How much better than a human can an artificial one be?

[MrDakka]

Not to get too off-topic, but this point is a good one. One thing that has always bothered me about the Ray Kurzweil school of transhumanist thought is the incredibly simplistic assumptions about the development of complex biological, psychological, and technological systems.

Glad to see I'm not the only one who thinks this predictions are rather simplistic in addition to being way too optimistic.

[Ariphaos]

Not to get too off-topic, but this point is a good one. One thing that has always bothered me about the Ray Kurzweil school of transhumanist thought is the incredibly simplistic assumptions about the development of complex biological, psychological, and technological systems.

Most of my peers that I've conversed with on the subject tended to feel that Kurzweil is ridiculously optimistic about the timetable involved, but there isn't a problem that understanding the human brain presents that is unsolvable.

That said, there's a difference between full-up simulating the human brain, and getting a computer program to have comprehension of a core set of concepts that would enable it to expand that core set on its own.

[cosmicalstorm]

Maybe I should add that I'm no big fan of Kurzweils schedule either. More likely, by the late 2020's we'll have a synthetic designer virus kill a third of the population or half of the oceanic phytoplankton and we'll be back on square one. Or if we do design some kind of AI, its probably just going to kill or ignore us.

[Ziggy Stardust]

Most of my peers that I've conversed with on the subject tended to feel that Kurzweil is ridiculously optimistic about the timetable involved, but there isn't a problem that understanding the human brain presents that is unsolvable.

Indeed. It's not that I think it is particularly unsolvable, but that the limitations are more troublesome than many would like to admit.

That said, there's a difference between full-up simulating the human brain, and getting a computer program to have comprehension of a core set of concepts that would enable it to expand that core set on its own.

That's a good way to put it.

My general feeling isn't that it won't be possible to simulate the human brain, but that the level of sophistication we would need to do so technological is such that there wouldn't be much of a point to doing it except for it's one sake. That is, if our technology level is that good, the computer programs we can make without shoe-horning them into a human-like architecture would be far superior anyway. So that by the time we CAN do it, we will do it just to prove we could.

[someone_else]

I will believe it when I see it. Has anyone managed to do it so far ? No. Untill then it is just fanxiful speculation like. Warp drives and time travel.

With this nonsense you can disbelieve anything from fire to modern tech, so I'd call it invalid reasoning. You know, before anyone managed to do it, none managed to do it (by definition), so IT WILL NEVER HAPPEN according to your broken logic.

For your knowledge, time travel and FTL (which are more or less the same thing) tend to violate stuff we know is true to very high degree. So they aren't called bullshit without a good reason.

That is, if our technology level is that good, the computer programs we can make without shoe-horning them into a human-like architecture would be far superior anyway. So that by the time we CAN do it, we will do it just to prove we could.

what about doing it as an effort to understand better our own brain? I mean, having a human-like brain to play with would be useful. Say we start with pieces like that simulation-ish of rat cortex, and eventually you build up the simulation until it starts to cross the line between simulation and "human" however you want to define it. Then hippies free them and they go in the wild and eventually get their rights recognized (or not and live on the run if you want a good cyberpunk setting).

[K. A. Pital]

I must admit that I'm sort of baffled by the attitude of people in the thread. Yes, perhaps simulations of the brain are not around the corner, because we need to understand it much better than we do now. However, there is no hard blocks on knowledge of material structures, and with the right instruments, et cetera, humanity will eventually figure it out.

In essence, to say that the human brain cannot be artificially reconstructed one must subscribe to a theistic notion of soul or something like "inviolable complexity" which humans can never master. That sounds far worse than Kurzweil's misguided optimism about the speed of transhumanist advancement. It implies capitulation before the fight even started.

[Ziggy Stardust]

I must admit that I'm sort of baffled by the attitude of people in the thread. Yes, perhaps simulations of the brain are not around the corner, because we need to understand it much better than we do now. However, there is no hard blocks on knowledge of material structures, and with the right instruments, et cetera, humanity will eventually figure it out.

In essence, to say that the human brain cannot be artificially reconstructed one must subscribe to a theistic notion of soul or something like "inviolable complexity" which humans can never master. That sounds far worse than Kurzweil's misguided optimism about the speed of transhumanist advancement. It implies capitulation before the fight even started.

On the other hand, to play Devil's advocate again, is it really any more logical to just wave your hands and say that one day we will reconstruct it because ... technology?

(To clarify, I don't think that it IS impossible, just that the challenges are greater than the transhumanist advocates usually choose to recognize. And it just bothers me when people make an argument that essentially amounts to nothing more than, "Technology is super cool, and I think it will one day be able to do X.")

[someone_else]

is it really any more logical to just wave your hands and say that one day we will reconstruct it because ... technology?

Maybe not totally logical, but necessary to advance further imho.
If you say "unless I find brick walls then I can do it" you are likely to try and then discover if you actually can in practice or if there are unknown issues.

If you say "if it isn't certain it can happen i cannot do it" you never start doing anything.

[Ziggy Stardust]

Maybe not totally logical, but necessary to advance further imho.
If you say "unless I find brick walls then I can do it" you are likely to try and then discover if you actually can in practice or if there are unknown issues.

If you say "if it isn't certain it can happen i cannot do it" you never start doing anything.

Well, yeah. I'm not saying we shouldn't TRY to model the brain. The potential advances and knowledge that could come out of failing to do so are monumental, nevermind if/when we actually succeed. My only real opposition is the same as if someone came in here and said we will build FTL engines because ... TECHNOLOGY! I guess that's a bad example because FTL is running against certain laws of physics that brain modeling would not, but you get my point.

[K. A. Pital]

FTL engines represent a hard limit with our current understanding of physics. The matter that composes the brain does not require breaking relativity or causality, neither extremely high energies that are beyond our generation capacity (which, for example, might be required to recreate some material phenomena in the universe). Hence why there's not much wrong with saying "Technology will be able to model a material object X one day".

I don't think it is super-easy.

[cosmicalstorm]

Watching Henry Markram's presentation of the blue brain project on Youtube, reading Steven Omohundro's papers on artificial intelligence. I have a very hard time seeing how the creation of self modifying minds will not play an extremely central role in our near future. When I come across people who scoff the concept of artificial intelligence I'm reminded of the (anecdotal?) gentlemen who thought everything had already been invented.
To me, the science of intelligence seems to find itself right now, in a similar position to where the science of nuclear physics found itself this time a hundred years ago, and the end result will be equally powerful. Intelligence is as real as radiation, and channeling its force is a matter of arranging matter in a very delicately careful way. If we are not extinct or set back by some massive disaster*, I don't think we need more than a hundred years before its done. The outcome won't necessarily be very positive for humans.

*Sadly, this seems fairly likely, IMHO.

[Guardsman Bass]

what about doing it as an effort to understand better our own brain? I mean, having a human-like brain to play with would be useful. Say we start with pieces like that simulation-ish of rat cortex, and eventually you build up the simulation until it starts to cross the line between simulation and "human" however you want to define it. Then hippies free them and they go in the wild and eventually get their rights recognized (or not and live on the run if you want a good cyberpunk setting).

You raised the ethical issue. If you run a good simulation of a human brain capable of responding to information fed to it, then shut the whole thing down, are you "killing" a "person"?

[Starglider]

If you run a good simulation of a human brain capable of responding to information fed to it, then shut the whole thing down, are you "killing" a "person"?

Technically you are suspending them. If you delete the simulation state file, then they're dead.

[Ariphaos]

On the other hand, to play Devil's advocate again, is it really any more logical to just wave your hands and say that one day we will reconstruct it because ... technology?

"We will" and "we can" are two entirely different concepts. Sarevok is proposing that the human brain works by magic of some sort. It's a laughable claim, and I suspect he's just trolling. Time travel and warp drive requires violating principles as we currently understand them. The human brain is possible as there are currently seven billion demonstrations.

(To clarify, I don't think that it IS impossible, just that the challenges are greater than the transhumanist advocates usually choose to recognize. And it just bothers me when people make an argument that essentially amounts to nothing more than, "Technology is super cool, and I think it will one day be able to do X.")

All programming is extending language. I'm going to write a few new functions tonight. Some of the functions I work on - that thousands of others are working on - actually build and extend the language's power as a language itself. While predicting far into the future is hard, it's much easier for me to say 'next year I will have X done', and X is something I'm writing solely for the purpose of building Y and Z on top of it. X wasn't possible five years ago, because T, U, V, and W needed to be ready first.

I consider myself a transhumanist, as denying that that sort of progression is occurring is simply denying reality.

[madd0ct0r]

Getting away from the brain - It seems I can pick any organ and 'design' a sci-fi one that does the same job but better.

The thing is, if you assembled a human from all these 2.0 organs, it'd be less then the sum of it's parts, as most of the things one organ does wouldn't be needed anymore due to everything else being upgraded.

[Guardsman Bass]

If you run a good simulation of a human brain capable of responding to information fed to it, then shut the whole thing down, are you "killing" a "person"?

Technically you are suspending them. If you delete the simulation state file, then they're dead.

Even that's ethically dubious. We don't allow people to forcibly keep other people under sedation, only to wake them up for specific tasks.

[K. A. Pital]

If you create a sentient software, you shouldn't "shut down it" and I doubt you would do it. Instead you would try to create a virtual space where this sentient program could develop further. Essentially, you gain more from trying to nurture this intelligence than from suspending it at will. By keeping it constantly running you could gather way more information than by switching it on and then off.

[Starglider]

If you create a sentient software, you shouldn't "shut down it" and I doubt you would do it. Instead you would try to create a virtual space where this sentient program could develop further. Essentially, you gain more from trying to nurture this intelligence than from suspending it at will. By keeping it constantly running you could gather way more information than by switching it on and then off.

Real general AI research, like all other programming, makes very heavy use of stop/edit/restart, pause/debug/resume and version control of both the source and the mind states. Leaving copies running is appropriate for A-life and some GA/GP and unsupervised learning experiments, but you still halt them when you deploy a new version. Ethical and safety concerns have frankly been rather slow to develop.

[K. A. Pital]

Until the machines really become sentient on a human level, the ethical issues are pretty much nonexistent. We kill animals routinely.

[someone_else]

You raised the ethical issue. If you run a good simulation of a human brain capable of responding to information fed to it, then shut the whole thing down, are you "killing" a "person"?

If the simulation is accurate enough to be called "person", yes. The fun part is how to determine when it crosses the line.

More than likely it will just depend from the laws in effect, just like for animals and abortion and staminal cell research.

Which in turn depends from the lobby of humans defending the programs, initially.

But note that I was saying brain research simulations could lead to a simulation so similar to become a fully-fledged person when they start working on the most important parts of the cortex, the decision-making areas.
Then that shit would happen and these synthetic brains could become citizens, or rebels in a darkish cyberpunk setting, hunted as monsters or somesuch.
I didn't say WE WILL CREATE FULLY FLEDGED HUMAN SIMULATIONS AND THEN TORTURE THEM FOR SCIENCE!!!!! MUHAHAHAHAHAH!!!!

The thing is, if you assembled a human from all these 2.0 organs, it'd be less then the sum of it's parts, as most of the things one organ does wouldn't be needed anymore due to everything else being upgraded.

Why? stuff will change but you still need the same basic functions every organ does to have a self-replicating autonomous biochemical machine.

Brain--> problem-solving to achieve primary goals
digestive system --> acquisition of power and building materials
immune system --> defence from inside and outside microscopic threats

and so on and so forth.

[Rossum]

Regarding human brain simulations: One could go the Hitchikers Guide to the Galaxy route and say that the only ethical problem with shutting down an intelligent mind is if the mind doesn't want to be shut down.

If you find a suicidal human who doesn't mind the thought of dying and create a simulation of their brain, then its possible that the simulation wouldn't mind getting shut down if need be. At the very least, its human enough to pass the turing test, self-aware enough to be declared legally able to make its own choices, and inclined to agree to sign documants letting you shut it down.

[Skgoa]

And it (he?) would know that it's not permanent. The great wangst-inducing thing about dying is the end of conciousness, at least it is for me. If I knew I could be rebooted later on, I wouldn't have a problem with being shut down. I expect suspicions of not being booted up again would go away after going through since often enough.

I must admit that I'm sort of baffled by the attitude of people in the thread. Yes, perhaps simulations of the brain are not around the corner, because we need to understand it much better than we do now. However, there is no hard blocks on knowledge of material structures, and with the right instruments, et cetera, humanity will eventually figure it out.

In essence, to say that the human brain cannot be artificially reconstructed one must subscribe to a theistic notion of soul or something like "inviolable complexity" which humans can never master. That sounds far worse than Kurzweil's misguided optimism about the speed of transhumanist advancement. It implies capitulation before the fight even started.

You might not realize this, but we are currently using encryption codes that would take longer to break than the universe existed. Complexity. Is. A. Bitch. There is absolutely no reason to expect simulating a brain to be feasible with the technology that we have right now or that is on the horizon. Especially since we now know that simulating a sufficient amount of neurons will not be enough, due to memories being encoded inside single neurons. This means we will have to go down to molecular level. Numerically simulating the laws of physics is computationally hard, especially since with that discovery the number of "actors" we have to simulate has increased by millions. From everything we know right now, we might have to simulate EVERY SINGLE ATOM in a brain to have a true representation. (Slight caveat: we might very well get away with mostly ignoring the cells' hull and axons.)
This is not just hard, this is riddiculously hard to solve numerically. Don't come crying when after billions of years your brain simulator just spits out "42".


edit: I honestly expect brains to be grown in vats before we fully simulate them in software. It's just that much easier.

[Starglider]

You might not realize this, but we are currently using encryption codes that would take longer to break than the universe existed. Complexity. Is. A. Bitch.

False analogy. Encryption is hard to crack because reversing the function requires search through a large space. That is not the same as algorithmic complexity; the irreducible complexity of all existing encryption functions is actually trivially small. Biological simulation does not involve any comparable brute force search; the closest equivalent in AI is genetic programming, and even that isn't directly comparable.

Especially since we now know that simulating a sufficient amount of neurons will not be enough, due to memories being encoded inside single neurons. This means we will have to go down to molecular level.

The article you linked (which is popsci level) states no such thing, unsurprisingly since it is at odds with the leading consensus in neuroscience. Rather it states that individual neurons are linked with specific memories in mouse brains. This is rather encouraging, because that kind of localisation makes tracing the functional role of neural wiring (and in reverse, the physical implementation of observed behavior) significantly easier. The best current research suggests that the signal processing functions implemented by neurons are nontrivial but only moderately complex, and duplicating their functional role requires duplication of the relatively gross electrochemical charateristics of synapses and dendrite trees (with some open questions about the influence of assorted cell biochemistry, but on the plus side that stuff happens on a much slower timescale and is hence computationally less demanding).

Numerically simulating the laws of physics is computationally hard, especially since with that discovery the number of "actors" we have to simulate has increased by millions.

On the contrary, every claim to date of meaningful computation at the molecular level including all the quantum consciousness nonsense has been thoroughly debunked. Unsurprisingly, as body temperature cell chemistry is a ridiculously noisy, transitory and unreliable medium and nervous systems only work at all by dealing with statistical aggregates at the neuron level or above.

edit: I honestly expect brains to be grown in vats before we fully simulate them in software. It's just that much easier.

Do you have a degree in AI and practical experience with biomorphic neural nets? If not then frankly your expectations have negligible predictive value.

[Ziggy Stardust]

nervous systems only work at all by dealing with statistical aggregates at the neuron level or above

Just as a side note, even this isn't entirely certain. There is a lot of work being done now that questions this model of neural processing. I need to try to find some abstracts/papers, I am just remembering this from some talks I went to a few weeks ago, but there was some really neat stuff about the neurological analogues of "attention".

[MrDakka]

Call me when I can upload my mind to a hardened substrate.

On a side note; if you can upload your mind to say a computer (with no EMP protection) and then you're hit with an EMP, what happens? Are you dead? In a coma?