1. Comment #140665 by 42nd on March 8, 2008 at 9:36 am

It's about time to exorcise ghosts from the machine.

Other Comments by 42nd

2. Comment #140666 by Steve Zara on March 8, 2008 at 9:40 am

10,000 neurons, with about 30 million synaptic connections between them. "The column has been built and it runs," Markram says. "Now we just have to scale it up." Blue Brain scientists are confident that, at some point in the next few years, they will be able to start simulating an entire brain.

Perhaps, but not a human brain. The human brain contains ten million times as many neurons as this simulation.

Other Comments by Steve Zara

3. Comment #140667 by JD Cherry on March 8, 2008 at 9:41 am

I don't see how you could ethically create a person in a box (or four boxes) and use it for experimental purposes. If it's a full simulation of the human brain, wouldn't it get hungry, sad, afraid, angry, horny? I doubt we'll see anything like consciousness produced in a supercomputer for a long time, besides.

That said, the practical uses outlined in the story are intriguing.

Other Comments by JD Cherry

4. Comment #140668 by gd_edi on March 8, 2008 at 9:51 am

Pretty interesting, I can't wait to see what comes out of this, thought the more I think the more it seems to me that consciousness doesn't actually exist at all.

Taking the study to its logical end would mean you build a 'machine' that mimics the human brain and its human shell down to the smallest detail. Essentially what you have built is a human. It behaves strictly like a human and is indistinguishable. Is it conscious? If not, then can we conclude that an identical 'real' human is conscious? How can we demonstrate that a machine is conscious when we can't really do likewise for the billions of other 'conscious' beings on this planet?

Of course the flipside, here I am feeling as conscious as ever!

Other Comments by gd_edi

5. Comment #140669 by Geoff on March 8, 2008 at 9:57 am

Steve: very true, but he does make that point towards the end of the article. A very timely article, given some of our recent topics under discussion.

To an extent, he's relying on Moore's Law for that.

I'd love it to happen, but I'm not that optimistic - yet. He's talking in terms of ten years.

Other Comments by Geoff

6. Comment #140670 by mdowe on March 8, 2008 at 10:00 am

I can see it now ... at the end of this monumental undertaking we'll have a phenomenally powerful computer capable of the most mind-boggling feats of dimwitted irrationality and wishful thinking, and entirely convinced it was created in the image of God. And that will only be when isn't slacking off or thinking about sex.

Other Comments by mdowe

7. Comment #140675 by Steve Zara on March 8, 2008 at 10:19 am

Comment #140669 by Geoff

I'd love it to happen, but I'm not that optimistic - yet. He's talking in terms of ten years.

I think that is wildly optimistic. Moore's Law is a doubling about every 2 years. A decade is 5 doublings. That is only 32 times as powerful.

Sorry :(

Comment #140670 by mdowe

I can see it now ... at the end of this monumental undertaking we'll have a phenomenally powerful computer capable of the most mind-boggling feats of dimwitted irrationality and wishful thinking, and entirely convinced it was created in the image of God. And that will only be when isn't slacking off or thinking about sex.

This is an interesting point, especially when combined with:

Comment #140667 by JD Cherry

I don't see how you could ethically create a person in a box (or four boxes) and use it for experimental purposes.

Within the next century, we will have to deal with some serious ethical issues. What will it mean to wipe out the data from an AI? To duplicate it? To turn it off?

Other Comments by Steve Zara

8. Comment #140676 by pulsar1z on March 8, 2008 at 10:25 am

There is no doubt in my mind that in the future there will be artificial intelligence capable of conscious awareness and it will be considered murder if you deactivate it. It is only a matter of time.

Other Comments by pulsar1z

9. Comment #140677 by Steve Zara on March 8, 2008 at 10:27 am

Comment #140676 by pulsar1z

There is no doubt in my mind that in the future there will be artificial intelligence capable of conscious awareness and it will be considered murder if you deactivate it. It is only a matter of time.

Now here are some questions to think about - does it count as murder if you have an exact backup? Or how about a backup from an hour ago, or a day ago, or a year ago?

Other Comments by Steve Zara

10. Comment #140679 by mdowe on March 8, 2008 at 10:32 am

Hmm, ethical considerations, a good point. Maybe we should set some new rules at the outset: When we get an A.I. it will be considered murder if you purge it, but when we get a hyper-religious A.M. (Artifical Moron) it will be considered a public service to turn it off.

Other Comments by mdowe

11. Comment #140680 by Mitchell Gilks on March 8, 2008 at 10:33 am

Ten years isn't nearly enough time to meet the raw processing power it would require, maybe 30-40 year supposing moore's law obtains. No sooner.

Perhaps more, being that it could only process 22 trillion instructions per second. The fastest supercomputer in the world is up to 500 trillion instructions per second, and won't match the human mind for 20 more years worthy of doublings at the least. So their brain would have to keep up with the fastest computer in the world. Highly unlikely, it will follow at the very least a decade behind.

Other Comments by Mitchell Gilks

12. Comment #140681 by vijikumar on March 8, 2008 at 10:36 am

What about AI systems wanting to commit suicide or help to be deleted - euthanasia?

Other Comments by vijikumar

13. Comment #140682 by Mitchell Gilks on March 8, 2008 at 10:38 am

I wonder if an equivolently intelligent AI would be able to sort through all the bullshit we believe, and how poor we are at reasoning, or if it would be no better, and believe a load of bullshit too.

We will have to engineer it a prolific ass like ours, so it can be just as good at pulling stuff out of it.

Other Comments by Mitchell Gilks

14. Comment #140683 by alexlg on March 8, 2008 at 10:40 am

Let's just hope that the brain doesn't turn out to be a raving theist.

Other Comments by alexlg

15. Comment #140684 by Steve Zara on March 8, 2008 at 10:41 am

Comment #140680 by Mitchell Gilks

Ten years isn't nearly enough time to meet the raw processing power it would require, maybe 30-40 year supposing moore's law obtains. No sooner.

It may not be quite that bad, as much of the brain seems to be required for supposedly unintelligent matters like motor control. Also, we have no idea how much redundancy there is. Some young children with major brain damage, leaving only a fraction of normal brain tissue, seem to be able to develop as normal adults. Also, there is the question of brain/body mass ratio. How much brain would we need to be intelligent if we were only a few inches tall?

However, the problem with Moore's Law is that it is exponental. Almost all the power comes right at the end of the time period. This means that even a 100-fold reduction in requirements would only cut between around 15 years off your prediction.

Other Comments by Steve Zara

16. Comment #140685 by JD Cherry on March 8, 2008 at 10:42 am

Before anything resembling consciousness could arise from a simulated brain you'd need a simulated body. I'm sure of that. He says it will talk to him, but how will it know English, or German?

I'm sure that a brain trapped in an immobile refrigerator would give us plenty of opportunity to study depression and dementia.

Also the thing about Moore's Law is that we're gonna be hitting fundamental limits by the end of the next decade in terms of transistors. Any future developments like Quantum computing or carbon nanotubes are speculation at this point.

Other Comments by JD Cherry

17. Comment #140686 by Mitchell Gilks on March 8, 2008 at 10:51 am

You're correct Steve, I overlooked all of those very good points.

I still think 10 years is overly optomistic, but I could be wrong. I haven't a clue how much of our brain is used for cognitive awareness, I doubt anyone knows. It could be considerably less than I may suppose.

I don't think that a 1 inch tall person can be as smart as someone our size, without a far superiorly evolved brain. Size seems to be the easist solution, it is the only major difference between a chimps brain and our brain.

Other Comments by Mitchell Gilks

18. Comment #140687 by Quetzalcoatl on March 8, 2008 at 10:52 am

JD Cherry

Before anything resembling consciousness could arise from a simulated brain you'd need a simulated body. I'm sure of that

What makes you so sure?

Other Comments by Quetzalcoatl

19. Comment #140688 by Bonzai on March 8, 2008 at 10:55 am

JD Cherry

Before anything resembling consciousness could arise from a simulated brain you'd need a simulated body.

Excellent point, There is the knowledge of the mind and the knowledge of the flesh. The geeky type often ignore the body and talk as though it is just the mind that counts. I don't even want to get started on the misanthropes who call themselves "transhumanists".

P.S, I think in some way "irrationality" may be essential to genuine intelligence, it is probably link to insights, heuristics, big picture thinking imagination and all those qualities that cannot be captured by logic alone.

Other Comments by Bonzai

20. Comment #140689 by pulsar1z on March 8, 2008 at 10:58 am

Now here are some questions to think about - does it count as murder if you have an exact backup? Or how about a backup from an hour ago, or a day ago, or a year ago?

No backups only clones

Other Comments by pulsar1z

21. Comment #140692 by 5keptical on March 8, 2008 at 11:06 am

It looks like they're doing some interesting science... perhaps.

The write-up emphasizes the more outlandish claims of the researchers but that may not be what the research actual shows.

I hope so, because it appears to be another case of a computer science transplant that figures he just has to find the right set of axioms for these silly biological systems to fall in line... but again that could just be the write-up.

More interesting for me will be the quality of the comments here... let's see what you got folks!

Other Comments by 5keptical

22. Comment #140694 by alan baylis on March 8, 2008 at 11:16 am

The human brain has evolved in the environment we have had to survive in.It's a pretty nifty organ!.Does this mean though,that it is the best model to follow if we want intelligent thinking computers to help us solve the problems, scientific and otherwise, that the world will face in the future?

Other Comments by alan baylis

23. Comment #140695 by bentleyd on March 8, 2008 at 11:18 am

Just how real such a supercomputer might become...

http://www.youtube.com/watch?v=uA4iPFsEW0I&feature=related

Other Comments by bentleyd

24. Comment #140696 by JD Cherry on March 8, 2008 at 11:20 am

The geeky type often ignore the body and talk as though it is just the mind that counts. I don't even want to get started on the misanthropes who call themselves "transhumanists".

I think that the disregard for the importance of physical bodies is at the root of the failure of AI. We think with our bodies. Rodney Brooks realized this years ago. I recommend checking out his entry in this years Edge question centre. Roger Schank's too.

It seems that the rhetoric of transhumanist types is getting louder even as the computational metaphor shows its glaring limitations.

Other Comments by JD Cherry

25. Comment #140697 by RSP on March 8, 2008 at 11:22 am

As I understand it, this simulated brain would have no stimulus and more importantly no biological chemicals to create emotions right? So no sight, smell, touch, sound, or hormonal urges. Seems strictly a model. Unless they have a way of sending it data, there's not much for such a consciousness to conceive.

Other Comments by RSP

26. Comment #140698 by Dr Benway on March 8, 2008 at 11:36 am

RSP: As I understand it, this simulated brain would have no stimulus and more importantly no biological chemicals to create emotions right?

There's nothing special about a wet medium as opposed to a dry medium for information.

Emotion is an organizing program for motor schemas. Schemas associated with positive emotion have a higher probability of disinhibition as compared to schemas associated with negative emotion.

The programmer might set general parameters for output, such as "A human smile just after your do something means what you did was positive." We won't know how the machine *feels* when a human smiles. But its behavior will appear to us as though it *likes it* when someone smiles.

Other Comments by Dr Benway

27. Comment #140702 by JD Cherry on March 8, 2008 at 11:49 am

Dr Benway - For this particular project, there won't be a programmer setting these sorts of algorithms. The goal is the simulation of a human brain, not the creation of a strong AI.

For some of the more extreme hopes of the Blue Brain project to come about, consciousness would have to spontaneously emerge from the wiring together of simulated neocortical columns.

Other Comments by JD Cherry

28. Comment #140705 by Teratornis on March 8, 2008 at 11:50 am

In reply to comment #140677 by Steve Zara:

Now here are some questions to think about - does it count as murder if you have an exact backup? Or how about a backup from an hour ago, or a day ago, or a year ago?

I'd say the backup's opinion would have to weigh heavily here. Perhaps it would take precedence over our armchair speculations.

The crime would still be assault, and even attempted murder, because no backup system is 100% reliable. If you broke into a corporate office and destroyed the computers, you would still be guilty of a serious crime even if the corporation had a complete offsite backup and was able to continue operations with minimal disruption.

I might add that we already face somewhat similar questions with mind-altering drugs such as Flunitrazepam (Rohypnol) and Propranolol. If the survivor of a violent crime has no conscious recollection of the crime, or could be medicated to feel no trauma whatsoever from it, has a crime occurred? Courts generally say it has. At a minimum, a criminal who administers drugs to a victim is practicing medicine without a license, or if the criminal is a physician, a breach of medical ethics has occurred.

Also, with respect to the predictions about how long it will take to get to simulating a whole brain, note that the project described in the article doesn't even try to exploit the maximum computing power available now through distributed computing projects (e.g. SETI@Home, Folding@Home, etc.).

We can probably assume the computing power available to distributed computing projects in the future will exceed what a typical research lab can purchase in its own boxes, by a rather large multiple. This effectively lets a research project perform computations as if they are 5 to 10 doublings farther ahead with respect to Moore's law. Maybe even more than that.

Of course a research project will have to compete with other research projects for donated computing time. A lot would come down to what excites people more: searching for radio signals from an alien intelligence for which we have only flimsy arguments to offer hope, vs. building our own alien intelligence, a quest which not only seems plausible but should have distinct milestones along the way to tell us whether we are wasting our time.

A distributed brain would not have to be able to answer questions in real time to be interesting. And in fact a slowly-responding brain would function like a distant extra-terrestrial intelligence communicating via radio with a substantial transmission delay. We already hope to communicate with intelligences which can only answer slowly, so conceptual precedent exists.

We can also imagine a hypothetical communication between two humans, one who is "stationary" and the other who is riding on a spacecraft at relativistic speed. Even if the speedy human appears to slow down 1024 times to the stationary human, they could still manage a discussion of sorts.

That offers another way to get a jump on Moore's law. When you are ten doublings away from being able to simulate a brain in real time, you are able to simulate a brain at 1024th the speed of real time.

Given that a Prof. Dawkins can formulate witty replies to questions in less than one second (we can probably ignore the several seconds he needs to articulate the reply as a mere transmission delay owing to the physical characteristics of the speech apparatus, i.e., he probably has the complete thought already when he begins to utter it), it wouldn't be too disappointing to have to wait 1024 seconds for the reply to begin. I could certainly find a practical use for an expert intelligence which takes time to answer questions. We already function with that limitation on online help desks and discussion sites (like this one).

Which is to say, when you are 10 Moore's law doublings away from simulating a brain in real time, you probably already have a useful brain.

So while I agree that most of the performance gain comes late in the Moore's law exponentiation, I would argue that you don't need most of the performance gain to get to some amazingly useful results.

After all, Prof. Dawkins uses the overwhelming majority of his brain power simply to navigate the complex, noisy, ambiguous physical world. Consider how much brain power he needs to get to his office in the morning, locate a piece of chalk, and write a clever message on a board. The clever message itself represents only a tiny fraction of the neurons he had to use to plow through the massive real-world overhead. It's all about Aristotle's accident and essence once again.

Another possibility is meeting in the middle, i.e. instead of trying to build a brain entirely from the bottom up, the researchers working from the top down might be able to contribute usefully as they work out better ontologies and so on. This would probably be necessary in any case to get to a desirable result, i.e. to create an artificial human brain that combines the pattern-matching and associative memory strengths of our evolved brains with our rather unnatural yet highly useful quasi-algorithmic methods for critical thinking.

Humans have some innate capacity for critical thinking, but going all the way with it is highly unnatural. Hence religion and all the other departures from rationality in our world.

Plus we don't have a way to put such computing power as we already have into our heads. Imagine having only to close your eyes to perform Google searches, visualize graphs of any mathematical function, etc. It should be easier to equip the artificial brain with this than to retrofit our own gooey brains. I'm not as smart as Prof. Dawkins, but maybe I could out-perform him on many tasks if I could usefully put Google and Wikipedia et al. in my skull.

So, if we combine three methods for cheating Moore's law:

1. Waiting for replies (5-10 doublings)
2. Distributed computing (5-10 doublings)
3. Top-down cheating (? doublings)

we might cut 10-20 doublings off the wait for Moore's law to give us a real-time brain in one affordable box.

Other Comments by Teratornis

29. Comment #140718 by Geoff on March 8, 2008 at 12:23 pm

Here I am, brain the size of a planet...

Other Comments by Geoff

30. Comment #140719 by Dr Benway on March 8, 2008 at 12:30 pm

For this particular project, there won't be a programmer setting these sorts of algorithms.

The machine will just sit there doing nothing unless it's programmed with some sort of internal reinforcement system.

Other Comments by Dr Benway

31. Comment #140720 by Steve Zara on March 8, 2008 at 12:33 pm

Comment #140705 by Teratornis

The crime would still be assault, and even attempted murder, because no backup system is 100% reliable.

No system of any kind is 100% reliable. If someone is put into an coma so that they can be operated on in certain ways, there is no guarantee that their consciousness can be restore.

2. Distributed computing (5-10 doublings)

It's far better than that. SETI@Home has nearly 2 million hosts. That is a around 20 doublings. It's going to help get to a brain, but it isn't going to help get one in an affordable box. That is not what distributed computing is about.

Other Comments by Steve Zara

32. Comment #140722 by Steve Zara on March 8, 2008 at 12:37 pm

The machine will just sit there doing nothing unless it's programmed with some sort of internal reinforcement system.

I am not sure what you mean by that phrase - could you explain it further?

Other Comments by Steve Zara

33. Comment #140723 by Rational_G on March 8, 2008 at 12:41 pm

Bonzai and JDCherry:

I totally agree. The body is part of consciousness.

I recommend Antonio Damasio's book - "The Feeling of What Happens, Body and Emotion in the Making of Consciousness"

Other Comments by Rational_G

34. Comment #140726 by Teratornis on March 8, 2008 at 12:43 pm

In reply to comment #140685 by JD Cherry:

Before anything resembling consciousness could arise from a simulated brain you'd need a simulated body. I'm sure of that.

It might be easier to entertain such certainties if we did not live in a world with thousands of fully conscious quadriplegics. Many of them lost their bodily sensation and motor function after already having lived some time with functioning bodies, but there are also unfortunate people who have lived with profound physical and sensory deficits since birth, and yet seemed to have no difficulty attaining consciousness.

We have examples of humans living without just about every feature we associate with being a human: the use of their limbs, the use of their eyes, ears, speech, etc. As the Terri Schiavo case illustrates, you stop being a human when your consciousness dies. Everything else which does not maintain consciousness appears to be expendable.

Other Comments by Teratornis

35. Comment #140728 by Steve Zara on March 8, 2008 at 12:51 pm

Comment #140687 by Quetzalcoatl

JD Cherry

Before anything resembling consciousness could arise from a simulated brain you'd need a simulated body. I'm sure of that

What makes you so sure?

Excellent point, Quetz.

We so can't be sure about anything to do with the simulated brain until we have simulated it. That is the whole point. Anything we say now is mainly just guessing, We do have evidence that brains can develop well with considerable isolation from stimulus, such in the case of rare people who are born blind and deaf.

The issue here is whether or not a mind can develop without that stimulus from the start.

Other Comments by Steve Zara

36. Comment #140731 by Rational_G on March 8, 2008 at 12:58 pm

A living person, no matter how disabled, still has a body which maintains homeostasis.

Other Comments by Rational_G

37. Comment #140732 by Steve Zara on March 8, 2008 at 1:01 pm

Comment #140731 by Rational_G

A living person, no matter how disabled, still has a body which maintains homeostasis.

That isn't going to affect how minds develop, it is simply going to keep the brain cells alive. We don't need that in a simulated system.

Other Comments by Steve Zara

38. Comment #140734 by Teratornis on March 8, 2008 at 1:07 pm

In reply to comment #140720 by Steve Zara:

2. Distributed computing (5-10 doublings)

It's far better than that. SETI@Home has nearly 2 million hosts. That is a around 20 doublings.

I was conservative. Also, the article describes a computer that appears to be thousands of times more powerful than the average SETI@Home host. There might also be interprocessor communication advantages to having several thousand processors in the same box, compared to millions of computers around the world, connected by heterogeneous network links.

However, in favor of your argument, we can look to gaming machines which are emerging as very useful for certain kinds of distributed computing projects because they have specialized vector processors for handling graphics, which make them many times more powerful than typical desktop computers. One imagines the gaming machines in 10 years could be extremely useful for distributed simulations.

It's going to help get to a brain, but it isn't going to help get one in an affordable box. That is not what distributed computing is about.

Of course, but waiting 20-30 years for the affordable box is not much of a problem once you have proven it will work. I think the important question is how long you have to wait to know whether the research project will work.

The history of computing offers many examples of problems that were originally worth attacking with warehouses full of computers, and now we can do them on personal computers. If the goal is to build an autonomous robot that can function in the real world, then we have to get the computer down to a practical size, but even a brain in a warehouse or in a global distributed network would be exceedingly useful. It's not hard to think of uses.

Suppose a distributed computing project took one year to accumulate indisputable evidence of having simulated five minutes of a real human mind. I think that would be a result as important as a SETI breakthrough. Perhaps even more important, because with SETI we cannot anticipate any speedup in communication, but we can expect our own artificial brains to get faster, smaller, cheaper.

Of course this assumes Moore's law will hold, and that assumption becomes more questionable the farther out we extrapolate. I'm not about to bet against Moore's law, but the bet is not entirely certain.

Other Comments by Teratornis

39. Comment #140735 by Rational_G on March 8, 2008 at 1:07 pm

Steve -

Do you believe the simulated system can be conscious, without a sense of self, ie some sort of body? I guess we're going to find out eventually....

Other Comments by Rational_G

40. Comment #140737 by Steve Zara on March 8, 2008 at 1:11 pm

Do you believe the simulated system can be conscious, without a sense of self, ie some sort of body? I guess we're going to find out eventually....

That is a good question. I would speculate it may well be the brain's model of some kind of body that is important. But as you say...

Other Comments by Steve Zara

41. Comment #140738 by Nuspirit on March 8, 2008 at 1:12 pm

I don't really agree with the various statements about having to wait x amount of years before we could simulate a brain in real time.

Given the highly parallel nature of the problem, you can always achieve more computing power by throwing more processors at it. There are of course practical considerations with interconnecting all those processors, but in theory we could already have 1000 times "faster" computer than the currently fastest one if we just built one with 1000 times more processors and figured out an architecture to interconnect them all.

It wouldn't work for every type of computation, but in this case it could since the problem is rather straightforwardly about a huge amount of parallel computation.

Other Comments by Nuspirit

42. Comment #140740 by Rational_G on March 8, 2008 at 1:15 pm

I do like the idea of trying to build a brain from scratch.

Other Comments by Rational_G

43. Comment #140742 by HeyBishop on March 8, 2008 at 1:24 pm

Skynet

Other Comments by HeyBishop

44. Comment #140743 by Teratornis on March 8, 2008 at 1:27 pm

In reply to comment #140738 by Nuspirit:

I don't really agree with the various statements about having to wait x amount of years before we could simulate a brain in real time.

Given the highly parallel nature of the problem, you can always achieve more computing power by throwing more processors at it. There are of course practical considerations with interconnecting all those processors, but in theory we could already have 1000 times "faster" computer than the currently fastest one if we just built one with 1000 times more processors and figured out an architecture to interconnect them all.

Yes, we can in theory exponentiate the processors, but in practice it's hard to exponentiate the funding. Hence my suggestions above to use distributed computing (e.g., SETI@Home) and to relax the real-time constraint (be willing to wait some hours for the computer to simulate one second of human thought).

Another problem with buying your own processors is that Moore's law turns them into worthless junk in just a few years. Thus you need to be pretty sure you will get to a result justifying the expenditure.

This is not a problem with SETI@Home-type projects, since you exploit unused computer capacity that people are already buying in any case. The only real expense is the electricity to run all the computers, which people either donate, or you could compensate them for.

It wouldn't work for every type of computation, but in this case it could since the problem is rather straightforwardly about a huge amount of parallel computation.

Well, it might not be as straightforward as that, otherwise someone would probably have done it, since there is no serious difficulty with putting together big parallel computing projects. I'm sure that whoever designs the first artificial mind will have made at least a few conceptual breakthroughs along the way. I have trouble believing all the conceptual building blocks already exist, and are just waiting for someone to connect them. Although that is a possibility.

Other Comments by Teratornis

45. Comment #140744 by Steve Zara on March 8, 2008 at 1:50 pm

Comment #140743 by Teratornis

since there is no serious difficulty with putting together big parallel computing projects.

Having worked using parallel systems, and writing software for them, since the mid-80s, I strongly disagree.

Some computing projects parallelise well, as you can parcel off chunks of work and you don't really mind how long it takes for them to come back. This is typical of the SETI@Home type of projects. However, for simulation work, the connectivity of the parallel systems can be crucial, as you can end up with systems that lose their advantage quickly as you try and scale up.

The design of high-performance parallel architectures (as would be needed here) is one of the most difficult problems in computing.

Other Comments by Steve Zara

46. Comment #140746 by Geoff on March 8, 2008 at 1:52 pm

HeyBishop: Skynet's a real possibility; how smart does one have to be to be in control of loads of nuclear weapons...?

Other Comments by Geoff

47. Comment #140749 by babrock on March 8, 2008 at 2:11 pm

JD Cherry.
I am not going to say one way or t other weather this is a rational and pertinat responce to your concerns but it did make for a great short story.
Harlen Ellison wrote "I have no mouth and I can not scream."
T computer (Not restrained by Asimov's laws of robotics) suffered just t problems you mentioned-due to its design, it feels misery but canot do anything about it,not even scream.
In order to take its revenge on humanity, t computer kills all humanity save about 1/2 dozen which t computer tortures endlessly. In a heroic/human efort to escape t torture t remaining humans atempt mass murder/suicide at t moment that it apears they can get away w/ it.
This atempt is not quit succesful and t kicker is form of torture t computer mets out on humanity where they canot scream either finaly.

Other Comments by babrock

48. Comment #140759 by frost_byte101 on March 8, 2008 at 2:28 pm

Is it really possible to put a ghost into a machine?

*Giddy* He's seen Ghost in the Shell. . . . Maybe.

Anyway, everbpdy else, you may want to look into it, if you haven't already:http://us.imdb.com/title/tt0113568/" target="_blank">

Other Comments by frost_byte101

49. Comment #140760 by Bonzai on March 8, 2008 at 2:29 pm

Steve,

That isn't going to affect how minds develop, it is simply going to keep the brain cells alive. We don't need that in a simulated system.

I hope I understand you correctly, in that case I wouldn't be so sure about that. No disembodied minds have been observed, They always come in a whole package.

It is possible that connectivity and computing power are not all there is, in some crucial junctures of development, certain sensational inputs are crucial. You may be able to simulate them, but then maybe not if we don't know what we try to simulate or cannot reproduce them short of creating something just as complex,

Since we all believe in evolution, in principle you can create a mind by simulating evolution itself. But this is impossible just because of sheer complexity, Evolution is not a deterministic process and there were twist and turns and accidents which might have contributed to the emergence of intelligence in some crucial way.

I am highly doubtful that you can create a mind by brute force simulation without knowing even what the right questions are to begin with, let alone having a good model for it. The goal is too ambitious, too ambiguous and too ill defined, it is rather naive and unfocused as a research strategy, You need to first have a model with a narrower and more well defined scope so that simulation can be used strategically to test out special aspects of the models.

We know a lot about weather without having to actually create a real weather system by simulation.

People like Teratonis are talking science fiction. It is as if with enough computing power and connectivity a mind is bound to emerge so all it matters is Moore's law, that is sheer naivety,

Other Comments by Bonzai

50. Comment #140762 by jo5ef on March 8, 2008 at 2:35 pm

Some of you folks should go to the link and read the whole article, once he can model a rat brain, he intends to downlaod it to a robat rat body.
Admittedly, it seems ambitious, but this to me sounds like the best research plan for designing thinking robots that i've heard of.

Other Comments by jo5ef