Scientists build simulations of neocortical column
Aim at testing antidepressants, Alzheimer’s drugs
In his office in Lausanne, Switzerland, Dr. Henry Markham muses about how little doctors actually know about depression and the drugs they use to treat it: “You try a drug for depression. Two months later you come back and say, ‘Well, that didn’t really work. Let’s try another one.’ And for none of the diseases of the brain—and there are hundreds of them—do we know what it is really doing in the brain. What is the effect of giving you a specific drug? It’s not going to kill you. We know that because the FDA wouldn’t allow that. But basically we don’t really know at all.”
Markham is the lead scientist in an ambitious project to reverse-engineer the most complex part of the mammalian brain—the neocortical column (NCC). These IBM scientists working in Switzerland have already simulated a rat’s NCC as a proof of concept. The rat NCC has 10,000 neurons and 30 million synaptic connections based on 15 years of experimental data. They used software to reverse-engineer accurate 3-D models of those neurons and interconnections. By simulating input stimuli and comparing the behavior induced with those in experiments the researchers iteratively tested and refined the model until they were confident that it behaved like a real neocortical column.
The scientists and engineers plan to simulate the NCC for a cat next, then a primate, and finally a human being. The scale-up is huge: while the rat brain has 200 million neurons in all, the human brain has somewhere between 50 and 100 billion. There are those who are skeptical that the Blue Brain project can be completed in the ten- year period Markham and his team have set as their goal for the finish date. Once the model of the human brain is completed, scientists anticipate studying how the brain processes sensory inputs and how consciousness arises. They also expect to be able to understand disease processes, eventually at the molecular level.
Time is important to this project. Within ten years, Markham predicts, people will be living an extra ten years beyond what is considered a normal lifespan now. That means that 50 percent of the world’s population could have Alzheimer’s disease, unless work progresses rapidly to understand the characteristics and causes of that disease. Traditional science methods won’t work because the problem is too complex and too interactive: “You’re not going to develop a comprehensive understanding of the cause of these diseases and you’re not going to develop treatments unless you’re able to put them together and stimulate the interactions. “ For the trained neurologist, using simulation speeds up the process remarkably. Adds Markham, “I’ve learned more from studying the neocortical column simulation in the last three years than in the twenty years previous to that.”
How does the Blue Brain Project differ from traditional AI projects?
Artificial intelligence researchers want to create intelligent machines using digital hardware and software. In many ways it would seem that the triumph of Deep Blue, the IBM supercomputer that beat world chess champion Garry Karsparov in the late nineties, sounded the death knell for AI. Far from making decisions about the game as a human being would, Deep Blue performed—and won—by merit of brute force, systematically evaluating 200 million possible chess moves per second. It was a triumph for the programmers and the hardware designers but it did not compete with human intelligence in a way that would help people improve their lives.
Recently, a new wave of energy around AI wants to create co-processors for the brain, which would enhance or augment brain function in various ways. Ultimately they want to program the human brain as one would a microprocessor. Marvin Minsky wants the brain to be more like his iPhone in being able to download applications. “I would like to be able to download the ability to juggle. There’s nothing more boring than l earning to juggle.” In the shorter term—over the next five years— researchers want to develop a program that can read a children’s book, explain the gist of the story, and ask meaningful questions about it.
In comparison to Big Blue, the AI effort at MIT does indeed sound like child’s play. The Big Blue researchers want to understand how intelligence and perhaps even consciousness arises. They intend to use the simulation to shortcut clinical trials and someday use simulators like fMRI machines in hospitals to model individuals’ brain functions based on genetic and other data. This is part of the dream of a personalized medicine, where you can simulate, for example, the action of a drug in the brain to quickly and safely test its efficacy for an individual and explore various hypotheses about brain function. While the AI researchers stress the imitation of certain human activities, the IBM scientists emphasize simulations at the cellular and potentially at the molecular levels.
Post a comment
For more photos and text about Big Blue
http://bluebrain.epfl.ch/
To see what the MIT AI community is up to
http://web.mit.edu/newsoffice/2009/ai-overview-1207.html
Aim at testing antidepressants, Alzheimer’s drugs
In his office in Lausanne, Switzerland, Dr. Henry Markham muses about how little doctors actually know about depression and the drugs they use to treat it: “You try a drug for depression. Two months later you come back and say, ‘Well, that didn’t really work. Let’s try another one.’ And for none of the diseases of the brain—and there are hundreds of them—do we know what it is really doing in the brain. What is the effect of giving you a specific drug? It’s not going to kill you. We know that because the FDA wouldn’t allow that. But basically we don’t really know at all.”
Markham is the lead scientist in an ambitious project to reverse-engineer the most complex part of the mammalian brain—the neocortical column (NCC). These IBM scientists working in Switzerland have already simulated a rat’s NCC as a proof of concept. The rat NCC has 10,000 neurons and 30 million synaptic connections based on 15 years of experimental data. They used software to reverse-engineer accurate 3-D models of those neurons and interconnections. By simulating input stimuli and comparing the behavior induced with those in experiments the researchers iteratively tested and refined the model until they were confident that it behaved like a real neocortical column.
The scientists and engineers plan to simulate the NCC for a cat next, then a primate, and finally a human being. The scale-up is huge: while the rat brain has 200 million neurons in all, the human brain has somewhere between 50 and 100 billion. There are those who are skeptical that the Blue Brain project can be completed in the ten- year period Markham and his team have set as their goal for the finish date. Once the model of the human brain is completed, scientists anticipate studying how the brain processes sensory inputs and how consciousness arises. They also expect to be able to understand disease processes, eventually at the molecular level.
Time is important to this project. Within ten years, Markham predicts, people will be living an extra ten years beyond what is considered a normal lifespan now. That means that 50 percent of the world’s population could have Alzheimer’s disease, unless work progresses rapidly to understand the characteristics and causes of that disease. Traditional science methods won’t work because the problem is too complex and too interactive: “You’re not going to develop a comprehensive understanding of the cause of these diseases and you’re not going to develop treatments unless you’re able to put them together and stimulate the interactions. “ For the trained neurologist, using simulation speeds up the process remarkably. Adds Markham, “I’ve learned more from studying the neocortical column simulation in the last three years than in the twenty years previous to that.”
How does the Blue Brain Project differ from traditional AI projects?
Artificial intelligence researchers want to create intelligent machines using digital hardware and software. In many ways it would seem that the triumph of Deep Blue, the IBM supercomputer that beat world chess champion Garry Karsparov in the late nineties, sounded the death knell for AI. Far from making decisions about the game as a human being would, Deep Blue performed—and won—by merit of brute force, systematically evaluating 200 million possible chess moves per second. It was a triumph for the programmers and the hardware designers but it did not compete with human intelligence in a way that would help people improve their lives.
Recently, a new wave of energy around AI wants to create co-processors for the brain, which would enhance or augment brain function in various ways. Ultimately they want to program the human brain as one would a microprocessor. Marvin Minsky wants the brain to be more like his iPhone in being able to download applications. “I would like to be able to download the ability to juggle. There’s nothing more boring than l earning to juggle.” In the shorter term—over the next five years— researchers want to develop a program that can read a children’s book, explain the gist of the story, and ask meaningful questions about it.
In comparison to Big Blue, the AI effort at MIT does indeed sound like child’s play. The Big Blue researchers want to understand how intelligence and perhaps even consciousness arises. They intend to use the simulation to shortcut clinical trials and someday use simulators like fMRI machines in hospitals to model individuals’ brain functions based on genetic and other data. This is part of the dream of a personalized medicine, where you can simulate, for example, the action of a drug in the brain to quickly and safely test its efficacy for an individual and explore various hypotheses about brain function. While the AI researchers stress the imitation of certain human activities, the IBM scientists emphasize simulations at the cellular and potentially at the molecular levels.
Post a comment
For more photos and text about Big Blue
http://bluebrain.epfl.ch/
To see what the MIT AI community is up to
http://web.mit.edu/newsoffice/2009/ai-overview-1207.html
RSS Feed