Scientists Build Mind-Reading Computer

An anonymous reader writes to tell us that researchers from Carnegie Mellon University have developed what they are calling a "mind reading computer." Using a panel of nine volunteers, the team built a "profile" of 58 test words based on brain scans taken while the volunteers were directed to think about the meaning of each test word. "'If I show you the brain images for two words, the main thing you notice is that they look pretty much alike. If you look at them for a while you might see subtle differences,' explains Tom Mitchell of the Machine Learning Department, which lead the study. 'We believe we have identified a number of the basic building blocks that the brain uses to represent meaning. These building blocks could be used to predict patterns for any concrete noun,' added Mitchell."

well that doesn't work

[ILuvRamen]

All you have to do is have a different connotation for the word and it doesn't work. The gays stole rainbows so now if people see a picture of a rainbow, they have a distinctly different reaction to it. Or you could purposely make yourself feel angry or sad or do a complex math problem as you're thinking of the word and it would throw the machine off. To get this to work I'd bet they have to tell you to stay calm and what to think about beforehand, during, and after they try to predict what word you're thinking of or whatever. Gee, they tell you how and what to think and then predict what you're thinking of. AMAZING!

And so it begins..

[multipartmixed]

..THIS is the basis for yet-another-trek-related-invention: the Universal Translator.

I always knew it had to work this way.

Objects and Nouns

[natedubbya]

This area of research has been growing more popular lately. Last year's big language conference had a keynote speaker address the question of brain waves and word recognition. Most of the progress though is based on nouns because they have a core rooted meaning in everyone's head...you basically visualize a generic version of that object in the world. You say hammer, I think of an actual hammer I've seen. It's not really mind reading because the approach falls apart when you start talking about verbs and actions...which are what most conversations and thoughts are about. Actions don't have a stereotypical physical representation in the world, but rather involve several objects with that action, all interacting in some way that defines it. As I understand it, the patterns they observe in the brain then become too complex to capture.

Re:And so it begins..

[the_humeister]

That's an interesting idea. Do different words that mean the same in different languages light up the same areas of the brain when a person thinks about it? Would a Spanish person who is told to think of "coche", have a similar brain scan of an English person told to think of "car"?

What about Pron?

[brunokummel]

I know it sounds funny but i would like to see the brain activity for pornografic pictures, since it already known that "bad words" are stored in a different area of the brain than regular words... it would be kind of interesting if "bad images (or nice depending on the person)" got also stored on different areas....

Re:And so it begins..

[zappepcs]

Doubtful actually, at least in all cases. In English, nova has one or two meanings that may bring different thoughts. In some Spanish speaking countries, they might be thinking 'no go' or some option for various value of go in Spanish.

Grammarians unite! Only those who understand language will be able to interpret the results of this machine.

It is quite interesting that there are parts of the brain that light up uniformly (or near it) for some processes. Puts the human brain more in the land of machine with wetware and further away from the land of magic and such. There is probably still a LOT of work to be done before that universal translator does anyone any good.

Re:And so it begins..

[Feanturi]

And then you put a person who was born blind into the MRI and ask them to think about a car. Now what?

Call me when it translates...

[icebike]

Quoting article:
"We believe we have identified a number of the basic building blocks that the brain uses to represent meaning. These building blocks could be used to predict patterns for any concrete noun..."

The implications of building blocks would suggest that the french word for "Desk" (bureau) would elicit the same response as the english word for "Desk", instead of some governmental unit.

That would be useful, (once we get cheap portable MRI hats).

However I doubt these building blocks are anywhere near that generic due to the excess emotional baggage that people associate with words. I suppose it might be able to detect the presence of such baggage even if it could not decipher it.

Re:Words chosen

[PopeRatzo]

I've got a feeling the words involved in this project were more along the lines of "Department. Defense. Homeland. Security. Surveillance."

Re:And so it begins..

[locallyunscene]

This reminds me of a psyche 101 exercise where students were asked to draw a map of the town in which the college resided. Upperclassmen drew far more detailed maps than freshmen. I suspect the upperclassmen would have thought of very different things(past experiences) compared to freshmen also, and that's within a very small subset of people. I doubt very much that this machine could "read your mind" primed with someone else's input, but it could be invaluable in determining how the brain works and what similarities do exist across regions/ages/cultures if any.

Re:As dangerous as it is useful

[ceoyoyo]

There are just a few physical limitations on fMRI technology.

The most important for this topic is that fMRI scanners measure changes in BLOOD FLOW. They do not measure electrical activity. The flow response is delayed by about three to five seconds and has a certain minimum time duration. Therefore the requirement to enter the password r...e...a...l...l...y s...l...o...w...l...y.

Secondly, in order to get any recognizable imaging signal at all (and if you want to measure letters you're going to need a REALLY high resolution image) you've got to have a uniform, intense magnetic field across the head. The only way we know of doing that is using a magnet that surrounds the head. That places a limit on the minimum size. Note that although my calculator is smaller than a room now, my gloves are still the same size they always were.

Third, the fMRI signal is low amplitude and noisy. The results you see are all statistical. Using a higher field magnet can increase the signal a little, but there are hard limitations on that too. The only way to decrease the noise is by cooling the subject. I know I'd object to typing my password for you if you tried to cryogenically cool my brain.

Maybe some day there will be a way of scanning someone's brain to steal their password unobtrusively. This isn't it, and there's nothing on the horizon that looks promising, so you're not going to have to worry about it for a LONG time.