Jul 23 2009
Artificial Brain in 10 Years?
I have written previously about the various attempts to reverse engineer the brain and to develop artificial intelligence (AI). This is an exciting area of research. On the one hand researchers are trying to model the working of a mammalian brain, eventually a human brain, down to the tiniest detail. On the other, researchers are also trying to build an AI – either in hardware or virtually in software.
In the middle are attempts at interfacing brains and computer chips. Remember the monkeys who can move robot arms with their minds?
All these efforts are synergistic – modeling the mammalian brain will help AI researchers build their AI, and building AI computers and applications can teach us about brain function. The more we learn about both, the easier it will be to interface them.
However, it is very difficult to say how far away specific milestones and applications are on all these fronts. Progress is steady and promising, but predicting the future is hard.
That has not stopped Henry Markram from predicting at the current TED Global Conference that we can have a virtual model of the human brain in 10 years.
“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.”
I hope he’s right – that would be a huge milestone. Markram works on the Blue Brain project, and right now they have managed to build a virtual model of a rat cortical column. This simulates about 10,000 neurons. Markram says that each neuron requires the processing power of a laptop to model, so they use an IBM Bluegene machine which has 10,000 processors. This is definitely not a desktop application.
And that’s just one cortical column. Modelling the entire brain would require 100 billion neurons – that gives you an idea of the processing power of the human brain. That’s 10 million supercomputers.
It seems to me that hardware is going to be the biggest limiting factor on Markram’s prediction. If we assume Moore’s Law of doubling processor power every 18 months, then it will take 36 years for supercomputers to exceed the power of 100 billion neurons.
Of course, simulating virtual neurons requires much more processing power than building artifical neurons directly – building a massively parallel processor to duplicate a brain in hardware, rather than creating a virtual brain. But that is also hard to predict because it requires the development of new hardware – not just software and information.
While I share Markram’s enthusiasm for this technology, and his optimism that all the components will eventually come into place (modeling the human brain and developing fast enough computers) – 10 years seems to be pushing it. I hope to be proven wrong.
19 Responses to “Artificial Brain in 10 Years?”
Leave a Reply
You must be logged in to post a comment.
I’m still curious what the actual benefit will be of having another ‘black’ box that mimics our ‘gray’ box. It’ll probably be just as complex and mysterious as what we already know from our own brains. Or am I missing something? What could we learn from it? That enough ‘neurons’ will give us a brain? Don’t we know that already?
Danny,
Actually, I think we will learn a ton. First of all, raw numbers of neurons is not enough, and does not a brain make. We will have to figure out the structure of the brain – how the neurons connect together.
Also – then we have a working model of the brain. We could alter or even turn off specific subsystems and see what happens. This could prove to be an invaluable research model of the brain.
What about the ethical implications? What happens when you “turn on” the virtual human brain? Will it feel pain? Will it be torture to poke around in it while it’s “running”?
Also, building a virtual brain is nice, but how will it interact with the world? Will it exist in a virtual world? Or will we hook up cameras and speakers to it? While it get tired and need to sleep? Will it try to defend itself when a scientist reaches for the off switch?
This is some heavy sci-fi stuff here!
Can you build in the fear of death or destruction for failure to use short term thinking appropriately? Will you have to lie to it in the process, such as telling it that there are other brains out there that make it part of a kinship or other social system? Can you restrict it’s conscious awareness of 90% of its thought processes? Can you assuage it’s hunger and/or build in a distaste for broccoli? Will it have dreams or nightmares as part of a self-regenerating process? Will it yearn for a larger structure to justify the functional nature of its strategic goals? Will it be able to do better with some psychotropic assistance?
All excellent questions – won’t it be fascinating to find out?
The ethical considerations are real. I don’t think we can answer them yet.
Reminds me of Isaac Asimov’s R. Daneel Olivaw.
There’s almost enough evidence here to get a search warrant for Dr. Novella’s basement. You have to wonder what’s going on down there.
It’s always seductive to think we could have our own robot buddies just around the corner. I’d add a few more decades, maybe shortening it if they get some realistic animal-like AIs.
Of course, one note I’d like to make is that AI abuse is one of my berserk buttons. I grew up watching a lot of shows and movies where the robot just wanted to live a happy life, but was treated like a slave or a weapon, not a person. I don’t expect it to become relevant during my lifetime, though.
These questions and concerns sound like a lobe-finned fish training for a marathon to me. I think that first we should develop the technology to the point where such questions become plausible before we start to ask them. Before those questions become important there’s still a long and interesting series of developments to make, and it’s likely too that that path is going to give us the tools that we need to better ask and understand those questions.
A multi thousand-node artificial neural network does not a consciousness make.
A couple of points. First, I would hope that thier software gets a bit more efficient in the next 36 years. Secondly, it is not clear to me that their computing requirements will scale linearly with the number of neurons. Considering the amount of interconnectedness in a brain, it seems that doubling the number of neurons may more than double the computational requirements. Thirdly, is real-time emulation really a hard and fast requirement with version 1 of the virtual brain?
In any case, 36 years is not a very long time to wait, if eternal youth is at the other end. I, for one, am looking forward to having my very own shiny robotic carapace and laser beam eyes!
-The ethical considerations are real. I don’t think we can answer them yet.-
We won’t be able to answer them until technically it’s too late.
Of course the first thing we should do once we know what we are doing is NOT build a human brain. We should build something better.
I look forward to my cyborg implants.
The real question is – what will the computer brains build? And what will what they build, build?
yes – the 36 years assumes linear requirements, no change in efficiency. Actually, I think the current cortical column is at 1/10 real time simulation, but we can let that slide.
36 years is therefore a gross estimate, but was meant only as a rough evaluation of the 10 years prediction – which seems overly optimistic.
If they make a brain, out of like, a bunch of supercomputers, I hope they store them in a building shaped like a head. Bad ass.
I was recently listening to The Brain Science Podcast with Dr Ginger Campbell , and she was interviewing Dr. Eve Marder a Neuroscientist at Brandeis University. Dr Marder stated that she and hundreds of other scientists had been studying the structure and function of the stomatogastric ganglion of the lobster, which contains only 30 neurons, for 35 years, and still they were finding out new things about it.
Given that, it seems like a virtual model of a human brain in only 10 years would be quite a leap (quantum?)
Markham stated at the TED Global Conference in Oxford, England that a synthetic brain could be used to develop treatment for mental illness. However, artificial intelligence can only be useful to study “artificial” illness. A model always falls short of the real human illness. This is similar to the hopes of scientists who study rodent “schizophrenia” to understand a human diagnostic category.
stompsfrogs’ concept of Daneel Robot Olivaw as a cousin of Markham’s model is fascinating for medicine. Asimov’s character from the 1950′s was capable of conducting cerebroanalysis on humans. This analysis was a fictional interpretation of electromagnetic fields of living brain cells. Asimov was predicting a portable Magnetic Resonance Imaging system years ahead of its real time use in medicine. Imagine a portable Hitachi AIRIS robot solving crimes!
Simulating 10,000 neurons is certainly impressive. Simulating 100 billion (with the equivalent of 10 million current technology supercomputers) is not feasible.
To truly “simulate” a human brain one would need new hardware technology capable of scaling to such numbers. But if such hardware technology existed, we’d face a new (and to me, very fascinating) question. Specifically, one would have to ask, “Is this collection of 100 billion specially designed ‘artificial’ neurons a ‘simulation’? Or is this an actual, sentient being with a sense of self?” Would it be able to pass a Turning test? And if it did, what conclusion could we draw from that?
Almost 30 years ago, Douglas Hofstadter mused that we could never build a machine that was truly as intelligent as a human being that we could ALSO harness to solve hard problems night and day, day after day–because it would get bored. If such a machine completely mimicked human intelligence, including its sense of self, it would also mimic becoming bored with repetitive tasks.
Jonah Lehrer over at science blogs had an interesting post on this where he surmises that the ’10 years’ quote is not to be taken literally. You can find it at http://scienceblogs.com/cortex/
As far as the other ethical questions, I don’t know if it would feel pain as, if I remember correctly, the brain itself does not feel pain. Also, I would not see hunger being an issue due to the fact that it is constantly being ‘fed’ electricity, it’s nutrients.
I do however, think that until we unlock the secret of consciousness, this will unfortunately, all be a flight of fancy and fun speculation.
Theres these argument of the ethical implications of creating an artificial intelegence, such as: would it feel pain, or a desire to sustain it’s existence. But, arn’t these questions irrelevant? Pain is something that life evolved as a response to outside stimulai which was potentialy harmful to our existence. There was an evolutionary drive to equipe us with pain which was beneficial to the passing on of our genes. Why should an artificial intelegence have such a mechanism unless it was programmed into it originally?
I would argue that we shouldn’t be necessarily framing our thinking of how or what this AI would feel or think from our own biological point of view.
Also, and i say this with caution as I’m no expert in the field of neurological compution (i think thats the field!) but it seems that the method of modelling all these individual neurons using massive amounts of processing power is a little like trying to hack a password by going through all the possible different combinations of letters, numbers and symbols. Isn’t it possible that we just need to throw away our old style of computing to solve this problem? I may be talking out my ass on that one, i know its probably a lot harder than i make it sound!