The garden of Eden

[hoserofdoom]

In research, there are two ways of looking at a project: the hype, and the hangover. Some researchers are really good at hype, and fuel the development of science fiction. Some researchers are really good at giving others hangovers by pointing out obvious flaws. Both approaches are needed, but I find that the former really annoys me.

Comments

[ghap]

So, if your society starts worshipping you as the one true creator god and starts crafting theological explanations for the various curses and blessings you send to them, would it be considered a success or failure condition?

creator god

[hoserofdoom]

Unlikely. The agents won't have anything like abstract thought going on... so they won't be able to attach a bunch of beliefs to some agent and tell stories about it. Thinking will be much simpler: "in this situation, my linear model predicts that this action will lead to the best outcome".

I have grand plans for the future, where this simple model may be enhanced by a rudimentary linguistic model, but there are some serious computational costs involved that I don't think current hardware would be able to handle, even with all of the short-cuts I intend to take.

Hangover: Part the first

[ghap]

I like the idea, but I doubt it would work.

I agree that human society follows emergent patterns, created by individuals making locally rational decisions. However, I doubt basic mathematical/computational tools would do a good job modelling the situation well enough to be predictive. It seems highly unlikely to me that you could get this right on the first iteration, or even on the x number of iterations you could reasonably expect to be able to try before the grant money runs out.

But I'm not an expert on this sort of modeling. I suppose no one is yet.

I have ideas about as crazy as this, but I haven't found the energy/guts to clarify them and report them to anyone. Your ideas are an inspiration :p

P.S.: All Hail The Demko. The Demko givith, and The Demko taketh away. Blessed be the name of The Demko.

Re: Hangover: Part the first

[hoserofdoom]

I haven't posted my hangover yet... but the truth is that I think I'll hit at least one of the emergent properties if I implement enough of the system. It won't be as pretty and complex as we might dream of, but it will be there. I've done similar things in the past, with some success. I'm not just building castles in the air here. I'm also very devoted to this project, even if I'm only throwing about 100 hours a year into right now.

What I'm really hoping for is to hit a couple of emergent properties in the same simulation... that would be a first for me. Yes, a lot of tuning will be required, but that's where challenge is.

Re: Hangover: Part the first

[hoserofdoom]

The really far-fetched stuff will come in a few posts, when I start to talk about the language model.