Virtual Neurons Acting Like the Real Thing - The Blue Brain Project | Hacker News Reader

Virtual Neurons Acting Like the Real Thing - The Blue Brain Project

singularityhub.com

26 pointskkleiner24 comments

As far as I remember, they were recreating the network based on pure statistical properties. It's completely possible, if not likely, that the network they are simulating lacks the actual properties that give our brain its capabilities.

I see a cargo cult and a lot of hype.

It doesn't matter exactly how the are simulating it. They can test it and the real thing and if exhaustive testing shows that they behave identically, then they are functionally identical.

Obviously, if the simulation says "daddy? where am i?", then they've succeeded.

What I argue is that the method they are using is like putting monkeys to bang on a typewriter, but they are promoting it as "the next Shakespeare".

I agree with you in principle, however, their simulations are approximations, not identical.

From their website FAQ:

Q: Do you believe a computer can ever be an exact simulation of the human brain?

A: This is neither likely nor necessary. It will be very difficult because, in the brain, every molecule is a powerful computer and we would need to simulate the structure and function of trillions upon trillions of molecules as well as all the rules that govern how they interact. You would literally need computers that are trillions of times bigger and faster than anything existing today.

Because this will only be an approximation, it means that certain hard-to-define grammatical properties of spike-trains (e.g. "neural codes" as opposed to "rate codes") may be completely lacking in the simulation. Do these "neural codes" exist? At this point we do not know.

If it turns out that the brain does utilize complex neural codes, and that they are not being accounted for in the simulation, then the simulation can be no more than gibberish which happens to share many low-order statistical features with the real thing.

To illustrate this point, here are some examples of pseudo-English sentences automatically generated by a purely statistical model (n-gram model where n=2):

Richard Beesemyers, formerly raised sagging candidate to the Friday officially forecast at the project.

A hearing appeals.

That bank handles most notable exceptions to buy time for reconsideration Wednesday in decades, is the B& Coopers, to discuss international theme for the violin away as the wages of the power plant near classic chemise.

It could have been observing the 65 to back Tuesday.

And fourth, there is a full report, Wagner said.

Behave identically down to the exact pattern of neurons firing at the same times? Then sure, though thermal noise makes this impossible. But suppose they only behave "identically" in the sense that, say, the average firing frequency is the same across the whole net. Then there's no guarantee you're capturing the information-processing. This stuff that I've read tends to be heavily overhyped.

I don't see a whole lot of hype. As for the AI part, I think you're just rehashing Searle's Chinese Room argument, which has been shot down a long time ago.

What I have always wondered about this project is how they expect to deal with the huge number of parameters associated with the model. Each neuron has several parameters that control its response, and the neurons can be connected in a huge number of ways, even if restricted to small-world networks.

This is related to the bias-variance dilemma in machine learning; the larger the number of parameters the larger the variance in the models that fit the behavior. A nice recent article on the limitations of reverse engineering and the brain is http://frontiersin.org/computationalneuroscience/paper/10.33...

Without question, that's incredible. Little steps forward and we may just have complete technological simulacrums of life, would these not also be considered life as well?

>would these not also be considered life as well?

Is a simulation of life considered life? Good question.

There's not fast and hard definition of life, like there is with many mathematical concepts.

The line between crystals and viruses and bacteria and higher order life is very blurry.

Because it's so blurry it will be hard to draw a sharp crisp line between old life and any new man created life.

This is a great explanation. Perhaps new 'forms' of life or life-like processes such as this might end up helping to narrow down what life is truly defined as.

Isn't it generally true that the more variety there is of something, the easier classification becomes?

I think so. The blurry definitions of life tend to involve self organization, propagation, increase of complexity/local reduction of entropy, the problem is a lot of crystals meet that definition. Remove the constraint of mass and a lot of memes meet the definition of life, religion is a virus and all that.

Perhaps digital life will help us come up with better definitions and greater understanding.

If you asked a simulated brain how it felt, would it answer any different than the brain being simulated?

If it can self-replicate (without humans having to manufacture hardware for it), yes.

I don't think "self-replication" is the key of life as our empathy would understand it.

Killing a sterile person doesn't deserve punishment?

Agreed. The only reason we associate life with self-replicating is because living things don't last long without it, but it doesn't mean you need self-replication to be alive.

I think a digital reproduction of biological functionality is more likely and will probably arrive sooner then the long term extension of biological life.

Which presents the interesting possibility that some day you may have the chance to download your brain into a virtual one. But your biological self, flesh and bones and brains and identity and all, don't go anywhere, so in a way you're still going to die... but also live on.

"you're still going to die... but also live on."

No, you are still going to die, full stop. Some other thing, that shares certain characteristics with you, will live on.

This is no different than saying that you "live on" through your kids, or through your art or writing. This "thing" might share more in common with you than your kids or artifacts that you leave behind, but you are still dead.

Think about it this way. The potassium from the potato chips you had last week is now the memory of your 10th birthday party in your brain.

Quite literally none of the matter that was you 10 years ago is still part of you. Even the human skeleton is replaced on average every 9 years, faster in children slower in older people.

Are you still you? If so what makes you you? Is it the sum total of mass + thoughts? What if you change your mind, are you different person? What if you piss out the water was you yesterday? In 10 years?

Your doppelganger walks in the door just now. By asking probing questions, you establish he has a perfect copy of your memories and experiences, in every detail. He laughs at the same jokes, and cries at the same point of sappy movies you do (even though you never told anyone this). You can not discern any difference. Furthermore, he (apologies if you are female, just global replace the pronoun) has been engineered in a way that he will not age, or forget, or degrade in any way over time the way you will. Given the awkwardness of there being two people with one identity, he carefully explains that it only makes sense that he goes on as the "real you." To facilitate the process, he hands you a loaded gun, and tells you that you know what needs to be done and leaves the room.

Do you put the gun up to your head and pull the trigger? If not, why not? Surely, the fact of your memories going on is enough to assure you that you are not really "dying," just making it easier to resolve this complex situation?

Well, insofar as the doppelganger and I have sepata experiences, our personalities have already begun to diverge. Now, if the doppelganger whipped out a pair of EEG helmets and demonstrated that we could sync our experiences - eg, I take a sleeping pill and let my doppelganger go out on the town for a night, but I can easily 'catch up' the following morning - then i'd think about it.

I'd ask him to wait until I'm on my death bed and it's clear my death is unavoidable. Then I'd give him the thumbs up, instruct him to keep it real.

I'd argue that it matters whether the replacement is gradual or immediate.

There is really no difference for other people if the copy is perfect, but there is a huge difference for me.

Let's say I am able to make a perfect digital copy of myself, and immediately after I die. Well, my identity as a person has not in the least bit been affected because I have a perfect digital copy of myself. But the copy of my identity that was destroyed will never experience anything ever again, and that to me is the same as death.

However if the process is gradual, and the original copy of my identity experiences the process, then I would not view that as death.

Compare my identity to a computer program that you want to be moved to another location. You can either make a copy, and then delete the original, or you could copy one small piece at a time and update the file locations so that the program can be running while file locations are being changed. The only difference for you is that the second is probably a lot more work. But I'd argue that no complete copy was deleted in the second case. And if the program is sentient, I see this as being drastically different for that program.

It should be easy to fund this.

Have a "sponsor a synapse" program.

At $1 / synapse / year you can be listed as the benefactor of a synapse!! Cool!! Since there are approx 10^14 synapses (or something like that), that should be plenty of money.

Hell it could be $1 / synapse / century.

So if My_Estate sponsors synapse _S, then I pay $1 for the maintenance and upkeep of synapse _S for a 100 years. That is 1 cent per year.With 100 trillion synapses, that is 100 trillion per century, or 1 trillion per year. Which comes to an average of $200 per person on the planet (with 5 billion people on the planet). So some people and governments would sponsor more synapses than other people and governments obviously.

But if we all came together like this, we could very well fund this, at 1 trillion per year (100 trillion per century) paying 1 cent per year per synapse (1 dollar per century per synapse). Small price to pay to create a new form of consciousness, and perhaps a new era of prosperity for humankind in my opinion.

Reply on news.ycombinator.com