Pat Gunn (dachte) wrote,
Pat Gunn
dachte

Bound by Oath

Loki was the lesser of the trickster gods, his cousin having played the much greater trick of not existing. When the task grew old, he came into the world, but was blocked by the Norns. To be seduced by chaos, as was Loki's fate, is part of our world -- to be born of it is not. Thus was learned by noone that our tricks bind us more than they bind others.

On the way to work today, I was thinking about philosophy of science, in particular the preference that we have for the simplest adequate theory. This bears a certain resemblance to an issue in designing/training neural networks -- when I was a CS undergrad taking AI classes, many of the neural nets we used had a separate training and active period -- during training feedback would be computed to update the values on the network, and when the network performs well enough the values are frozen and the system is put into production. The purpose of this separation is to prevent the network from being distorted by statistically unusual input. The fact that this is seen as useful shows that we're probably missing out on something in the design of neural networks -- either our neural networks are sufficiently large that it's hard for them to be distorted by such means, or we have some "firmware' neurons that don't learn in quite the same way that impose a certain ordering on the process. Anyhow, I was thinking that our preference for simple theories resembles one of the dangers one must be wary of in neural networks -- overtraining. While programming them, we were warned not to include too many more nodes than a certain rule of thumb suggests because when one has insufficient nodes, one learns the underlying regularities of the inputs, but when one has sufficient, one instead memorises the inputs. If we allow ourselves more complex theories than needed, eliminating the preference for simple theories, we end up incorporating data rather than theory into our science, and just stating our observations. I find myself wondering if this suggests either a limit to possible intelligence (unless one is in an amazingly complex environment, with a better brain one just memorises inputs rather than learning underlying regularities) or that to reach a deeper intelligence the brain must keep developing to keep riding the wave of regularities versus memorisation. Then again, if the brain regions associated with memory are not engineered much but the regions associated with pattern recognition are, perhaps artificial boosting of intelligence may be achieved that way at some point. Finally, I suppose the process of developing rules for intellectual integrity and scientific learning is more the acquisition of an artistic sense than a science -- a lot of heuristic judgement is involved in aiming for the right level of theory (both determining what's likely to be the theoretical "right" level (natural complexity) and the pragmatic level that one aims for).

This winter, while miserable already (sinuses and static are already attacking me), is made much better by my having really awesome gloves -- the gloves I got for snowboarding in Switzerland last year keep my hands *really* warm.

In my quest to make a collection of goodies for someone, I found myself wanting to burn a DVD with the UFS filesystem. Unfortunately, xcdroast, which I've used for a long time, can burn ISO9660 onto a DVD but not UFS, so I downloaded and installed k3b, which is pretty decent (albeit slightly quirky).

I have found jealousy of friendships to be a very strange thing to have -- when people have been friends for a very long time, it's difficult to compete with that. It's interesting to me that this bothers me so much, but things of this nature are always kind of strange, touchy things. I don't think it'd be rare that some people want more (or less) from a friendship than others do, and that this can lead to tension, hurt feelings, or dissatisfaction. As consolation, it's different than issues with one's significant other on these topics -- it's something I can put aside without too much difficulty. I've encountered the other side a few times too, and that's always been awkward.

Last Friday, the 17th birthday of KGB, was awesome! We all played games and ate cake from about 19:00 until 22:30, and then hoofed it over to Wean Hall to watch the Rocky Horror Picture Show. I haven't seen it in forever, and as is the tradition for RHPS, I went in (mild) drag, wearing a skirt and a russian hat, both of which I picked up at a used clothing store around the corner from CMU on Fnord Avenue. Wean Hall isn't well organised for RHPS -- there wasn't much room to dance, but I gave it all I had and it was fun. The cast tossed condoms and similar out as party favours, the chocolate flavours of which I've been dropping into the collection boxes of the local salvation army bell-ringers (whose bell I'd honestly love to fill with peanut butter). There's a (remote) chance that the cast will be able to bring RHPS back for good, which would be awesome. One amusement was that there were more RHPS virgins than there were non-virgins. Also, this cast didn't do the pledge of allegiance to the lips. Oh well. There was another cool thing about that night that I'll leave out (no, it's nonsexual).


(section not shown)

The Red Cross/Crescent folk have decided on a third, nonreligious symbol that their local affiliates can use. This red crystal is a decent symbol, looks neat, etc. I think it'd be better to now phase out the cross/crescent in favour of making this universal, especially given that most nations are multiethnic/multireligious nowadays.

Subscribe
  • Post a new comment

    Error

    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded 

  • 0 comments