The Prof discussed the experiment whereby humans neural activity was charted by a computer as they were exposed to different 'artifacts.' In the example given, one person's perception of a knife was recorded by the computer and it was shown that the computer could tell what the person was looking at simply by the neural patterns. Researchers had discovered that different people's brains exhibited the same characteristics when processing the knife. The class seemed very interested in this and discussed this for several minutes. Some questions I had were:
1. Was the person asked to 'say' the word representing the object they saw, or just to look at it?
2. Did everybody look at the exact same knife? From the same perspective?
3. How would/should people's experience with knives play into their neuro-biological 'output'? For instance, would a chef who used knives like the one displayed on a daily basis have the same 'reaction' as a person who had little familiarity?
4. How would adults' perception of the knife differ from that of children?
In the end, the Prof explained the 'magic' behind the computer's ability to read minds, but it seemed that some of the class was still unsatisfied. (Note that the answer to number 1 is NO; I must have misunderstood.) What I gathered from his explanation was the following:
The computer detected neural activity, but the portion of the activity which actually spoke directly to the object in view was the portion connected to the visual center. That is, the computer looked for (and discerned) whatever happens in the brain when the eye falls on a particular color, depth, pattern, shape, blah, blah, blah. The information picked up by the eyes is somehow mapped and the device read and interpreted that mapping. That doesn't however, answer all of the questions. Would not a person looking at the knife from a different perspective (or angle) have a different resultant mapping? (the answer might be that the mapping is different but similar enough for the machine to detect the pattern)
What about the question of familiarity? Wouldn't the chef 'perceive' the knife a bit differently? (perhaps there were no chefs in the study)
Are not children's perceptual-cognitive representations of knives different than that of adults? That question raises the next question: is there a perceptual 'cutoff'? I guess (answering my own question) that a human is a human and if the eye captures the image, the mapping would be the same. What about a chimp's representation of the knife?
Just some quick notes so that I could say I participated in the blog and not be lying. Really, though, this class is thoroughly enjoyable to me and I love to listen to the discussions raised by the students from time to time. This blog should be a great tool. See you guys in class.
Subscribe to:
Post Comments (Atom)
Thanks Jarrad for summarizing the experiment. I was very unsatisfied with this experiment and the result and I try to explain why:
ReplyDeletewhat does this experiment want to show is that all human beings have the same pattern of neuron activity when we see a certain object. but why should somebody expect something else? isn't it a simple conditioning? we train the machine by presenting A-->B relation and then we present A, the machine will say B.
"A" is "knife" here and "B" is "the pattern of neurons for knife".
the excitement part of the experiment was supposed to be that "B" is unique for a given A. but why shouldn't it be? I don't think the guys on the other side of the spectrum (like Pylyshyn) ever wanted to say that we don't have any unique pattern of neuron activities in our mind when we actually "see" a knife. the debate has been misplaced or mis-presented in this experiment. The debate was about the "images" in mind not about the actual object outside the mind. the argue was about the symbols in cognitive, analytical part of the mind and not even about the perception.So, as you have correctly pointed in your first question, we should see if this account holds when they imagine a knife or not. Otherwise, this experiment at most tells us that we have the same visual receptor system in our brain. something that I think we all knew from the high school!
Hi, here you can watch the videos of the experiment on CBS 60 minutes.
ReplyDeletehttp://www.cbsnews.com/video/watch/?id=4697682n&tag=contentMain;contentBody
http://www.cbsnews.com/video/watch/?id=4892613n&tag=contentMain;contentBody
http://www.cbsnews.com/video/watch/?id=4892617n&tag=contentMain;contentBody
http://www.cbsnews.com/video/watch/?id=4892615n&tag=contentMain;contentBody
http://www.cbsnews.com/video/watch/?id=4892632n&tag=contentMain;contentBody
To jump in on the discussion...
ReplyDelete2. Did everybody look at the exact same knife? From the same perspective?
I think this question is critical to our previous discussions on the visual knowledge representation issue as well as possibly introduce a new concept: is the way in which knowledge is represented in our brains a function of the task we are expected to perform, something we know a priori. I bring this up because if we, as the subjects, know our only task is to "view" or "regard" or "think about" an object, then I would expect that orientation information would be filtered out rather than taken in. BUT, if we go into an experiment told up front that we will be asked to "grasp" and "manipulate" and "handle" said object, this is probably information we want to process once we've viewed the object. Since it doesn't appear that the experiment involved any interaction like that, I'd venture that it did NOT matter.
3. How would/should people's experience with knives play into their neuro-biological 'output'? For instance, would a chef who used knives like the one displayed on a daily basis have the same 'reaction' as a person who had little familiarity?
This is VERY interesting as it, I believe, brings the issue of experience into play. Where the layperson may see the knife as a tool for eating and maybe even preparing food for one's self, a subject like a chef may regard the knife as a tool that represents a subset of a larger toolset for making a living. Clearly, I'm teetering on the line of being a "folk psychologist" but just so that I can try and stay on the right side of things, I would propose that the frames we create and populate during this type of experiment must be experience-independent given the proven results.
--!To digress for a bit, I just saw a snippet of the CBS clip promoting the actual story and (as all of us scientists and engineers should expect) there probably won't be too much technical jargon used to explain the in's and out's of the experiment. One nugget of info that I did notice is that the items shown to subjects were pictures, not actual objects in 3D. I think this was a wise starting point for the experiment given how complex the debate must be between not only thinking in terms of images, but 3D images at that!--
4. How would adults' perception of the knife differ from that of children?
I definitely would offer that the system may not work as well being trained by data from children and adults as it would with adults only, simply because the frame of reference may not be as broad for a child and the frames a child may develop may not be as extensive, which may leave gaps for the machine learning algorithm to fill in. This doesn't address all of your comments, Jarrad, but I did want to chime in on some of the questions you raised.
Well, now that I've seen the FULL story on CBS, I guess I have to amend a few of my earlier comments. 1)As expected, there wasn't too much technical jargon and the reporting went beyond just the capabilities of the system at CMU. 2) Professor Just eluded to the fact that when we see these images (2D) we think different thoughts about them, so not just that they exist in our visual line of sight, but also what they may be used for and even how we would use them, so this challenges my 2D versus 3D comment, which is good because I didn't consider that we probably (more than likely) project what we see in 2D into a 3D plane in our minds. 3) Just as a general comment on the ethical issue, I REALLY liked the question of whether FMRI information was "DNA-like" evidence or "testimony" in a court case. I'd like to hear others' thoughts on that one...
ReplyDelete