By 
Is an artificial human brain only 10 years away? Prof Henry Markram, director of the Blue Brain Project, says yes:
“It is not impossible to build a human brain and we can do it in 10 years,” he said.
“And if we do succeed, we will send a hologram to TED to talk.”
They’ve already simulated elements of a rat brain and are working their way up to humans. They’re using supercomputers with 10,000 processors to simulate and reverse engineer the brain.
I highly doubt an artificial brain will be created in 10 years. It would be great if they can, but it doesn’t seem feasible — the human brain is far more complicated than a rat brain, and they don’t even have that simulated yet. My guess would be at least 20 years. More realistically, 50-100 years.
Hopefully I’m wrong though and it’ll be quicker.
It’ll be interesting when we can download/duplicate all the information in our brains to an artificial one. We could live in a virtual environment when we die… until someone pulls the plug.
Do you think the prospects for this are exciting or frightening (or both)?
Entropist: These idiots are wrong on so many different levels, its hard to... 
Just a Guy: Me an my wife have negotiated infidelity. Meaning she can be with other guys as long as she doesn’t hide... 
Gwenny: Absolutely not! The world has plenty of people having kids. A child consumes HUGE amounts of resources, which from both a... 
Siberia: Or maybe they know they have neither patience nor inclination to deal with the fruit of their loins and thus, rather than be... 
DDM: Repubs and Dems. He’s saying they only take first impressions and... 
CM: Good point, most religions encourage having a lot of children. From an evolutionary standpoint you sort of owe it to your... 
Jasowah: “They don’t go beyond those emotions, and that’s the... 
27 Comments
Ah, the wonders of science journalism. ‘Creating an artificial human brain’ as in simulating one in software is highly plausible and 10 years sounds about right. Given the mention of rat brain simulation and animal testing I suspect that this is what the researchers mean by ‘artificial brain’ and that someone in the journalistic chain got confused or deliberately used misleading language.
Oh, I should add; I suspect that the ‘complexity’ of the human brain in comparison to rat brain is not related to the difficulty of simulating it, most likely you’re just going to need a proportionally larger amount of computational resources.
Based on the quotes (which I’m suspecting are at least somewhat out of context) I would assume that the real difficulty is the actual ‘reverse engineering’. I can’t find any specific details on how this is done, though wikipedia does say “Markram mapped the types of neurons and their connections”.
Maybe we can be downloaded into a robot and go in space without a spacesuit. But then we could be infected by computer viruses.
There’s always a catch, isnt it?
I think the problem is twofold: the first problem is to model the human brain, the second is decent simulation time. If you can simulate one second of brain activity in a week-long simulation, then the model will be only useful in analyzing reflexes and instinctive behavior. In 10 years we will have the technology to drastically reduce simulation time, since the “supercomputer” they’re using today will be probably considered just an high-end PC in 10 years.
We could use the Internet to distribute the simulation of brain activity over thousands of computers. It could be like “Skynet”.
Oh, wait…
I think it’s a great plot for a SciFi story. As to being factual? Yeah, sure, and we were supposed to have flying cars by the late 70’s. We can’t even get agreement on solar power plants, but we can claim that there will be an artificial human brain in 10 years.
Sorry, the skeptic in me is having a really hard time with this story. I work in the technologies industry and have been a computer systems engineer for more than 20 years now. As good as we are and as quickly as the equipment changes, no, not in 10 years, or 20, or 50. The human brain is not a CPU/Storage system. It is a functioning organ of the body, the most complex organ we have. Science does not even understand clearly how we see and hear, much less think, reason, and store our memories.
Great SciFi story, lousy reporting.
It was a great sci fi story in 1930.
Anyone who follows the field knows that this plot is done to death.
Isn’t the real question, “Do you want the blue pill or the red pill?”
You could never “download” your consciousness into a computer. If anything an exact clone of you, with all of your experiences and personality would be created, but you would still be dead. For all intents and purposes to everyone else, you would be around still. Its like making a cd, and then copying it, the information on each cd is identical, but there are two cds, not the “transfer” of data from one to another but the copy and pasting thereof.
Are they taking advance orders?
Moan….
Braaaains!
Braaaaaaains!
DOWN, Zombie Jesus! DOWN!
Scientifically, this is exciting.
But I have qualms.
I think tums can help with that.
I’m with Proto on this one, there is far too little information in the snippet to be either excited or frightened. I do think that as computer modeling of neural networks and such advances that there should be some debate as to how such technology should be used. Considering how often my laptop crashes, I wouldn’t put my faith in living forever in a computer.
Well, computing power is likely to be the least of the problems. If Moore’s law (cpu power doubling every two years) continues, we’ll have a 1000-fold increase in cpu power every 20 years.
Various scanning technologies will also be drastically improved in the next decades, making it easier to map the brain in detail (its biological configuration but primarily the electrical activity, which is probably more interesting). I doubt very much this will enable us to map the human brain to the point where we can duplicate it, though. Biology is terribly messy.
mode = armchair philosopher
There are some interesting ethical problems with simulating a human brain. Assuming that the simulation is accurate, then is the simulation a person? If personhood is a consequence of the ongoing computational process in the brain, then the simulation is a person. Even if the simulation doesn’t exactly simulate a human brain, it may still be a person.
mode = computer scientist
I’m not convinced that we can simulate an organic brain of any species on a digital computer. If neurons act a digital switches or have only a finite number of distinct activation levels, we can simulate them on a digital computer. If however neurons have an analog component, then we can’t accurately model them on a digital computer.
I also suspect that our cognitive processes are directly influenced by being embodied. This would lead to the consequence that the person being simulated would not be a human, making is bit difficult to transfer our consciousness to a computer.
I suppose that the prospect is frighteningly exciting.
I also suspect that our cognitive processes are directly influenced by being embodied.
My personal pet theory is that the experience of consciousness is due in part to the brain being a highly asynchronous parallel processor, due more than anything else to the fact that it takes a non-trivial amount of time for a signal to get from one end of the brain to the other. If that is the case, it should be (at least in theory) possible to simulate the asynchrony using synchronous processors (like most modern processors).
Artificial brain the size of a human does not mean that it will be like a human. At least not in real-time. With the tecnology in 10 years we probably could simulate a 1 second brain activity but have to calculate a few weeks to simulate that one second.
If you would like to read about the possibilities of recording all your experiences and being able to be downloaded to a new body, I would recommend Richard Morgan’s SF books about Takeshi Kovacs; Altered Carbon, Broken Angels and Woken Furies. Morgan also has the good taste of speculating about how this technology affected the society, in quite realistic terms. Are you really the person you claim you are?
The books are not to everyone’s taste, they are very violent and the author regularly try to chock the reader in one way or the other. But I never miss an opportunity to recommend them to anyone I meet…
We have “laws” for everything; Murphy’s Law, Goodwin’s Law, Moore’s Law – there ought to be a law for this as well: If you predict that a great breakthrough in science or society will happen within ten years, then it will never happen.
That is, if you predict it to happen within nine years or eleven years, it might happen, but never if the prediction is ten. Similar to Pratchett’s “one in a million chances happens in nine cases out of ten”.
I’ve enjoyed Altered Carbon, Broken Angels and Woken Furies, but to call them very violent is a little misleading. To say they occasionally make A Clockwork Orange seem like a happy book is closer to the mark.
Lol :D
I like them already. [/bloodthirsty mode]
What I would find most interesting is whether they can teach a machine to be irrational. If not, then they can’t duplicate humans. And if they *do* make an irrational machine… ooh, watch out!
Seriously, we are only beginning to discover why and how humans think what we think, it’s going to be another hundred years before we even really know what it means to ‘think like a human’.
How comes that, suddenly, everyone is an expert in the matter? What entitles you to assume you know better than the guy that is researching this for 15 years?
The project has, to the date, simulated 200 different types of neurons working as neocortical columns and functioning as elements of rat brains. As the main page of the project states, they have proven the feasibility of the simulation and plan to move forward on more detailed simulation.
BTW,
- The simulator takes “two orders of magnitude” more time than it’s biological counterpart (1:100) and they are working to down that to real time.
I’ve been following the Blue Brain project for a couple of years now, am a computer science student/programmer, and have been following science since I was a kid, and since there seem to be some questions I feel like maybe I can shed some (extremely unqualified, despite that long-winded introduction) light.
What they’re doing in the BB project is modeling a brain up neuron-by-neuron, at very high resolution. Like, they’ve basically written a program that simulates a neuron — receiving neural impulses, transmitting, generating proteins from its DNA, changing the way it operates based on virtual hormones — and run it several thousand times in parallel. My understanding is that they do have to specialize a few different kinds of neurons and such, but that’s the basic idea.
The reason they’re fairly confident about being able to simulate a whole human brain in 10 years is because they’re not trying to model a brain, they’re trying to model a few billion neurons in basically the structure of the brain. The problem is highly parallelizable and unlikely to grow significantly more complex. Basically, throw an extra few billion processor cycles at what they’ve already got and you’ve got a human brain. Not really, but, you know.
Also, based on that, I’m sure you can imagine that this project has basically no hope of providing a decent download host for our thoughts. Not that you were suggesting that. The main point is just to be able to have a brain, apply some voltage to some part of it, and be able to watch — in 4-d — as it cascades around.
And of course speaking to the Rev PJ: that’s why everybody’s so freakin hopped up excited about memristors. But also there’s no such thing as analog when you’re talking about discrete objects like atoms and molecules and electrons. It might seem like there are smooth functions when you look at the evaporation on the ocean surface, but if you had enough computing power you could model it perfectly. Same goes for the neurons of the brain: they are made up of atoms and electrons, so.
The memristor article is definitely interesting. I have a suspicion that “Cortical Computing” may be very exciting and ultimately not a Turing machine (which would be exciting itself). While we can in theory model neurons, or any other analog system, to any desired degree of precision we pa a substantial price in performance. I would love to have something analogous to human cognition on a Turing machine but I don’t believe it is a tractable problem. Change the model of computation, say with a machine that uses memristors for computation somehow, and we may be able to have a silicon-based intelligence.
My own research would benefit from an electronic analog of the human perceptual system for spotting large scale structures. The memristor may allow us to build some exciting pattern recognition systems.