I've just read (quite a bit of) the very interesting (long) document on Ultra-Large-Scale (ULS) systems.
As you probably realise, the Semantic Programming research project is trying to address very similar issues to those explored by this ULS system research.
I guess a big area of disagreement, however, is that I don't have as much faith in genetic algorithms as Richard Gabriel clearly does.
Indeed, I think his MetaDesign2 / MetaDesign3 distinction doesn't look that strong as evolution is just large amounts of trial and error: MetaDesign2.
So, I think we need humans not computers doing large amounts of tinkering on a massive, peer to peer, interconnected system.
3 comments:
Hey Phil,
I've just read (quite a bit of) the very interesting (long) document on Ultra-Large-Scale (ULS) systems.
As you probably realise, the Semantic Programming research project is trying to address very similar issues to those explored by this ULS system research.
I guess a big area of disagreement, however, is that I don't have as much faith in genetic algorithms as Richard Gabriel
clearly does.
Indeed, I think his MetaDesign2 / MetaDesign3 distinction doesn't look that strong as evolution is just large amounts of trial and error: MetaDesign2.
So, I think we need humans not computers doing large amounts of tinkering on a massive, peer to peer, interconnected system.
We need Semantic Programming :)
Oli
I'm writing a longer respone to the Gabriel work, just haven't finished it yet.
I tend to agree that you can't rely on some naive kind of GA and you need human intelligence in your distributed design process.
OTOH, understanding of GAs and artificial evolution can definitely be helpful in understanding what you want from your distributed design process.
I'm looking forward to seeing how SemProg solves all this, of course.
Post a Comment