"My name is Professor Hugo de Garis. I'm the head of a research group which designs and builds "artificial brains", a field that I have largely pioneered. But I'm more than just a researcher and scientist - I'm also a social critic with a political and ethical conscience. I am very worried that in the second half of our new century, the consequences of the kind of work that I do may have such a negative impact upon humanity that I truly fear for the future."

http://www.cs.usu.edu/~degaris/artilectwar2.html

So, Esoteric

What do you think about the professor's brains?

Anyone seen one work? OR is it one of those fictionalized documentaries? My BYU friends in the EE department think, he is smoking or might have done too many loops on the roller coaster....:D

From what i have been reading on the subject artilects seems plausible. The terrans/cosmist stuff is pretty much pure fantasy, albeit entertaining.

heres another site on artilects.

http://www.artilect.org/

Definitely interesting. But all these pundits forget that one just can not linearly extrapolate how a million times powerful AI would behave. They just try to extrapolate their own behaviour if they become Gods and then get scared only if their co-worker becomes first.

As life goes....do you know that there is a whole different plant kingdom out there that are gradually getting smarter thanks to our genetic meddling? We should worry more about that.

And perhaps there are a lot of natural hyper AI computers out there the size of pulsars....just imagine...

From Ray Kurzweil site:

"Once computers achieve a level of intelligence comparable to that of humans, they will necessarily soar past it. For example, if I learn French, I can't readily download that learning to you. The reason is that for us, learning involves successions of stunningly complex patterns of interconnections among brain cells (neurons) and among the concentrations of biochemicals, known as neurotransmitters, that enable impulses to travel from neuron to neuron. We have no way of quickly downloading these patterns. But quick downloading will allow our nonbiological creations to share immediately what they learn with billions of other machines. Ultimately, nonbiological entities will master not only the sum total of their own knowledge but all of ours as well."


This assumes that the complex process of AI is a simple process of downloading information to an artificial brain even though such does not work for we humans. If that is so easy, how come K-Mart is going bankrupt? All they need to do is download the software and all their supply chain logistics will hum like Wal-Mart. What if the software is too complex to emulate the hardware? What if you need a real hardware to develop the AI. So, the downloading the next upgrade is more like building the next 3-D cube.

What burns me is that these physicists who spend better part of their life learning material science and physical laws suddenly become experts in information science while tinkering with hardware. When you ask them, they say - oh! just a little software, I can not do it, but my 12 year old son could do it. We can not find God, but we surely can build it.

Well, we have been building God since Man came out of the cave. And when we make a Monkey translate Chinese to English, then we will have the software to do it. At least, we will learn the mechanism....

Here is our fictional take on the artilect/ai situation;
http://www.orionsarm.com/sophontology/archailects.html

who Knows if this will ever come about? What will be the status of humanity in such a universe?
Is it, or will it soon be, possible to detect the infra-red emissions from an alien artilect thousands of light years distant?

don't worry- the software for artificial intelligence will emerge, sooner or later, one way or another; will we even recognise it as such if it is radically unlike human consciousness? And as Kurzweil points out, another thing to worry about is how friendly will it be...

One of the first things this guy appeals to is Moore's Law... most of the stuff he mentions in this article is neither cutting edge nor particularly clever.

Maybe that's a little unfair but, getting past his weird advertisement of his beliefs about gender relations, I find that he seems strangely non-technical in a lot of fields he draws from.

He states that "If Moore's Law continues unstopped until 2020 or thereabouts, it will be possible to store one bit of information (a zero or a one, a "0" or a "1") on a single atom. " I hesitate to make this point, but Moore's Law is not restricted by time... either we'll find out how to do that little trick (storing information on a single atom, which in terms of current photolithographic techniques is not very likely) or we won't. No law of even progression will govern this.

I won't say much about Avogadro's number other than the fact that it's the number of atoms/molecules in a mole of an element/compound, not the number of atoms in "an object of human scale".

In terms of physics, he says "In physics, the concept of "entropy" is used to measure how disordered a physical system is. For example, ice has a lower entropy than water, because it is more ordered, less chaotic."

Now since the heat-death of the universe - the ultimate expression of entropy, I suppose - will probably be extremely orderly, this description of entropy is rather backwards. However, this is not the most serious problem here...

He begins equating information to energy, in the belief that a system with no information loss will also generate no heat. This seems at odds with the physics of the situation as I understand it...

In general, there's a lot of "Moore's law will save us" assertions -
"All this information could be dumped into a "hyper-computer", that Moore's Law will make possible" - apparently ignoring the fact that Moore's law is not a law, but just an observation of a trend. His descriptions of artificial embryology betray no knowledge of biology, and his maunderings about "gigadeath" and "artilect gods" are no different from plenty of other speculative fiction.

Any reasonably smart high school student could have written this document, the misconceptions about various scientific fields are certainly in line with the regular public misconceptions. His descriptions seem imprecise and at times impossibly hopeful, and he still has not made any attempt to explain how making a computer bigger will make it intelligent; he is simply following the old science fiction saw that any computer that is big enough will become sapient, probably by accident while no one is looking.

After this he goes off into a long wander about Cosmism, the reasons why everything will be the way it will be, and a bunch of tiresome ideological claptrap which I didn't bother to do more than skim through.

From the quality of his composition it's hard to tell whether this guy is even a real university professor. I sincerely doubt that this document will provide any insight for anyone into artificial intelligence.

One of the first things this guy appeals to is Moore's Law... I don't think he's legit in AI or even computers.

After looking into it for 10 minutes or so, he seems legit. He's a prof at a university. I've seen nothign to say it works, or is even actually promising, but I'm pretty sure he's somewhat legit with computers and AI. His appeal to Moore's law could easily be viewed as an attempt to relate to readers. Most nerds have herd of it, so he captures interest that way? I dunno. Quite interesting I suppose. I'll have look into it more.

Sorry Wes, I edited quite a bit.

from the linked text above:

There are people (for example, Sir Roger Penrose, of black hole theory fame, and arch rival of the wheel-chaired British cosmologist Stephen Hawking) who claim that there is more to producing an intelligent conscious machine than just massive computational abilities. I am open to this objection. Perhaps such critics are right. If so, then their objections do not change my basic thesis much, since I feel that it is only a question of time before science understands how nature builds us, i.e. before science understands the "embryogenic" process, used in building an embryo and then a baby, consisting of trillions of cells, from a single fertilized egg cell.

I think that sums it up. I'm with Penrose on this one, and this dude's baby may be programmable, but until other avenues of research offer up key info, I doubt there's little worry of a machine becoming self-aware... unless of course it could happen by accident.

Wes, that's crap. That's the CS version of waiting for the Rapture.

Sorry Wes, I edited quite a bit.

You suck. Hehe. Not a problem. I still stand by my comment! ;) Oh, and I think I was much more succinct in my assessment as well, so :p

:D

Wes, that's crap. That's the CS version of waiting for the Rapture.

What exactly is crap? Penrose's point?

You mean the part about it happening by accident? Oh man yeah that's a seemingly remote possibility for sure. I didn't mean to present it as otherwise.

No, waiting for another field to provide an answer to the AI problem. Why bother even studying it in the first place?

No, waiting for another field to provide an answer to the AI problem. Why bother even studying it in the first place?

Oh, well that's silly. You're serious?

Okay, assuming you are serious (pardon if that was obnoxious), I'll offer the following simple notions.

There is no way to know if Penrose is correct. (like I said, I think he is though) If you don't think he is, you'll pursue AI via a computational route. A number of positive things remain even if penrose is right:

you study nueral networks and get to know the mechanics.
who knows what stuff you can learn from that


Hmm..

how about this. since you don't know how consciousness works, you can't say what is the right route to figure it out. so all routes that seem reasonable shoudl be pursued, especially if there are side benefits, like those offered by current AI research. Better computer games, stock market prediction and all kind of other AI apps that don't require a conscious entity. could be that the conscious entity could be comprised of a myriad of these sub systems... so if you don't pursue this route you still couldn't do it even if other areas of science offer up the fundamental conditions for conscoiusness.

oh and "waiting for the answer from somewhere else" is the silly part. you don't wait, you look as hard as you can in your area. maybe you don't need the other answers, maybe you do, but you'll never know if you just sit around waiting. i didn't intend at all to imply "sitting around and waiting". I just mean that I think that ultimately you won't have a choice if you're seriously trying to make a conscious machine. you still should research all you can in the mean time.

oh and further, maybe the scope of AI research is too narrow eh? rather, it's possible that for it to actually result in intelligence, the scope of the research may have to expand to solve the problem.

Of course I'm serious. You don't devote your life to studying a problem, and then add a footnote to your efforts that says

1. This part is somebody else's problem.

when you're referring to the fundamental pursuit of your study. Making computers really really large and complicated is just clouding the issue; the real point of the study of AI is to try to develop an intelligent/sapient computer, or come to the conclusion that it cannot be done for a concrete reason, be that quantum effects or whatever. That doesn't mean that you can just cite an apparent problem as a reason to sit on your backside and wait for someone else to fix it.

The nature of intelligence is such an unknown that it's not really possible for us to just say "Well it's obviously quantum" because we don't know. To do so is to effectively appeal to an unknown force.

Of course I'm serious. You don't devote your life to studying a problem, and then add a footnote to your efforts that says

1. This part is somebody else's problem.

when you're referring to the fundamental pursuit of your study. Making computers really really large and complicated is just clouding the issue; the real point of the study of AI is to try to develop an intelligent/sapient computer, or come to the conclusion that it cannot be done for a concrete reason, be that quantum effects or whatever. That doesn't mean that you can just cite an apparent problem as a reason to sit on your backside and wait for someone else to fix it.

The nature of intelligence is such an unknown that it's not really possible for us to just say "Well it's obviously quantum" because we don't know. To do so is to effectively appeal to an unknown force.

Hehe.. I like your footnote.

Yeah I agree wholly, as you can likely see from my preceding comment. Note that I didn't imply that one shouldn't pursue that AI research is pointless or that they should stop, even in the post to which you objected. I merely stated that I agree with Penrose. I'm pretty sure AI research will not produce conscious machines until they get a bone thrown to them from somewhere else. I don't think that means their efforts are wasted.

I'm pretty sure AI research will not produce conscious machines until they get a bone thrown to them from somewhere else. I don't think that means their efforts are wasted.

It is not meant to. The hope is a brute force machine that runs on a Bayesian or similar network will hopefully, suddenly become self aware. If that does not happen, we are back to square one. Another thought is that we can have a machine that improve and design another machine and so on, until several generation down the road, the machine becomes self aware.

I think such is possible when machines start interacting with humans for a prolonged period of time with some serious adaptive algorithms. Like that Robin Williams movie....

It is not meant to. The hope is a brute force machine that runs on a Bayesian or similar network will hopefully, suddenly become self aware. If that does not happen, we are back to square one. Another thought is that we can have a machine that improve and design another machine and so on, until several generation down the road, the machine becomes self aware.

I think such is possible when machines start interacting with humans for a prolonged period of time with some serious adaptive algorithms. Like that Robin Williams movie....

Of course Penrose could be wrong. I'm sure he's aware of it. Have you read at least skimmed "the emporer's new mind"? I read most of it a while back. Seems like a pretty strong case.

Regardless, I think that when you're tryign to study something about which little is really known, any rational approach should be taken if you have the resources to throw at it.. so I think the CS approach to AI is a good route, but again.. I think it probably won't work because of how I think consciousness works, which I've been tryign to figure out how to explain since I've been here. :) Of course, it's most likley that I'm wrong.. but that doesn't mean that I think I am. ;)

First of all, I am talking from experience. That does not mean, I am an expert - even though, I get paid as one. The experience comes from two areas. One was due to my son's migraine and resulting activities with a series of top doctors, neurologists, research, analysis of the brain scans, EEG, heavy duty discussions with experts to the point that I now know as much the current knowledge of how the brain works as any expert out there and perhaps more.

The second connection comes from my work in advanced automation in complex chemical manufacturing and robotics, vision systems and emulating certain human processes for aerospace - which required to write adaptive algorithms such as weather change and other functions that is partly based on a precanned if...then...else...data and partly fuzzy self learning decision matrix.

When I try to merge the two, what looks like parallel paths - I see a convergence somewhere way out there. I have faith, but as I learn more new areas such as Cellular automata, cognitive engineering, fractal science - the merge point still eludes me. I am pretty sure it can be done, but not the way what you read because they are each talking from a single vantage point. It is a combination of technologies and science that will get us there. As you said, it will come from unexpected sources....

Asimov's Three Laws of Robotics put to rest any fears I had of automatons taking over the human universe. Make the laws a part of their reality that cannot be changed, for example, we humans have to breath every second or two--there's absolutely no way to get around it. No one's even tried. It should be the same thing for robots, but with the 3 Laws substituted for breathing. They have to remember and practice the three laws, or die, just as a person dies when he or she stops breathing. They should be an inherent trait to artificial life, just as spines are inherent to vertebrates. No matter how intelligent we get, at the most we'll only be able to substitute breathing or substitute our spines, likewise, no matter how intelligent the robots get, they'll be unable to work past the 3 laws, or even desire to, for that matter.

As for the article, or book, there's just way too much conjecture. I don't think anyone can really predict what the world is going to be like twenty years from now, because no one twenty years ago succesfully did the same thing. I do agree with him when he says that robots are going to become heavily involved in politics, just as oil is now, and just as slaves were to the United States in its antebellum years. Should they be allowed to vote? Should they be allowed to make money? Are they really people? White Americans asked the same questions about Blacks, and it is easy to postulate that the same questions will be asked about robots. Hopefully these questions and disagreements will not lead to power struggles or civil wars. The American Civil War over slavery was not apocalyptic (in spite of how horrific it was...something like 1 in 30 Americans was a casualty in the War), but the next Civil War over Robot Suffrage, Robot citizenship, etcetera, may be exponentially worse, when nuclear weapons are added to the equation.

As for human breathing goes, the caveman could not stay more than a few minutes under water. He must have thought that man can never be under water or out in space. Since we can work around it, so can the future robots by redesigning those Asimov rules.

As for conjecture, we can predict a lot of stuff for the next 10 to 15 years because it takes that many years for a new idea to take shape. We can even predict to next 50 years for complex science to add value. There is not really any earth shattering inventions or discoveries that changes the planet overnight. Remember the cold fusion - which could have, but did not. It is the elbow grease that can produce stuff from carbon nanotubes or high density DVDs using Purple lasers or OLED based big screen TVs.

And do not forget, we may end up joining the caveman if a big asteroid hits us in the next 30 years, since we do not have any protection yet willing to spend our wad screwing, I mean going to Mars in 2030. :D

Even if Penrose is right (and I don't think he is) the creation of self conscious AI is not ruled out; if it is necessary to incorporate quantum effects into a conscious computer then this will be done-
what law says that a machine could not have microtubules in the design?
Obviously to make a human-like conscious entity every detail in the brain must be emulated in some way.
It might happen that other lines of research produce apparently conscious entities before that is possible... if a machine consistently passes the Turing test, then it will appear to be conscious, even if it is just a glorified search engine.
An apparently conscious entity is effectively identical with a real conscious entity, as no-one can prove the existance of consciousness in any entity outside of themselves anyway.

And I really don't think that anything like the Three Laws will be possible in practice...
a thinking machine can't also be a pre-programmed machine.
Animal instincts provide a set of nested imperatives, but the behavior of an animal following its instincts is very diffrent (it seems to me) to the behaviour of a car assembly robot following its program.

When you get to human level and beyond, instincts and preprogrammed behaviour will almost certainly be overruled by conscious thought.

"As for human breathing goes, the caveman could not stay more than a few minutes under water. He must have thought that man can never be under water or out in space. Since we can work around it, so can the future robots by redesigning those Asimov rules."

You're missing the point. We can't survive without breathing. We may be able to go underwater or into space, but we still have to breathe when we're there.

"As for conjecture, we can predict a lot of stuff for the next 10 to 15 years because it takes that many years for a new idea to take shape."

Alright. Give me someone who, fifteen years ago, accurately predicted the world of 2004. Then I'll believe you.

"When you get to human level and beyond, instincts and preprogrammed behaviour will almost certainly be overruled by conscious thought."

We still have to breathe, we still have to eat, we still have to drink. We wouldn't breathe if our instincts didn't tell us to. Robots wouldn't follow the 3 laws if their instincts didn't tell them to.

The three laws of robotics are a poorly thought out concept and probably have no algorithmic description sufficient to convey the spirit of the concept.

#1: A robot may not injure a human being or, through inaction, allow a human being to come to harm.

This is the killer... we can't even manage this ourselves. As a categorical imperative it is impossible to follow, since human beings are allowed to come to harm constantly as a result of our inaction.

The other two also represent significant logical problems...

You're missing the point. We can't survive without breathing. We may be able to go underwater or into space, but we still have to breathe when we're there.

But you can break the rule of not able to interact in space or under water. It is the result you are after not the laws. Get it?

Alright. Give me someone who, fifteen years ago, accurately predicted the world of 2004. Then I'll believe you.

I did. I spent sometime in 1983/84 China and predicted the growth just as it is happening now. I have posted this here before this thread began. I even predicted Russia's change and also difficulty in managing its economy.


We still have to breathe, we still have to eat, we still have to drink. We wouldn't breathe if our instincts didn't tell us to. Robots wouldn't follow the 3 laws if their instincts didn't tell them to.

In case of robots, the instinct is a line of code and can be removed. The robot will function just fine since they do the same today in the assembly line without such laws. They just do not have the higher functions today because we do not know how to program them. Besides, breathing is not an instinct, it is organic mechanics - just as an internal combustion engine that has to have air to operate. Now imagine, if we can program the car not to have an accident by cutting off its air supply.

But you can break the rule of not able to interact in space or under water. It is the result you are after not the laws. Get it?

No...


I did. I spent sometime in 1983/84 China and predicted the growth just as it is happening now. I have posted this here before this thread began. I even predicted Russia's change and also difficulty in managing its economy.


Forgive me for asking, but is there proof that the materials you mentioned are dated at the time you mentioned? Fork over a URL!


In case of robots, the instinct is a line of code and can be removed.

Then this programming should be far more inherent to what it means to be robotic. It should be deeper than just a "line of code." I'm just speculating, and for me to really go any further would mean talking about things that I do not have the credentials to fully understand. It should not be something that they would have the ability to change.

Failsafe machines might be useful as well. They could be just as intelligent as their robotic buddies, just as much of a soul, I guess, but would have a different purpose to their life. Everything does have a purpose--humans are supposed to have sex, computers are for the moment used to do complex or long calculations, etcetera. Future robots should have purposes as well. Some would have the central purpose of keeping other machines in line, if the 3 laws somehow failed. What better to fight fire than fire?

Or they just shouldn't be made any smarter than people. Let them do complex computations, but beyond that, nothing.


This is the killer... we can't even manage this ourselves

That's because it isn't instinct for us. We kill because we can. Instinct is prevalent throughout every form of life, why should it not exist for robots? Their primary instinct, just as ours is to procreate, should be the 3 laws.

Nono Pollux, not killing is easy. It's the "not allow a human being to come to harm through inaction" that is impossible to follow. Your inaction is killing hundreds right now. There would have to be a scope placed upon the robot's consideration of harm to others, which would make it just as easy to kill those people through indirect means as it is for you to let them die by doing nothing.

In any case, categorical imperatives like the three laws are open to interpretation... I seem to remember Asimov had a story where the "no harm through inaction" directive was removed from one robot, and it managed to interpret shooting people as "harming them through inaction" because it was fast enough to run and catch the bullet before it hit them... thus by not doing so it merely "failed to save them".

If the categorical imperative is not open to interpretation, then the definition of harm will have to be structured in some way that people are still permitted to harm themselves without intervention from the robot; certainly no sort of police or combat robot would ever be possible with the 3 laws, and people are already making remote combat machines.

Also, an intelligent thing's purpose is defined by the intelligent thing, even if it is only to follow orders... if the robot is really intelligent it will choose its own purpose.

Okay. So it's open to too much interpretation. The laws are too relative.

We could be exact then. I think it's possible. We use a machine to calculate every single possible way of killing a person through action or purposeful inaction. Then we implement these rules into a robot at or above average human IQ and test drive it for a few years to see what happens. To see if we covered everything.

Why are you afraid of a robot killing you anyway? Aren't you worried that Asimov's widely loved stories of robot slavery will be terrifying to the intelligent machines? Trying to remove their ability to defend themselves may make them more dangerous than if they had human freedom anyway.

I believe such imperitives as the 3 Laws will have to be introduced not as strict "Laws" like they are written, but more like fears, or OCD in humans. A robot will have an "urge" to never hurt another human. An urge so strong (yet for no reason) that he probably will never be able to break it. Notice the probably. Many humans overcome many fears because it helps in their life. A robot that killed a human (or failed to save a human) will probably be destroyed. This doesn't seem to me to really help in robot survival, thus the robots (AI) would never have a reason to try and break the phobias.

-AntonK

Another post.... just because I don't wanna stick on 666 posts.

I've mentioned Hugo de Garis on these pages before; he has some very interesting ideas...
he coined the terms Picotech and Femtotech, for example, which we use liberally on our website
http://www.orionsarm.com/main.html
but I am not aware that he has actually achieved anything concrete...
even the concept of picotech is somewhat controvertial, but it is certainly fun.

Their primary instinct, just as ours is to procreate, should be the 3 laws.

Is it? tell that to the large gay community? and find out how many complete the law in their life time. If we intelligent beings can break laws, so can any AI. Remember the German cannibal? If nature could not hardwire any abstract laws to humans, I am sure we would not be able to - simply because we are part of that nature. May be once we evolve out of our natural boundaries - then that is where the AIs come in.

Is it? tell that to the large gay community? and find out how many complete the law in their life time.

It's an urge that doesn't guarantee completion. While homosexuality is still somewhat a mystery, I believe it's simply a different manifestation of the craven sexual desires within us all. We have an urge, but it doesn't matter that we complete the reason for the urge (at least on an individual level). I'm not talking about completion, just the urge itself. Just because I'm a heterosexual doesn't mean that I'm definitely going to have kids some day. You're taking my ideas out of context, here.


Why are you afraid of a robot killing you anyway?

Because each one could possibly be a murderer, but a murderer with a mind millions of times faster than my own. I doubt many humans would stand a chance, one-on-one or when engaged in large battles. They would be superior to us in too many ways.

Another solution to this problem is perhaps binding man and machine together. Don't make robots human, make humans robot. Give us the intellectual capacity of a quantum computer and immortality. Humans would still want to kill each other, undoubtedly, but the playing field would still be fairly level.


Trying to remove their ability to defend themselves may make them more dangerous than if they had human freedom anyway.

I don't see what's wrong with defending themselves. It's not that hard to subdue someone and at the same time not kill them if they're attacking you. Plus, robots could just radio for aid as soon as necessary.

I think you're hooked on the Asimov-like idea that robots will be machines that can move at half the speed of light and lift one million times their own weight. We don't have any other machines that can do that.

Even their super-genius is not really guaranteed, for it may come about that intelligences can only be made possible through giant associative engines that resolve many very complicated problems in parallel. If this is true, they may not actually think as fast as we do.

On the other hand, if quantum computer intelligence and immortality come up for grabs I'll be first in line, so I agree with your last statement there.

Perhaps...we all will be dead or go back to the stone age before quantum computing intelligence emerges. I agree with BBH that giant associative engines should drive the future - whether that would be a reality...your guess! BTW, there was this story about massive computation with fragments of information on StarTrek Voyager....very interesting.

I think you're hooked on the Asimov-like idea that robots will be machines that can move at half the speed of light and lift one million times their own weight. We don't have any other machines that can do that.
Of course AI could be incorporated into an interstellar spacecraft, or into a giant excavator, or the traffic control system for a city of ten million people; there is no limit on the size of an artificial mind, although as you say spped is a limiting factor.


Even their super-genius is not really guaranteed, for it may come about that intelligences can only be made possible through giant associative engines that resolve many very complicated problems in parallel. If this is true, they may not actually think as fast as we do.
Exactly right. An intelligent dyson sphere (one of Robert Bradbury's 'Matrioska Brains') would take days to have a coherent thought.

But such an entity could have trillions of smaller thoughts simultaneously; an entirely different type of thinking being to a human, and almost totally incomprehensible.

Of course AI could be incorporated into an interstellar spacecraft, or into a giant excavator, or the traffic control system for a city of ten million people; there is no limit on the size of an artificial mind, although as you say spped is a limiting factor.


There is already a giant I operating at the very moment to run the planet known as Earth. :D