Wednesday, March 30, 2005

Questions From Kai

I know, not THE most creative of questions.

Questions are only as creative as the answers they get. I think these were fine...

1) favorite song, movie and album? Why?

The hard one first, eh? Honestly, I don't know. I don't have such things. For movies I'm afraid I've just never been that interested. I prefer television to movies, and books to either, purely because of the extra scope for story-telling each tells. So any favourite movie would be a random name picked from a bunch of films I don't really care that much about. Sorry.

For music, the problem is the exact opposite. I adore music, so narrowing things down to one answer is impossible. Next week the answer will almost certainly be Deadwing by Porcupine Tree. I haven't heard it yet (it comes out on Monday) but they've never let me down yet. (By the way, anyone wanting to join me at the Garage for their gig a week on Friday should be able to get tickets pretty easily. Easily one of the top five imaginative rock bands on the planet today.) In general, the answer changes daily, depending on mood and what I've just listened to. Normally, bands which are a bit tuneful, a bit experimental, and a bit different are favoured. That covers lots of jazz and rock artists (at the Peter Gabriel, Marillion, Radiohead end of rock) but also more normal stuff like KT Tunstall or Savage Garden.

As for single songs - you can't make me choose just one song from the hundreds nay thousands I like! Eek! Oh, alright, last week I went through a phase of playing Signify just to hear Dark Matter (Porcupine Tree again). I don't really know why - something in the riff, probably.

2)Do you believe in predestination, free will, or something in between? Why?

Predestination has been dealt a huge blow by chaos theory and quantum mechanics. Chaos theory starts from the fact that any system can only be modelled accurately by a system that is (a) more complex and (b) aware of the initial state of every variable in the system. (And even then, this is only true of a closed system...) Or put another way, for sufficiently complex system, the most exact prediction is little better than a wild guess. Then Heisenberg's Uncertainty Principle states that we can't know the exact value of most variables that we'd need to know to understand the human brain.

All of which means that we can't, even in theory, predict future human behaviour exactly. Hormones, chemicals, genetics, external stimuli might give a strong hint (if only on the level where we understand "personality") but there is always room for people to surprise and even shock us. On a very practical level, if it looks like free will why not call it free will and have done?

And, yes, this does all tell you a lot about my religious beliefs...

3)Why genetic algorithms, and what are they?

Okay, I'm going to be very vague here, but partly because the answer is vague. A genetic algorithm is, roughly, an algorithm for solving problems using techniques inspired by genetics. The simplest variants encode a possible solution as a string of symbols (so if we know the solution is a whole number, the encoding might be a string of 1s and 0s, read as a binary number) and have some way of saying this solution is "better" than some other solution (so if we want the biggest number, a solution is "better" if it encodes a bigger number). You randomly generate a number of potential solutions (called the population) then generate new solutions by randomly selecting pairs of solutions (biased towards "better" solutions), randomly combining material from each parent to create a new child solution (say, the first X symbols from one string, all other symbols from the second string - this is called crossover) and finally possibly randomly changing a small number of symbols in the resulting child string (called mutation). If the child gets the best bits from both parents, it will be "better" or "fitter" and therefore both a better solution and more likely to pass on its good bits to the next generation. There's lots of variants (how you do crossover, how you mutate, how you choose parents, how you encode, population sizes, etc.) including some that don't fit the simple scheme (solutions are graphs or trees or program code).

Why? We're talking computers having sex here - do you need a reason? Oh, alright. Genetic algorithms are good for exploring problems that are easy to define but hard to solve. They have the advantage that at any point you have a valid solution, albeit probably a sub-optimal one, so they work well when you don't know how long you have to find a solution. But they are difficult to write well, and they very, very rarely find optimal solutions.

4) Where do you see the future of computer sciences being?

Computing science is a strange beastie, split into two inter-related disciplines. The real science lies in understanding functions, computation, and how they interact on the level of maths and physics. There will be some real advances here, involving dna computers, non-binary computers, and quantum everything. There may also be some interesting advances in the pure maths end - primarily around the area of P=NP, and especially if anyone ever manages to prove the unexpected result of "yes, P does equal NP". (I'm not going to explain in any detail, but essentially if P=NP that means that some currently very difficult problems must actually have relatively easy solutions, and we probably will know what they are.)

Otherwise, computing science is really about engineering. That's needs lots of work. At the moment computers are engineered in much the same way that novels aren't. If we're ever going to get the cost and time over-runs on every big project under control, we need new paradigms, possibly involving automated code-generation, massively parallel systems based on heavily standardised hardware, and programming languages, code libraries and operating systems that are stable over decades if not centuries.

You did ask...

5) What's the WORST system you've ever played in? Why?

World of Darkness. It may just be the players I've encountered, but far too many of those playing Vampire or whatever turn into book-quoting, stereotype-playing, dull, boring, munchkins. Even those who are perfectly reasonable when playing other games. People saying "you can't do that 'cos it says in some book that you've never read that's out of print that..." just annoys me, so I don't do World of Darkness anymore.

Phew!

[<< Catching Up] [New Doctor Who >>]