When I said I found out “a while ago” – I did originally discover it yonks back, forgot, remembered whilst talking to Ian, forgot, and discovered it again a month before this blog post.

May do an updated version someday, especially if a simple API was created… though currently working on a rather more complex python app that finds optimal arrangements of “items” that have preferences on which other “items” they’re near, for a given configuration of “slots” that these “items” can be placed in. Savvy?

I think a flag is actually easier, given that there’s only really one that matters (-d). -r is there for completeness but it defaults to a random lie anyway. I’ve always preferred these to the longer-hand form that apt/bzr/etc use.

1) It’s best to write python scripts like C, with a main() function, and then put

if __name__ == “__main__”:
main(*sys.argv[1:])

at the bottom. This makes it feel more like a program than a hacky script (even though it’s no more effort), and encourages you to write it in a clean, readable top-down style.

2) If it’s xml/xhtml, then try using xml.etree.ElementTree.parse() (etree’s API is a lot nicer than DOM) or something. Rexexps and hard-coded line numbers are fragile.

3) Try using “if ‘-r’ in sys.argv:” if a flag is really what you want. single-letter flags are very old-style unix though. I’m not going to recommend optparse for this because it’s overkill, but maybe an aptitude/svn/bzr-style subcommand (like $ tellalie.py random or $ tellalie.py daily) might be more appropriate in this day and age.

4) Please Please Please read pep8: http://www.python.org/dev/peps/pep-0008/. Style is very important, especially if you’re ever thinking of reading your code in the future. Obvious thing include comments in the wrong place and multiple imports on the same line.

5) It is currently recommended that you write API-compatible 2.6 versions of any python3 modules you write (so i’d be surprised if urllib/urllib2 didn’t have some backwards compatible functions. In the worst case, you can wrap your imports in try/catch). If you want to “from __future__ import print_function” then that’s cool too.

Holly says she’s getting me into Mills next. It’d say that my utility function approach looks a lot more like general “egoism” than specific “utilitarianism”, but I will have to read up on both if I want to convince anyone that utilitarianism isn’t what I mean.

If you wish to maximise the integral over all time of p(at least 1 thing similar to yourself existing at time t), then I think you would probably get utilitarian behaviour emerging as the best way to act. Now I just need to prove it.

Biologically, the “evolution utility function” is, unfortunately, the only real one. Life only exists to propagate itself – there’s no higher aim in a biological sense. However, since we *are* here anyway, and happen to have self-awareness, there’s no reason why we can’t layer other things on top.

I shall look over your discussion so far when I have a mo… probably not for a couple of weeks now, I’m afraid But “+5 insightful” on your first comment anyway, for putting something vaguely like what I rambled on about into a more specific mathematical format

I think that all aims should be able to take the form “maximise E[u(world)]” where “world” is the state of the world at all times in the future, “u(world)” is your utility function and “E[.]” is the expectation operator.

One person’s utility function might be
u(world) = A*health + B*wealth + C*wisdom
and another might be
u(world) = E*sexual_activity + F*drug_intake + G*rockandroll_volume
(where rockandroll_volume is a real-valued number between 0 and 11)

I’m trying to think of the simplest possible utility function (preferably well-behaved/measurable). One of the ones that came out of our discussion was the evolution utility function u(world) = count(copies of yourself in the world) or similar, but I don’t think any of us really like that as the “meaning of life”.

For now, we’re looking at the question “There exists a utility function. Does it have to include the happiness of others in order to make us act morally?” I think the answer is “no”, but I’m waiting for Alex’s blog post on the subject.

If you feel like contributing, give us a shout.

d) Ask which orifice you take the tablets through, asking questions of the form “So, can I take them through my ear/nose/etc.?”