Consciousness, Neuroscience, and Me

Passing of Time

2006-07-13T13:51:00.001-07:00

Here is a nice review in the New York Times on recent work with the BrainGate. I think it's interesting to note here that for the cost and effort, there wasn't a spectacular outcome (although there was no disaster, either), but it's so sci-fi that people are dying to throw money at it.

I wrote a term paper on many of these issues, and my basic conclusion is that people are not going about motor cortical implants in the right way. The neuroscience is pretty clear, actually - put as many electrodes as you can in the brain, and your brain will figure out how to use them. What's not clear are the technical details - how long can an implant last and give a good signal, how can it be powered, how can transmissions be made wirelessly, what are the risks for infection, what kind of device will be controlled and how can it be robust enough to be controlled constantly? It will be interesting to see if there are any breakthroughs here in the next few years. If technology comes around, we could have some very exciting work ahead of us figuring out how the brain manages to figure out how to use the system to which it's connected. So when I say the neuroscience is clear, I mean it's clear from an engineering point of view. With respect to understanding the underlying mechanisms of learning and cognition, it's awfully grim, it seems.

Passing of Time

2006-07-13T13:51:00.000-07:00

More On Sensory Substitution

2006-01-23T16:00:00.000-08:00

I've recently been doing some thinking about sensory substitution, which is a very fun and seemingly neglected area of neuroscience. With tactile-visual substitution, would it be possible to get color? I think the way to do this would be to take three color channels and feed them all to different tactile areas - would people ever experience color qualia from this? I have no idea! But I kind of want to try it!

Here are some links for our reference:

http://www.seeingwithsound.com/sensub.htm

This is a great little page, with lots of general sensory substitution information.

http://kaz.med.wisc.edu/Publicity/FAQ.html

This is the webpage of one of the central researchers in the tactile-visual substitution studies at Wisconsin, with Bach-y-Rita

http://www.wicab.com/technology/scientific_publications.html

This is Bach-y-Rita's company

"Bionic Man"

2005-12-11T15:06:00.000-08:00

At the talk that I went to last week, people eluded to some sort of human artificial limb research that I hadn't heard about. I'm pretty sure this is it, and I have no idea how it works yet, but it looks awfully cool, doesn't it?

Alva Noë

2005-12-09T04:51:00.000-08:00

I have a hunch that Alva Noë really knows what he's talking about. I should probably read his book Action in Perception, where he makes the claim (maybe?) that perception is not established by activating certain parts of the brain, but rather by the "lawful connection of perception and action", which makes a lot of sense. One of the big issues in consciousness, for me at least, is that certain parts of the brain seem to just be associated with certain modalities. But why should vision be visual, how does that part of the brain know that its visual? By an action-type argument, we're able to make a connection between how we act, that is, moving around in the world, or manipulating objects in the world, and how we interpret this, that is vision. Perhaps once this is established, other sensory-substitution type stories will make sense, that is, explain why certain non-visual information, e.g. tongue stimulation will produce visual stimuli. What quality of that information is inherently visual as opposed to inherently auditory? If we glean "visual" information reliably from the environment, does that mean that it will always have a visual quality to it? Hmm.

Procrastinating

2005-12-08T23:23:00.000-08:00

Right this moment, I am trying to write a paper applying parts of Patricia Churchalnd's Neurophilosophy to the VITE model, as described by Bullock & Grossberg in 1988. Problematically, I feel like I don't have a whole lot to say about the VITE model, as it's kind of an old version of the model and none too exciting a decade and a half after it's been developed. And yet, I am supposed to discuss whether it is at the correct grain of analysis.

Anyway, doing so I discovered a neat blog that is trying to apply VITE to a robot arm, which seems like a really good idea. I hope that more people realize that their intellectual struggles have a warm home in the world wibe web.

As far as my adventures, I went to see Lee Miller come give a talk at MIT, which appeared as though it may have been hosted by Emilio Bizzi, although I can't be certain, because I was late due to the very, very cold nature of the day and the very, very big nature of the McGovern Institute. I wish I understood more of it, but the big idea seemed to be the high correlation between neurons in Motor Cortex and muscle activation as recorded through chronic EMG electrodes. I didn't know people implanted that sort of thing, but it seems useful, especially if you're aiming at stimulating the arms that you have if you're paralyzed. Of course, there's still the "recruitment problem", which is that when you stimulate a muscle, the larger, fast-fatigable muscles are recruit more quickly, which is the reverse of what takes place when muscles function normally. Overcoming this problem is non-trivial, and there are also apparently problems with getting muscle fusion, or tetanus, at low enough frequencies. Unfused muscle movement is tantamount to twitching a lot, which is not the smooth movement that we're all familiar with.

It's also dawning on me that if I ever hope to be a "motor guy" I'm going to have to learn some hardcore anatomy and physiology of muscles, and also the spinal cord, in serious detail. We'll see.

Oh, and next time, I'll figure out consciousness.

Useful Resource

2005-11-20T11:20:00.000-08:00

MIT has one of the best Cognitive Science style programs in the country, and being MIT, they decided that their course resources should all be online. They don't all have the relevant materials online, but the syllabi and links to reading are generally there, and sometimes lecture notes, too. Particularly interesting is noting the differences between Introduction to Neural Networks at MIT and Principles & Methods of Cognitive & Neural Modeling (CN 510) at BU, in which I am currently enrolled. I think I might be a little more pleased if there were more of a consensus here about what is the best way to go about, oh, teaching an introductory course in neural network type stuff.

Going to graduate school has some unpleasant properties that are akin to religion, in that you kind of choose one approach to whatever you're going to study and then kind of ignore all the rest. My opinion of religion can be summarized in that there are N religions, so all things being equal, you have a 1/N chance of just being completely wrong. I think the same thing applies here to a certain extent, and that if something isn't part of the curriculum in my department, I should at least know what it is. And what's even stranger is that my department seems to agree - each year there is a huge conference here at BU covering a huge array of relevant studies, not just what's in our syllabus. But if it's important enough to show up at a conference sponsored by my department, then why do we have such a high level of redundancy in the courses here as opposed to covering this other material? Mysteries abound. I'm pretty sure that similar things are taking place at other universities, but in truth, I have no idea.

Consciousness

2005-11-19T12:08:00.000-08:00

So recently I've come to discover that what I mean by "understanding consciousness" is different from what other people mean. Of course, this isn't exactly a surprise, but I still want to clarify what I mean. I think that it's been pretty clearly established that the brain is crucial for consciousness, whatever you think consciousness is (and in animals that have brains). That is to say, if I stimulate the right parts of your brain, you get certain conscious percepts, and if I stimulate an apple, you get no conscious percept at all. Or, if you stimulate someone else's neurons. Or if you stimulate the wrong neurons in your brain. What is it about neurons in V1 that lead to a percept? Is it their pattern of connections to other neurons, or is it some intrinsic quality about that set of neurons itself?

Come to think of it, I don't even know what layers are stimulated when you perceive things in V1, but let's work it through, I guess. Here is some background on V1. It would be convenient to say things like "Layer 6 only projects to lower cortical regions, so if stimulating it leads to a conscious percept then this arises from either Layer 6 or the lower cortical regions", but I think that the canonical A connects to B discourse is probably false.

I know that some people think that the focus shouldn't be so much on neurons, but rather on global activation, that developing a theory of consciousness should start from what we can observe, eg, fMRI type patterns of activation, as opposed to what we cannot observe, eg, quantum phenomena in microtubules. But I don't know how this is going to actually help us understand consciousness on a scientific level, that is, why stimulating in V1 yields a visual percept, whereas stimulating in M1 does not.

I mean, you can always look at broad patterns of connectivity, and say, in obnoxious form, V1 gives visual percepts because it is connected to the eyes. This sort of ties back to a Jeff Hawkins type story, of the cortex identifies patterns, and as such identifying visual patterns simply is a visual percept. Of course, this is equally unsatisfying. But the goal here is to frame the questions underlying consciousness in a way such that they can be investigated.

A lot of this, I feel, ties in to development - do babies have visual percepts? We assume so, but when are these developed? Is it innate? We know that ocular dominance columns develop in utero, are these actually visual percepts? Of course, this is also more useless questions with no clear path to investigate them.

In the meantime, you can go read more about this from smarter people than me, and watch some lectures on the subject which are pretty basic.

An Introduction

2005-11-11T01:42:00.000-08:00

Lots of other people manage to have quite capable "work blogs", where they can work ideas out and refer to them later in an online forum that allows for comments by interested parties. I think I can, too!

So, I'm a first year graduate student at the Department of Cognitive & Neural Systems at Boston University. I have no idea what I'm going to study here, but I (oh so modestly) want to, you know, figure out how this consciousness thing works, and then come up with ways to utilize this knowledge that don't involve evil government mind control and tacky consumer products. I'm going to do my best to keep things at a legible level, although I'm sure that this will be pretty much exclusively for my benefit anyhow.

I studied Cognitive Science and Electrical Engineering & Computer Science at UC Berkeley, where I took a very good class by John Kihlstrom on Scientific Approaches to Consciousness. I very nearly went to graduate school at the University of Pittsburgh, to work with Andy Schwartz on neuroprosthetics, and that's still one of my central interests, academically. Neuroprosthetics is exciting because it manages to restore lost function in a profound way (via a mentally controllable prosthetic arm, in Schwartz's case) and also addresses some key features of internal motor representation, how this representation changes over time and how it relates to our own conscious experience.

One tough question though regards what is the best way to restore lost function to people who have lost a limb. Of course, a chronic neural implant is an option, and is being pursued by a pretty large number of labs (Schwartz, Donoghue, Nicolelis), but there are a huge number of issues that arise here. First off, how long does an implant last? Signal decays over time when it comes to such implants, keeping a microelectrode array in the brain for a decade and expecting to get strong signals from even a quarter of the electrodes is most likely a pipe dream. What does this mean for Matthew Nagle, then, who has a permanent array in place, and has for around a year, if not longer?

If an array dies, do you do brain surgery and implant another one? I think there are probably a lot of lessons to be learned here from cochlear implants, and probably also pacemakers, which are presently beyond my ken. Also, how can such an implant be wireless? How does it get power? Of course, these are all implementation details, but that's really the guts of any sort of neural engineering scheme. I have a hunch that if you could record from pretty much any population of neurons and use this to, say, control a race car, your subject would figure out how to control it, and find it quite natural to do so. Of course, what do I know, but still - the brain is good at things like that. Look at development - you basically take all sorts of neurons and randomly hook them up to each other (or do you?) and when you're born you spend a few years tuning the system, figuring out how to control things, and you end up doing a remarkable job. There's no instruction manual here, everything is embedded, and there is no innate "this is visual input" kind of information tagged to the retina, LGN, V1, whatever.

So, let's say you have a system that is a hybrid biological and neural system, such as a neuroprosthetic, or something even more fundamental like DeWeerth is doing at Georgia Tech. If I recall, he managed to teach neurons ex vivo to perform some sort of task, which is pretty remarkable, really. First off, wow, how can we use this, and second, what are the implications of this with respect to consciousness? DeWeerth is also special because he graduated from a program similar to mine at Caltech (Computational & Neural Systems), and he's trying to put his models to good use in biological systems, which is pretty fantastic.

Also, it looks like some of his models deal with the spinal cord as opposed to the brain itself, which ties in to other questions of neuroprosthetics. Is it better to go at the brain directly, or try to tap into the spinal cord, which is where a biological arm would have synapsed anyway? Certainly, a lot of the neural structure is still intact, and if you can harness that at a lower level, then you have a lot to gain. Doing surgery on your arm five times is OK, but doing surgery five times on your brain is probably not so hot. And plus, if your surgery goes horribly wrong, you don't destroy any brain tissue at all, ever. Of course, damaging the spinal cord is also really heinous, but there could be ways to try and access the sensory and motor fibers that remain in what is left of the limb.

For example, could sensory input from an arm be given by direct electrical stimulation, such as through explorations by Paul Bach-y-Rita in terms of sensory substitution? For those who aren't aware, Bach-y-Rita managed to provide visual input to blind subjects through electrical stimulation on the tongue. Subjects report visual phenomena, and are able to identify and interact with objects in the world. And this isn't even an invasive technique, putting a chip on the tongue. But how could this be integrated into something long term? Again, this is the sort of thing that could be very appealing for a neuroprosthetic device.

I should probably do some real chatting with Robert Ajemian, who shares many of my interests, and also graduated from my program here at BU.

Anyway, there's some food for thought. Later on I'll figure out consciousness. Yep.