Cybernetics
Today I attended a seminar in which the speaker (Dr. Krishna Shenoy of Stanford University) discussed his research which is aimed at integrating signals from human brains with prosthetic arms for those who have lost limbs or are paralyzed. Even he admitted the long term goals of his research are far into the future, but they seem quite achievable.
He began by discussing how they monitor motor neurons in the brain which control arm movements. They've implanted very small electrode arrays into very specific regions of the brain which are responsible for motor control of the arms. They then spent many many months training the monkeys to perform this basic task: A square appears on a screen in one of 8 possible locations. The monkey touches the square. A second, smaller square appears in another of those 8 locations. The monkey knows that it is supposed to touch this square, but not yet. The second square grows to the size of the original, which vanishes. The monkey then reaches and touches the second square. As a reward it gets a small amount of juice. (I am sure that if you really wanted to help advance science, you could volunteer to have a microchip implanted in your brain and perform this mindless task thousands of times per day for sips of juice. Me? I require steak for such labors.)
By having the chip monitor this region of the brain they have actually designed an algorithm that can interpret the brain signals of the monkey and can provide an output of where the monkey intends to reach, out of the 8 positions. So what next? Well, they altered the monkey's task slightly: The monkey is supposed to keep its finger on the original square, but when it sees the next square appear it is supposed to think about moving its finger toward it, but it is not supposed to actually move its finger. The neurons in the brain still fire, the chip reads this output, and a small circle appears at the position the algorithm interprets as what the monkey intended to touch. As the algorithm is very accurate, the circle almost always appears around the second square. When this happens, the monkey gets more juice.
What is this supposed to help us with? Well, the algorithm reads brain signals which are a plan of arm movement. It then has an electronic output (the appearance of the circle) as a result. This, theoretically, will be extended one day to an algorithm that reads brain signals indicating a person's plan to move their prosthetic arm. The algorithm interprets this plan and outputs instructions to the arm, which moves exactly as the person intended. Probably to reach for a glass of juice, I suppose.
One thing that struck me during the seminar is that in regards to the second task, the monkey is rewarded not when it correctly thinks about moving its finger to the second square, but when the algorithm correctly interprets the monkey's brain signals. This could lead to the monkey's brain 'adapting' its signals to optimize accurate readout by the algorithm. I don't think that this is an intended consequence, but it may be beneficial. If the monkey can have its brain 'learn' to interact with the microchip/algorithm more efficiently, then a human could do so with a prosthetic arm. Pretty cool.
He began by discussing how they monitor motor neurons in the brain which control arm movements. They've implanted very small electrode arrays into very specific regions of the brain which are responsible for motor control of the arms. They then spent many many months training the monkeys to perform this basic task: A square appears on a screen in one of 8 possible locations. The monkey touches the square. A second, smaller square appears in another of those 8 locations. The monkey knows that it is supposed to touch this square, but not yet. The second square grows to the size of the original, which vanishes. The monkey then reaches and touches the second square. As a reward it gets a small amount of juice. (I am sure that if you really wanted to help advance science, you could volunteer to have a microchip implanted in your brain and perform this mindless task thousands of times per day for sips of juice. Me? I require steak for such labors.)
By having the chip monitor this region of the brain they have actually designed an algorithm that can interpret the brain signals of the monkey and can provide an output of where the monkey intends to reach, out of the 8 positions. So what next? Well, they altered the monkey's task slightly: The monkey is supposed to keep its finger on the original square, but when it sees the next square appear it is supposed to think about moving its finger toward it, but it is not supposed to actually move its finger. The neurons in the brain still fire, the chip reads this output, and a small circle appears at the position the algorithm interprets as what the monkey intended to touch. As the algorithm is very accurate, the circle almost always appears around the second square. When this happens, the monkey gets more juice.
What is this supposed to help us with? Well, the algorithm reads brain signals which are a plan of arm movement. It then has an electronic output (the appearance of the circle) as a result. This, theoretically, will be extended one day to an algorithm that reads brain signals indicating a person's plan to move their prosthetic arm. The algorithm interprets this plan and outputs instructions to the arm, which moves exactly as the person intended. Probably to reach for a glass of juice, I suppose.
One thing that struck me during the seminar is that in regards to the second task, the monkey is rewarded not when it correctly thinks about moving its finger to the second square, but when the algorithm correctly interprets the monkey's brain signals. This could lead to the monkey's brain 'adapting' its signals to optimize accurate readout by the algorithm. I don't think that this is an intended consequence, but it may be beneficial. If the monkey can have its brain 'learn' to interact with the microchip/algorithm more efficiently, then a human could do so with a prosthetic arm. Pretty cool.
posted by josh at 2:01 PM
0 Comments:
Post a Comment
<< Home