genetic programming is back! (in my mind)

[gusl]

I used to think of genetic programming as "dumb", producing bloaty, hacky solutions to simple problems. When I think of "automatic programming", however, I think of clean, high-level languages and macros, which are basically better tools for making human-produced code. But why not mix the two? Have algorithms play with your clean code and see if they get you anywhere.

Apparently genetic programming has gotten smarter than I thought:
Genetic Programming II: Automatic Discovery of Reusable Programs.

Genetic programming has 16 attributes of what is sometimes called automatic programming or program synthesis or program induction

Trends and Controversies - Genetic Programming

Comments

[vi-z.livejournal.com]

There is a way to write short elegant programs; you just search the space in the order of increasing complexity. You can search through Juergen's page www.idsia.ch/~juergen for "OOPS" and skim through some papers and presentations. It has its own problems, though.

[vi-z.livejournal.com]

OFFTOPIC: I would be on IAS-8 in Amsterdam in March, you seem to live in Netherlands?

[gustavolacerda.livejournal.com]

Yup, I live in Amsterdam, not too far from the venue of the conference. (I had no idea about the conference, btw: I just looked it up). If you email me at lj at optimizelife.com, we can arrange to meet up, and I'll show you around the city.

[radiantsun.livejournal.com]

You might also want to google "John Koza" and genetic programming. He spoke at the IAC 2003 conference. He was pretty good.

Re:

[mathemajician.livejournal.com]

With this user icon I didn't even recognize you for a moment...

"smarty pants" indeed.

[mathemajician.livejournal.com]

"producing bloaty, hacky solutions to simple problems"

but aren't YOU a solution due to this process?

[gustavolacerda.livejournal.com]

but aren't YOU a solution due to this process?

Sorry, I can't parse that. My natural language module is pretty restricted.

Re:

[mathemajician.livejournal.com]

Your brain could be viewed as an algorithm that was developed by a long biological experiment in genetic programming right?

In which case, referring to the solutions created by genetic programming as "bloated and hacky" is in fact an insult to yourself!

[gustavolacerda.livejournal.com]

oh... I see what you mean. Yes, I had a very restricted view of GP.

Then, my question is: what am I a solution for?

"making babies who will make babies", etc. is a means, but to what end?

Re:

[mathemajician.livejournal.com]

Who said there should be an end?

That's the beauty of evolution... it is its own end.

Re:

[gustavolacerda.livejournal.com]

It sort of begs the question of intelligent design... which I am absolutely agnostic about. Is it even falsifiable?

btw, if you believe that our children are very likely to simulate their descendants (seems reasonable if you believe in strong AI), then by anthropic selection it's very likely we're in that sort of "Matrix" (there can be millions of simulations, but only one "real" world)