Amazing Facts.

[peewee_RotA]

I think humanity at its core has something that can't be mechanically reproduced. Our "soul" for the lack of a better word, I'm an atheist myself. But even when computers become far more advanced and capable of far more intelligence than our smartest "us", it'll still be a box that will never know lust, or fear, or hate. Maybe approximations of them, if some programmers get bored, but not the real things.

http://data.antonindanek.cz/Harlan%20Ellison-I%20hav%20no%20Mouth%20and%20I%20must%20scream.pdf

[Stinger]

So what is the purpose of creating artificial intelligence? To replace humans? If by 2020 we have a computer with the processing power of one of my 3 year olds, I still don't trust my three year old to fill her own cup with juice. And by 2030 we have a computer capable of becoming a telemarketer or a cab driver. Then I guess the trillion dollars per unit is going to replace the guy in india working for 6/hr.

Telemarketers are already being replaced en mass by technology. Telemarketers and "cold-calling" are on their way out in favor of highly-targeted lead generation systems being powered by systems like Google's Adwords.

I'm with you that anytime statistics are thrown out as "fact", it's time to start taking them with a grain of salt. But our history is filled with jobs being lost when technology got advanced enough to replace them. I'm positive it's only a matter of time before you can walk in to a McDonald's and have everything, from the ordering to the cooking to the serving, all be done by one machine. It's amazing how many jobs are out there that really can be done by computers only slightly smarter than we have now, and the cool thing about technology is that it gets cheaper as it gets older.

Basically, machines will take over for the guy in India cold-calling for $6 an hour after the technology has been around for 5-10 years. It never takes too long after a technology is invented for it to become affordable for small businesses. Take digital camera's for example.  About 9 years ago my dad bought a top of the line digital camera for $1500 so that he could do professional photography. For christmas last year I got a camera for $200 that shoots at 4 times the resolution, and does video at TV quality, and shoots to multiple formats.

Yeah, the technology won't go straight from "It has been invented" to "It replaces people." But, 6 dollars an hour, 8 hour days, 250 days a year? That's $12,000. And if you have a couple of guys like that, buying a big-ass expensive machine for 30,000 once seems like a real bargain. Especially if the company that makes it has good leasing options!

I make my living as a freelance consultant in a field that didn't exist 10 years ago. It really won't surprise me too much if in another 30 years, taxi-cabs drive themselves and my tech support calls are handled by 100% a virtual female voice.

[quadz]

If computers get that powerful, then I think the development of a thinking machine is probably inevitable.

You're attributing a "law" formulated for computer hardware to a concept that is inherently only possible through software. If the only program that ever existed was solitaire, then a supercomputer would only be able to play solitaire. The software may be limited by hardware requirements, but you are assuming that the software either exists or is possible.

I'll have to apologize for oversimplifying.  As a programmer who has been interested in AI for quite some time, I've got no shortage of relevant textbooks on my bookshelf, as well as books by futurists such as Hans Moravec (from a robotics point of view), and Ray Kurzweil (from an exponential technology growth slash brain structure simulation point of view.)

I'm well aware of the difference between hardware and software.

Here's a project simulating the neocortical column of the brain down the the molecular level:

http://www.research.ibm.com/journal/rd/521/djurfeldt.html
http://en.wikipedia.org/wiki/Blue_Brain

Despite that, assuming that the software is possible, it's inevitability is not guaranteed.

Not much is guaranteed.  There's similarly no guarantee we'll ever figure out what all the genes in our genome do.  But there are an awful lot of smart people working on that, and more computer power will help.

Keep in mind our entire genome is only 750 MB uncompressed.  And only 1% of that differs from chimpanzees'.  And only some percentage of that constitutes the instructions that lead to the differences between their brains and ours.

The problem is being attacked from many angles.  And best of all, we have an existence proof -- working brains to study, and the genetic instructions to build said brains.

Direct modeling of human brains in software seems unlikely to be feasible from a practical AI standpoint... but modeling can help us understand how the brain works... And the better we understand how it works, the better our chances of translating those mechanisms into their software equivalents.

And increased computing power will accelerate the research.

After an extreme cost of R&D Honda has developed a functional walking robot. Does that mean that our cars will inevitably be replaced by mechs? Or that robots will ever replace workers? To do what? Assemble Toys in Mexico? The labor that a robot would be used for will never be cost effective. And the labor that should be is currently being done by machines on assembly lines and it will continue to be more cost effective for assembly machines to be improved rather than replaced with robot workers.

So what is the purpose of creating artificial intelligence? To replace humans? If by 2020 we have a computer with the processing power of one of my 3 year olds, I still don't trust my three year old to fill her own cup with juice. And by 2030 we have a computer capable of becoming a telemarketer or a cab driver. Then I guess the trillion dollars per unit is going to replace the guy in india working for 6/hr.

To borrow slightly from your own rhetoric, this is the same failure of imagination that led some otherwise very smart people in the 1950's and 1960's to proclaim that computers would never find their way into average consumer households.  After all, computers were very bulky, prohibitively expensive, and what would average people do with one anyway?

This is the same logical fallacy as the planet "running out of oil". We'll never run out of oil because once gasoline is less cost effective than another source of power, gas will no longer be used.

It's hard to believe you could have written that with a straight face.  People use the phrase "running out of oil" as a shorthand to refer to some point to the right of the peak of the graph where production capacity falls off, driving the price up to where we are increasingly forced to pursue alternative energy sources.  The catch being no current alternative can even begin to compete with the cost/energy yield of a barrel of oil.  So yes, we will solve the problem out of necessity somehow--but experts who don't see any easy solution are worried the transition will be rough.  That rough part there is what we mean when we talk about "running out of oil".


Regards,

[peewee_RotA]

You're making the wrong assumptions. There is a big difference between a robot, and a machine that operates based on human input. Stinger is absolutely right. Machines have been a blessing to labor and have revolutionized production. They make humans not have to do unsafe work.

Would it really ever be effective to replace a factory assembly line with 100 robot workers? That's literally a step backwards in technology. And the computers that control these factory machines are really quite unimpressive.
http://en.wikipedia.org/wiki/Programmable_logic_controller#Features

If a supercomputer in 2030 is the size of a desktop today, then based on the fact that all banks still use COBOL (negating the "things you learn in a 4 year degree are irrelevant" half-truth) and that in a world of quad processors entire factories are run on 100mhz processors, I doubt that the existence of the new technology automatically translates to it's adoption or it's necessity.

Which begets the next point. The measure of a computer "being as smart as a chimpanzee" is just an abstract measurement. It does not indicate that we have AI, and it does not swing from trees. It's a comparison of numbers, most likely a comparison of number of synapses fired per second vs. processing speed of the computer. This is why I said "a computer as smart as my 3 year old." Obviously a computer "1/100th" as smart as my 3 year old can do an entire calculus problem in less than a second and my child at her current state given 100 years could not. Assuming that a computer as smart as a 3 year old would spill juice just like a 3 year old is silly. That computer couldn't even manipulate a physical object like a juice carton without being given a peripheral and additional programming.

If that machine could be instantly given a new peripheral and given X amount of time to learn how to use it on it's own. That would be AI. That does not exist, and you can't just assume that that's going to exist just because the computer capable of processing it does. This "amazing facts" video does not say anything about the existence of AI. Only the prediction of the processing power of computers in the future.

Which is why you indadvertantly proved my point. Do we need cars that can go 4000 miles per hour? Do we need a truck that can haul 1,000,000,000 tons? The existence of the computer proves that new technology replaces old technology. We most likely will never see desktop super computers because focus will be most likely given to mobile devices. Mobile devices as powerful as our desktops is probably a better prediction. Some other technology that replaces the mobile device is a better prediction.

Technology is not measured by how powerful you make an old invention. It's measured by the administering of a new idea. Quite frankly the idea that SkyNet will control all of the worlds' defenses, or that a central brain will control all of the robots trying to kill Will Smith is a quite old one. If we have desktop super computers, they will be used in ways we never imagined today. If AI ever exists, it will be used in ways that was never imagined today. (Because it imagined those new ways by itself?)

The example of the robotic cab driver is not an issue of AI because it's an object that simply has to accept inputs and react logically based on it. More powerful and smaller computers will help enable this, but it will still be at it's core a logically driven computer that needs new sensors and peripherals installed on all the highways. In fact it's a technology that already exists but is not used, because it's not practical. If every city on earth purchased a "johnny cab" then the price would come down and some day it would be practical. But if no city on earth ever purchases them then it never will become practical.

True AI I never see existing because the human thought process is the logical opposite of a computer computing. The entire concept of a computer would need reworked. From the fact that our brains are analog, to the fact that humans are illogical and make mistakes. A computer that makes a mistake or is illogical, is not a profitable investment. Also how can a computer ever understand good from bad? Not morally, but logically. If a baby is learning how to stand, he is able to dynamically determine that one thing is bad and one thing is good. Applying upward force using his legs is good. Applying upward force using his legs while not holding onto something is bad because he falls and feels pain. Standing is good because he sees other people do it. Standing is good because when he stands everyone in the room cheers.

So the limit of AI really is that of emulating humans. You have to assume that this will always be logical and that without inventing a computer capable of making mistakes you can never have a computer that can learn, program itself, invent more of itself, adapt new functionality without the input of a human, or make moral decisions let alone have feelings.

I think that Bicentennial man is the best example of a future assuming that AI is possible, not Terminator or the Matrix.


P.S. The comment on peak oil production. Most people don't even understand how the market determines that. The movie Tooth and Nail is about running out of oil. It can't really happen because the market plays such a big part.

[quadz]

You're making the wrong assumptions.

On the contrary, your argument is cluttered with irrelevancies. 


While I agree with Stinger to the extent that we don't yet know enough about consciousness and its relation to our biology to understand the mechanics of qualia (i.e. how and why we feel things), and I agree that without this understanding we shouldn't expect machines to have biological feelings or emotions, I do not perceive this as an obstacle that would block the so-called technological singularity from being able to occur. (*)

* Unless perhaps intuitive creativity turns out to be impossible without access to conscious qualia.


To me the key question boils down to: Is it possible for a software program to exist which is smart enough to improve its own design.


If so, the implications are staggering.  All the other questions, like whether it could ever have what we regard as true feelings, are interesting, but subordinate.

By the way, in my view nature has all but answered this question for us.  DNA is a program capable of improving its own design. (*)

(* Yes, said improvements occur indirectly without any goal or intent driving them.(**) And DNA is not quite software, but instructions for building a biological machine. We know all that... Nevertheless, it's a hell of an existence proof of a powerfully adaptive program..........)

(** Actually, DNA is now producing biological machines that are studying their own DNA and working with intent toward eliminating defects and making improvements in their DNA, so....... Pretty amazing program indeed.)


Regards,

quadz

[Stinger]

Quadz, I share your enthusiasm for this being something that we may be able to achieve through some programmable means, but I think your first asterisk is actually the biggest real obstacle to against this occurring.

Making a program that handles a task with the scientific method sounds way doable, and that it how you always guarantee improvements to a technology while it's in motion, but the problem is in quality of the input these machines could provide without human intuition to guide it.

I'm in marketing. When I start a new project, there are a series of rules I have to follow. I take an educated and creative guess on the best possible ad to get someone to buy a product. Then I save that one, put it out to the market, and take another educated and creative guess to make a second one. Test one against the other real time, and one always proves itself better. So you delete the loser, and write a new one to try to beat the winner. As long as you never stop testing, you never stop improving. But the system can't do this itself. Wouldn't know how, and I don't think it could be taught.

There are steps to follow, instructions that could be programmed, but when it came to the step where it had to enter something genuinely creative it would have to rely on algorithms. Random numbers, or math problems that could maybe cheat their way through a basic approximation of human interaction. And they would all suck, and fail.

It takes no human intuition to replicate something exactly, that can be taught. DNA, like you said, is an awesome example of self-propagation, but like you said, it's just following instructions. But to create something completely, or to improve on something that is creative in nature (all technology is creative in nature), that requires the spark of higher consciousness. I think that comes from a place of human irrational thought and feeling. I'd be really stoked if we figured out a way to cheat around this, but I'm not optimistic about seeing it anytime soon, if its even possible.

[quadz]

It takes no human intuition to replicate something exactly, that can be taught. DNA, like you said, is an awesome example of self-propagation, but like you said, it's just following instructions. But to create something completely, or to improve on something that is creative in nature (all technology is creative in nature), that requires the spark of higher consciousness.

Except that it would appear DNA was "improving itself" for a few billion years before it ever arrived at a program sequence to build a biological machine that has any access to higher consciousness.

So nature has demonstrated that it's possible, given the right algorithm, to begin with something less than an amoeba, and crank that algorithm for a few billion years, and end up with a program that specifies the construction of thinking biological machines.

Where was the spark of higher consciousness during that multi-billion year evolution of that program sequence?


On a related note, this would seem to beg the question of whether there's anything magical about biological machinery. We know that DNA is an instruction sequence for the construction of biological systems.  And apparently if a biological system is constructed a certain way, it can have access to consciousness and the potential to think and reason.

So: is access to consciousness and thinking and reasoning available only to carbon-based biological constructs?

If not, then this should imply a non-biological system, constructed in a certain way, could also have access to consciousness and the potential to think and reason.

Or, if there is something special about the chemistry of our biology as relates to consciousness--why is that?  As far as we know it all reduces to elements of the periodic table, bound together by physical laws and obeying chemical processes - so where does the magic occur?  (Even if the functioning of our brains were to depend at some level on quantum mechanical behavior, it doesn't seem obvious that such effects could only be achieved through biological chemical reactions. We've successfully designed a few small quantum computers, for instance.)


Regards,

[kren.Z]

                                     

 We've successfully designed a few small quantum computers, for instance.)

 
 
That statement is not entirely true. D wave (the company that specializes in quantum computing) has supposedly announced a working prototype, but i think its highly skeptical since they have yet to release any technical information. I don't think the general public would be informed of such achievements that D wave claims, since a fully function quantum processor would be of no use to them. It's actually quite a huge technological advancement if such a thing were built. Cryptography would be rendered obsolete for example...so you can ponder the consequences of something like that going mainstream.
 
 
 
 
 
We are in need of a breakthrough in science if we ever want to see a revolution in AI or whatever else is being discussed in this topic. Current physics is stuck. Early developments in quantum theory led to the birth of modern electronics, but the same can't be said for our advanced theories of today (string theory, loop quantum gravity, unification etc..). It's true they've given us some insights and new mathematics, but having any real practical application remains improbable. We need another genius like faraday or tesla.
 
 
 
 
 
 
 
 
 
Henry Ford was once asked what the people were expecting from transportation of the future. He simply responded with "faster horses".
 
 

[quadz]

 We've successfully designed a few small quantum computers, for instance.)

That statement is not entirely true. D wave (the company that specializes in quantum computing) has supposedly announced a working prototype, but i think its highly skeptical since they have yet to release any technical information.

Thanks, I overstated that a bit.  What I had in mind by "small" was the entanglement of multiple qubits has been demonstrated in research labs. [1]  And that we do have virtual machines for quantum computing simulation. [2]
 
 

We are in need of a breakthrough in science if we ever want to see a revolution in AI or whatever else is being discussed in this topic.

The one thing I think everyone here is agreed on, is that very fast computers does not automatically imply very smart computers.


Regards,

quadz

[Arm0r]

 

Major necro post but the conversation in this thread is very interesting.  Thought it could use one.

[reaper]

We're not even close to having the hardware or software to do something like this.  However I suspect eventually we'll have the capability to produce machines like in iRobot.  Possibly in thousands of years.  But we will only produce what is better of designed by god, and it will never be produced since we will never have the capability of producing it perfectly even with a purpose of being a tool.

It comes down to the input being taken from sensory devices similar to ours, and the instructions take this input and grow the machine (as they say a learning computer, so to us it just doesn't feel everything is a set function like given any value of x you can predict the output of f(x)).  Would we want the computer "thinking" randomly?  I believe quadz is hinting at there being no difference in theory, between these machines and humans; in other words, a machine similar to us is feasible.  But that is not true imo.  You can take look and imagine where we will be, but there's still differences.

 

[quadz]

But we will only produce what is better of designed by god, and it will never be produced since we will never have the capability of producing it perfectly even with a purpose of being a tool.

I'd think one of the most outrageous insults you could make to your deity would be to assign it responsibility for the rather crappy unplanned design produced by evolution.

A few intelligent tweaks to our DNA would put most of the non-emergency medical, dental, and pharmaceutical industries out of business.

I believe quadz is hinting at there being no difference in theory, between these machines and humans; in other words, a machine similar to us is feasible.  But that is not true imo.  You can take look and imagine where we will be, but there's still differences.

If you could please construct an actual argument against my claims, instead of outputting this sort of weirdly self-contradictory word salad, I'd love to read it.

The essential premise from which my arguments are based is the view of the brain as an information processing system.

I'm saying: If there's nothing magic going on inside the brain, so that consciousness truly arises out of the brain's electro-chemical neural processes, then it should not be a stretch to assume that equivalent information processing could be performed on a non-biological substrate.

(I am not saying that a robot would be "the same" as a biological creature.  I am simply focused on the implications of the brain as an information processing system.)


Regards,

[reaper]

I'd think one of the most outrageous insults you could make to your deity would be to assign it responsibility for the rather crappy unplanned design produced by evolution.

A few intelligent tweaks to our DNA would put most of the non-emergency medical, dental, and pharmaceutical industries out of business.

This means almost nothing and remember what doesn't kill you only makes you stronger and life is harmonious.

If you could please construct an actual argument against my claims, instead of outputting this sort of weirdly self-contradictory word salad, I'd love to read it.

The essential premise from which my arguments are based is the view of the brain as an information processing system.

I'm saying: If there's nothing magic going on inside the brain, so that consciousness truly arises out of the brain's electro-chemical neural processes, then it should not be a stretch to assume that equivalent information processing could be performed on a non-biological substrate.

(I am not saying that a robot would be "the same" as a biological creature.  I am simply focused on the implications of the brain as an information processing system.)

This is not a simple subject and it's uncharted terrority, so it's not the easiest place for arguments; however I do have some theories about how we would fail at the start when creating a machine like a human.  As well as a few others. 

Of course you can have a machine brain be the same if you can feed it the same input and give it the same processing capabilities.  You're copying, that's one of the first techniques we learn.  But there is no such system, at anything past the most basic level.  Attempting to create anything like a human would fail from the start.

[quadz]

I'd think one of the most outrageous insults you could make to your deity would be to assign it responsibility for the rather crappy unplanned design produced by evolution.

A few intelligent tweaks to our DNA would put most of the non-emergency medical, dental, and pharmaceutical industries out of business.

This means almost nothing

It means that our species has numerous sub-optimal characteristics from a design standpoint, as evidenced by problems with our teeth, problems with our lower back, problems during childbirth, susceptibility to a multitude of diseases and viruses, etc., etc., etc.

It means any designer who would have produced our species intentionally with all of our myriad flaws would either have to be incompetent, or kind of a total dick.

and remember what doesn't kill you only makes you stronger and life is harmonious.

The cognitive dissonance produced by this statement makes it self-refuting.

Of course you can have a machine brain be the same if you can feed it the same input and give it the same processing capabilities. 

Interesting.  With this statement you have agreed to my entire premise.  In fact, you have stated my premise more strongly than I did.  (When I stated it, I stipluated some preconditions with a bold and underlined If.... clause.)


Regards,

[reaper]

You have to believe before you can see the evidence.

Of course you can have a machine brain be the same if you can feed it the same input and give it the same processing capabilities. 

Interesting.  With this statement you have agreed to my entire premise.  In fact, you have stated my premise more strongly than I did.  (When I stated it, I stipluated some preconditions with a bold and underlined If.... clause.)

If you can copy the brain, you have a brain.  In theory if you built a quantom processing environment that learned from its input, and its input was similar to ours through sensory systems, it could learn and think.  This is based of the original theory of a "computer" from Turing, now you are just extending his theories saying you have parallel processing similar to a nueral network with live input.   Basically it's the logical conclusion of his theories and now we have more information.  In some perspectives it's been tested, so I would state it like I did.

the end though, you wouldn't of created what should have been created, and there's to much room for error.  If you look at modifying the human instructions, if it was controlled properly with fallback methods, you're still playing god, and you're still taking risks.