GPU hardware

March 3rd, 2010

This is probably of limited interested to most of my readers, however for those of you who care…

UC Davis has made available a course, EEC277: Graphics Hardware@ which, as the name suggests, is a course about the design of modern GPUs. There are discussions of how GPUs differ from CPUs, why Open GL (and DirectX) are structured the way that they are (as opposed to alternative possible graphics programming models), what types of computations GPUs are good at and how GPGPU programming environments are structured.

On the one hand, I was frustrated by the course because I hoped for more detail than was provided. On the other hand, to be fair, this course was really interesting, putting in a single coherent framework the few scattered pieces of information I previously knew about GPU programming. I really hope that Davis (or some other university) provide a successor course, more detail and less introduction, which they make available for podcast.

Crowdsourcing work

March 3rd, 2010

This was a remarkably interesting talk, given by Lukas Biewald at Stanford, on September 25, 2009. The subject was Crowdsourcing Work@, by which is meant specifically providing intellectual piecework, for pay, anonymously via the web. This is the sort of thing done by Amazon’s Mechanical Turk, along with various other me-too organizations. The sort of work made available is, to take a simple example, scanning photographs to look for certain objects, or scanning text message to rate them as positive or negative — fairly simple-minded work that is, nonetheless, not (yet) doable by computer.

Biewald runs a company that mediates between companies that have such work to be done, and the infrastructure companies like Amazon. He discusses a large number of aspects of this situation, from the types of work that are done, who is doing it, how long they do it for, and how (and how much) they are paid, including such issues as how do you guard against fraud and make sure that people did the work, and did a decent quality job.

Obviously Biewald sees the good in these sorts of arrangements, and is quick to point out issues like at least some of the people doing this work are in refugee camps, and have no other work available to them. Nonetheless one is driven to wonder if we are seeing here, in embryo, an important part of the future labor environment. Depending on your politics and obsessions, there is much to get excited about here, from the fact that this sort of work does not care if you are a Palestinian in Israel (or any other oppressed minority, including the handicapped, or women in Islamic countries), to the fact that this provides a wage neutral and labor-law blind access to the entire world’s pool of labor.

A good summary of the US health care system

March 3rd, 2010

From Stanford’s 2009 homecoming, we have Lawrence Baker discussing Where is the Health Care System Headed?@.

If you want a single one hour summary of how the US health care system works and the sorts of ideas that are in play to change it, this is about as good as you are going to get. (There is not much discussion of the way the health care system should be structured, and thetypes of bills that should be in Congress, but you can’t blame Baker for that — he is discussing the world as it is, not the world as it should be.)

Update to the Hubert Dreyfus post.

March 3rd, 2010

The Hubert Dreyfus post has been updated.

Dollhouse and Avatar — what do they mean?

February 23rd, 2010

We’re all familiar with discussions of how aspects of past pop culture reflected intrinsic fears in society — that monster movies of the 50’s and 60’s, for example, are a manifestation of fear of nuclear war. With this in mind, it seems to me that Dollhouse and Avatar are both sufficiently different from what has gone before, yet have sufficiently similar themes, that we can start to ponder what is going on here.

I think it’s worth stating, up front, that I think we have to see these both as not simply modernized versions of demonic possession or zombiehood. In these previous genres, the focus is on the body, and on how the body has been “stolen”; in the works I’m discussing, the focus is on the mind, with the body as disposable soma — what appears to matters is that the mind can be poured from one vessel to another, not the vessel in which it happens to reside.

So why should this subject now, in the late 2000s, become a hot topic? One possibility is the issue of “Identity in Cyberspace” (all capitalized), the suggestion that the internet has permeated enough life, and people have had enough experience, both with their fake facebook pages, and with meeting internet dates who were not what they claimed to be, that the fact that we can construct fake digital personae has become mainstream. This, however, strikes me as wrong — the emotional resonances of fake digital IDs and identity theft just don’t match these two works.

Here’s my offering of a possible interpretation. We start with my claim that the last thirty to fifty years have seen economics in the US transformed from a social science, occasionally useful for political purposes, into an all-consuming theology/world-view. No matter what the subject, whether it’s marriage partners, how people choose a church, the meaning of the law, or the point of society, economics has an opinion on the issue, and that opinion is becoming the defining opinion in the US. It is just taken for granted that the right way to view law is through the veil of “law and economics”, that the purpose of politics is to maximize GDP, that success in life is measured by wealth; these are considered such obvious points that they are not part of the debate. Health-care in the US right now is being discussed exclusively in terms of who will pay for private insurance; the idea that insurance should not be run by private corporations, let alone that health should not be run as a for-profit enterprise are just not considered serious viewpoints.

One central tenet of this world-view is that humans are atomized individuals, born into the world as little calculators, and interacting with other humans via contracts which can be broken at any time if there’s any profit to doing so. I have to wonder if Dollhouse and Avatar are reactions against this. Of course, prima facie, they stand precisely in the economics camp. giving us a vision of minds that can be shifted from body to body as convenient, an obvious extrapolation of the atomic individual unconnected to society. But in Dollhouse, the larger story is that the bodies rebel, that even as successive minds are poured into them, they are not empty vessels; and, of course, that minds disconnected from bodies leads, in the end, to disaster. The Avatar story is simpler, much more traditional and less interesting — the Americans are individuals, to the ultimate extent that they’ve developed the technology to pour minds into new bodies, whereas the smurfs are all connected as one happy family. Cameron, however, goes out of his way to emphasize the strength of that connection, so that we don’t just have mysticism about the great world spirit, we see it made concrete via the nerve roots at the end of the smurfs’ pigtails; one could argue that this is done precisely to state that this connection is not just words, it is every bit as real as a fiber-optic cable.
In other words, I see both of these works as pushback against the idea that humans are nothing but individuals, as attempts to assert that it is precisely our ties to each other that make us human.

With only two examples from larger pop culture (certainly I can’t think of any more) it is a bit much to declare a trend. Nevertheless, I do feel there is something here, and if one of my readers ever gets to meet Joss Wheedon or James Cameron, it might be worth asking them their opinion.

What patterns are there in dreams?

January 15th, 2010

G. William Domhoff at UCSC is interested in understanding dreams from a rather different perspective than the “meanings of dreams” model (including the Freudian scheme). Rather, he and his collaborators are interested in questions like@ “how often do animals appear in dreams? what about violence? how many strangers vs people you know?” and so on; with these numbers tallied against men vs women, or different cultures.

This is interesting stuff, and it is interesting to see the ways in which the dreams of men differ from those of women, likewise across cultures. One could certainly, if one wished, spin tales of the pre-occupations of different people. There is, however, a nagging worry in my mind that severe editing and translation occur in the mind in the transition from dream to recollection of the dream; and that it is perhaps this translation that is strongly culture dependent, not the dreams themselves. Obviously this insight doesn’t give us much to work with — how do you get at a dream before it goes through this translation? — nevertheless I do think it is worth bearing in mind; and thus that grand conclusions drawn from this sort of research should be taken with a grain of salt.

Is the US civil legal system as bad as is claimed?

January 15th, 2010

We are constantly barraged with claim about the pathologies of the US civil legal system, how terrible large awards are, how “ambulance-chasers” make it impossible for professionals to do their jobs, etc etc. It’s nice to have some perspective on the subject, which is what Stephen Yeazell provides in this talk from UCLA, entitled What’s Not Wrong with the Civil Legal System – And What Is@. As always, it’s not enough just to carp about the problems with a system; serious objections have to understand the problems that are being solved, and to suggest better solutions, and it’s not clear, for many (not all) of the complaints against the US system what a better solution would be.

LED lighting

January 15th, 2010

LED lighting is starting to become a subject of some discussion, and, as an outsider, it can be somewhat mystifying to try to reconcile the great things claimed for it by some parties with the reality that one simply cannot walk down to WalMart and buy an LED bulb to screw into one’s existing light sockets — how hard can it be to manufacture such a device?

From a UC conference in October 2009 we have this@ very nice overview of the current state of LED lighting. It should come as no surprise that the most enthusiastic proponents of the technology play somewhat fast and loose with the fact and compare apples to oranges. Some of the problems include

  • There are spectrum issues. The very high efficiency LEDs generate a very blue spectrum, which, while fine for flashlights or LCD backlighting, people apparently find distasteful for use in homes. Lighting for homes requires more yellow and red, and the existing LEDs produce those very inefficiently.
  • Which gets us to — I assumed all along that these white LEDs were basically optimized red. green, and blue LEDs in some cunning way fashioned together on the same die. Apparently not. Few details were given, but as I understand it, the current model for cheap white LEDs is in essence the same as fluorescents — generate high frequency (blue, not UV) light, and downconvert some of it to yellow through a phosphor coating. Obviously this means there is scope for improvement via dispensing with down-conversion and generating R, G, and B directly; on the other hand it also suggests that the technology for manufacturing multiple different frequency LEDs in the same device is immature and/or expensive. A related issue is that right now LEDs prefer to generate a spiky spectrum rather than a smooth black-body-like spectrum. Again it seems people prefer a smooth spectrum, and forcing this results in some inefficiencies.
  • There are packaging issue. A selling point of these devices is that they are very long-lived, but this is undermined if the packaging is poor. The optics cannot, for example, be made of cheap plastic because that yellows substantially before the LED dies. A similar example is the use of electrolytic capacitors in the packaging. One wonders who will police these issues in the real world once these devices become common. Certainly my experience has been that while compact fluorescents last longer than incandenscents, they don’t last anything near as long as the lifetimes claimed.
  • Apparently a substantial problem with heat dissipation in fitting these devices to existing sockets. You’d hope they are so efficient heat dissipation is not an issue, but sadly no; not yet anyway.
  • It’s not all bad news. There are environments where LED lighting is already superior to the existing alternatives. One unsurprising example is lighting for the entertainment industry (movies and TV) where the robustness of LED lights (for mounting on cars and such) is a real boon, as is the longer lifetime and much lower heat generation. More surprising is cold environments, where fluorescents apparently do not work well, like in the refrigerators of supermarkets.

Ben Polak course on Game Theory

December 17th, 2009

Game Theory is a subject I’ve long felt I ought to know something about, but the two previous courses I tried on the subject were, to put it bluntly, just awful. Not so with Polak’s course@, given at Yale. (I would recommend that, unlike most courses, this is worth viewing as video, not just listening to as audio. I watched about half a lecture at a time over lunch.)

The first impressive aspect of the course is just how carefully Polak has chosen his examples. he constantly switches from a theoretical game, described on the blackboard, to a “real-world” game, described to the students, after which they are asked to make decisions. At each step, as we alternate, there is the obvious addition of some new idea along with repetition of a previous idea in a way that is quite glorious.

The second impressive aspect of the course is seeing the games play out in a real world situation. It is one thing to read about one of these games and try to imagine how a variety of different people might react; it is something else to actually see the variety of responses before your eyes.

The third impressive aspect of the course is the extent to which the students participate, and the extent to which the students are, generally, bright and articulate people. The course is not as interactive as I understand an HBS or Harvard Law class to be — mostly Polak speaks, with the occasional live-action game, followed by asking people why they did what they did — but there are plenty of questions directed at the class (which is also being recorded). For every muddled or outright silly response from the students, there is another powerfully smart response, phrased more elegantly than I could probably have managed given the circumstances.

David Sloan Wilson on Evolution

December 17th, 2009

Evolution seems to generate a whole lot more patently stupid statements, from supposedly intelligent people, than most science. Yes, you get idiot statements about physics, or global warming, or whatever, but from people that are confirmed twits. It is evolution that gives us frog-mouse battles with such inane starting points as “Intelligence: nature or nurture?“, or “Group selection: myth or fantasy?

David Sloan Wilson, a much more patient man than myself, here gives a marvelous talk on just what group selection really means, and how it plays out in the real world.
To be sure, that’s not his aim. His first few minutes discuss the basics of natural selection, and how sad it is that Americans are so stupid about it — all stuff you’ve heard before. Then we get a few minutes on how the social sciences are locked in their own little bubble of refusing to believe that evolution has anything to teach them; again mostly familiar stuff. But with these preliminaries out the way, we get to the good stuff:

How would evolution play out if we put one good and one evil person on an island?
OK, now we put a community of good people on an island, and a separate community of evil people on a different island. What happens?
And it’s not just speculation; we then get a discussion of some real-world consequences of this sort of thinking with respect to the breeding of farm animals.

And so it goes. Fascinating.

If you liked that talk, you’ll probably also like this one@. The first half covers the same material as the previous lecture, but the second half goes on to discuss these ideas (how appropriate responses are shaped by the surrounding environment, and the consequences for the evolution of altruism) in the context of a large, on-going study of teenagers and how they react to the social world around them.