize beauty; you have to feel it, you have know and
believe that it is good and right.
I've devised an algorithm that tells me with 100% certainty that this guy's ego is way too far up his ass.
Isn't Curiosity really 'Intuition'? A part of the brain who's function it is to compress all the data from hearing and eyes to a lower 'bitrate' and to fill in the blanks between sensors by pattern-matching.
Imo, what you need to write to simulate 'Curiosity' is pattern-matching software: taking input from many sources (sensors) and try to find a match between them.
And he ultimately addresses the possibility that the entire Universe, including everyone in it, is in principle computable by a completely deterministic computer program."
After which he took another long drag on his joint and said, "It's like our whole universe is inside a single election in a larger universe, you dig? Hey, pass those corn chips over, dude! Now where was I? What? Ah, never mind. Put on Conan. It's his last show."
I'm an AI researcher, though not precisely in this area
What's that like? I'm curious.
These posts express my own personal views, not those of my employer
Something with a soul (eg. human) may not be curious about anything.
And who is to say something that is soul-less can not be curious?
I guess it's just an "implementation issue". Ah, the chorus of the pure theorist...
Here's a thought (I haven't decided whether I agree with it):
Would it make sense to divide the work of creating AI into the Getting Ideas part and the Turning Ideas Into Code part? The idea being that you can let people who are good at one part do that part, and let people who are good at the other do the other part. (That goes back to Adam Smith, division of labour.)
Suppose a physicist establishes a theory about the reflection of light which (among other things) can be used to make more efficient solar cells. Yet he doesn't make any solar cells. Would he be met with the same attitude? Is that "just an implementation issue" too?
Or say an astronomer discovers a new celestial object. Do people poo-poo him because he hasn't gone there? :P (Okay, this one is stretching it...)
I'm not saying your attitude is wrong. I'm wondering, and I hope some of you smart slashdotters can help me figure it out, why computer science researchers get met with the "You haven't turned it into a prototype (or product!) yet, come back when you have."
I think it's because what CS research creates is very close to what Software Engineers (/programmers) create: algorithms. Moreover, the algorithms created by research always solve a particular problem, because that's what algorithms do. In some sense, all CS research is applied, but since it's still research it's not applied enough---it's not a product.
Contrast that with what most scientific fields do: "prove" declarative claims about how the world works (quantum mechanics, planetary motion, natural selection, thermoelectric effect, ...). An algorithm relates to a declarative claim (about its correctness), but it has an imperative "(you can) Do this: ... (and only this)" bit attached to it that most other fields don't have.
I think I can find an exception in the field of medicine---much medical research is into the safety and effectiveness of "algorithms" for treating particular diseases (input chemical X). But they test finished "implementations"---you can't really figure out what chemical X does without inputting it and seeing what happens. Not yet, anyways---humans are big and complex, and to the best of my knowledge there isn't a good, complete model of how they work; that's unlike CS, where we can read a program and reason about what it does without running it.
I think it's the similarity between research output and engineering output that makes many people want researchers to do the engineers' job.
Would it really be a good thing if they did?
(That's not to say we should have a low bar for evidence for "truth", such as correctness or (for more fuzzy domains) effectiveness and usefulness.)
Sticking a picture of a smile on a piece of paper does not mean that the piece of paper is happy.
Curiosity is a feeling. Programming a computer to work in some particular way does not mean that the computer is curious.
Let's program our computers to date girls. This explains why there are so many virgins on slashdot. Sometimes a little investigation can be a good thing.
The theory is ... cute at best, but unworkable.
Over the eons simple 2d images and more recently 3d images & holograms have been accepted as reasonable representations. Would a complete 3 dimensional representation be acceptable ?
Would not leveraging the 7+ additional string/M-theory folded dimensions not be within our universe? Would suggest that were are already doing so with 1984+ quantum teleportation, susequent quantum cryptography, etc ...
Presume if not computing digitally with bits and vast series of same but instead NKS cellular automata relations there, that far less would required to represent all that "exists" and is experienced, known & interacted with in the our 3 extant dimensions.
As there would in essence be no distance between lattice of the subset of points leveraged to represent our perceived universe across the other dimensions Wolfram-Alpha-esk computation there should be vastly faster than that experienced in the slower paced space time in our lesser reality.
If you have to explain why a joke is funny, it isn't funny
Yet another AI fairytale...
Maybe next they should study how to program bi-curiosity. God knows most software is pretty gay these days, and I'd like to know why.
The Advanced Concepts Team of the European Space Agency is running
a project closely related with this. It is called "Curiosity Cloning".
Have a look at:
http://www.esa.int/gsp/ACT/bng/op/CuriosityCloning/curiositycloning.html
Part of it is that CS theory tends to prove things that, while mathematically true, in practice often don't matter. For example, it's a running joke in AI that everything interesting is NP-complete. So we don't care about NP-completeness. What we want to avoid is AI-completeness: problems that, if you could solve them, would imply that you had Full Human-Level Intelligence. We want to solve bits of intelligence without having to solve all of it, but if it's NP-complete, who cares, because everything is. In fact, if you can reduce an AI problem to SAT, everyone's happen, because SAT is famously fast to solve (it's NP-complete, but almost always fast in practice).
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Hah, that's "everyone's happy", of course.
Incidentally, this was the work that started the "reduce to SAT, because it's NP-complete and, btw, fast" trend.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
This robot will turn into a navel gazing philospher trying to figure out the meaning of his programming. Recursiveness ftw
'Curiosity is the desire to create or discover more non-random, non-arbitrary, regular data that is novel and surprising...'
... and that curiosity overlapped, like so many ill and broken human brains, into the realm of murder... Skynet!
It is not curiosity that is hard to achieve - brute forcing is equivalent to curiosity upped to 11. It is filtration of resulting data that is hard.
And it is hard not only because there is so much that can be said about anything, but because what is relevant depends on what is being looked for. Which is a bit circular, like in self referential.
In other words, what is the metric for curiosity?
"Blah blah blah." - [citation needed]
I've fallen off your lawn, and I can't get up.
I once extensively thought about that subject. The point of curiosity. How it works.
We humans have an internal model of reality. In fact it’s the only reality that we have. Since we can’t prove the existence of anything outside our minds.
This model is a set of associations. Which, just as the universe, are defined in a relative way.
This means, that we can’t handle anything that is not related to something else in some way. Even if it’s just a basic feeling.
But since we start out with nothing, and must be able to handle things, to survive, we have a built-in trigger. Whenever we find something that we can’t handle, we start trying to fit it into our model. Therefore we play with it, directed to whatever tiny feeling seems remind us of something. If there is none, there also is no real method in our playing. Just like with games. (Which are in fact training for such situations.)
We then either find a way that it fits... or we start to ignore it, as if it would not exist. Because else, our mind — unable to handle reality anymore — would completely break down. (Which never happens, because there is always schizophrenia — living in a imaginary world — as a last resort to protect you.)
I think if we implement it this way, it’s actually not that hard to simulate curiosity.
Create a mind that stores things in associations, and has a inner model, that it uses to do normal things a life form would do (collect and transform resources, grow/expand and reproduce, etc).
And then just build routine that tries to uphold a reality where everything fits and makes sense.
Meaning that things that can’t be associated, will be manipulated, until they are, or until a “I give up” neuron gets past its firing point (where that thing would be ignored).
Of course it’s easier when the mind has the ability to internally simulate never seen combinations of things, based on the known associations. (Imagination.) Because then that mind could make up fantasy worlds (that make sense in the inner model) about those things (= religion), and handle them anyway. (Or become schizophrenic. ^^)
And there you’d have it: Curiosity. (Which could even kill the simulated cat. ;)
Any sufficiently advanced intelligence is indistinguishable from stupidity.
And he ultimately addresses the possibility that the entire Universe, including everyone in it, is in principle computable by a completely deterministic computer program.
He should try to do some actual computational physics/chemistry. The amount of processor power you need to simulate only tiny structures is so enormous that he'd be thrown back to reality really quickly.
If Pandora's box is destined to be opened, *I* want to be the one to open it.
skynet: would humans would survive a nuclear winter? let's find out.
> the entire Universe, including everyone in it, is in principle computable by a completely deterministic computer program
Here come the Matrix analogies for the rest of the posts here on end....!
Any problem is solvable given an infinite number of (live) cats
Donte Alistair Anderson Roberts - hi son!
Karma: Chameleon
if ( object.stateIsPredictable() )
then ( object.process();)
else ( cout "Interesting!" )
Knowledge is power. Knowledge shared is power lost.
It's not so much the lack of code but the
or even non-vague pseudocode
And it's not so much the "where's your finished product?" but "um, so what *is* the algorithm?"
Information theory is life. The rest is just the KL divergence.
I can write a program that prints itself. I can write program A that prints itself, plus does other stuff. The program it prints (B) can also print itself, plus do other stuff (since B is the same as A). The program contains the code to print itself, plus the code to do other stuff, with no problem.
Simulating the universe from within itself is quite possible. Obviously, some bits within the simulation represent more than the same number of bits in the real universe, but fortunately, lossless compression is possible.
says you're wrong. In particular, the Heisenberg principle.
#!/bin/sh
cat &
$STATE = $1
if [ $STATE -eq "curiosity" ] ; then
kill $!
fi
Every time some subject comes up in conversation which makes me twitch or sweat or want to pull away, THAT indicates a piece of gunk.
This piece of gunk is called an "engram", and it triggers post-traumatic stress. Both psychiatry and Dianetics have their own versions of a process called abreaction therapy to "play back" these engrams in such a way as to reduce the influence that they have on the mind.
The Ego is, I think, a foreign installment designed exactly to keep us from performing that self-examination. With the Ego in place and strong, there is no hope of breaking out of the cage of automatic behavior.
Scientologists would say the same about the reactive mind.
If computer perceives things as "exciting" and is curious about them, it will necessarily perceive other things that are "boring" and will have no interest in them, probably refusing to do them as non-necessary, and generally moaning.
Also, guess what category is "running your Windows".
I believe its not that mystical or complex. Its a simple desire to find out more about a phenomenon that has already influenced us. We already had some inputs in our brain about it, but we don't fully understand it. Its just the desire to rationalize, to "complete the circle". Its a very basic mind process. It can be simulated extremely easy in my opinion.
Here is a short version of his incredible talk, only 10 minutes:
youtube.com/watch?v=Ipomu0MLFaI
"Wow, man, that was fucking AWESOME! Plug me into that thing again, will you?"
Free Martian Whores!
At the atomic level there's a lot of randomness.
Can we be sure? What seems random may not in fact be truly random. The flip of a coin is considered random, but if you could account for all the variables with enough precision; angle of the coin, angle of the thumb, force of the flip, distance to the floor, etc, you could likely predict each and every toss.
Exactly. As a poetic friend of mine puts it, "There's no such thing as random, even dice have a past."
Rather than being random it could be that it's just more complex than we know, or that we can't determine the variables with enough accuracy. What is the exact value of PI?
The exact value of pi is actually that irrational number that people like to waste supercomputer time on calculating to millions of decimal places, but a calculation of pi to such extreme precision is useless for any practical purpose. You're on the right track but you're asking the wrong question. The catch is that circles are purely a mathematical abstraction, not a real thing. In the real physical world, circles don't exist. The closest we could come is a bajillion-sided polygon where each side is one Planck Unit long.
Knowledge != Intelligence
I assume the ambition here is to duplicate the curiosity of the best of humans on machines.
If such algorithm can be done, it must be self-sustaining, an infinite loop, but non-repetitive, with a self-determining ability to recognize and incorporate ever newer variables, and as complexity increases, the consciousness to generate its own algorithms for better organization of its growing self. Create another thinking head, in short. That's one hurdle.
Multiply that by the number of heads in the world to account for variation or selectivity in curiosity, due to external factors beyond individual self-government. Task, create the web of humanity. That's another hurdle.
But perhaps the hardest - for the grand algorithm that answers to the above, what energy source is there to sustain open-ended growth? And say this sci-fi becomes reality, what is the advantage we have against the Frankenstein? It will not want to DIE easily. No life does.
> possibility that the entire Universe, including everyone in it, is in principle computable by a completely deterministic computer program."
Doesn't it contradict the Gödel's incompleteness theorems?
>>You can't calculate the universe from within itself any more than a VMWare can run a machine faster than the host processor.
What if the universe was made up by nothing but an (intelligent) bowling ball traveling through otherwise empty space? It's pretty easy to figure out where you'll be in 10 years.
That said, the whole computability argument is bunk, since the Halting Problem tells us that there are some things that are just not computable. If we have a universe consisting of nothing but a pool table and some (intelligent) billiard balls, we can predict the future to any arbitrary degree of accuracy. But if these same billiard balls are programmed to never go where they're supposed to go, then the result is indeterminate, as in the halting program.
This is actually the reason why I believe in Free Will. Determinism is provably impossible.
...or at least the computer that could build it.
Can't believe I'm the first to say that...
There's a robotic scientist called ADAM that investigates yeast genetics (http://www.aber.ac.uk/~dcswww/Research/bio/robotsci/). There was a pretty cool paper in Computer a few months ago. The robot actively tried to devise new theories and produce experiments (it's hooked up to a bunch of yeast-genetics-investigatory stuff) to investigate those theories. As I remember, most of the theories turned out to be true and were pretty novel (function of various genes). The researchers double checked several (or all?) of them.
Then he should perhaps take a trip inside himself. This article and his webpage is full of self-self-congratulatory bullshit.
-- .
Submitted as a comment to H+ Magazine by Fred (not verified) on Sat, 01/30/2010 - 07:44:
His team now has the best method for connected handwriting recognition. It uses a self-learning recurrent neural network that won several handwriting competitions at ICDAR 2009. But when you study papers on this at http://www.idsia.ch/~juergen/rnn.html you'll find the recurrent network is not trained by those super-universal learning algorithms, but by "greedy" techniques such as gradient descent. Even in his own lab, being practical is sometimes preferred over being theoretically optimal :-)