Thursday, July 24, 2008

Massively parallel algorithms, agents and game theory

I'm getting bored of programming in high-level programming languages.
I'd like to do more with writing less code: there is too much distance between modeling the problem domain and getting a prototype simulator of it...

So I'm looking into using agents and how to let them do the work. I have an intuition that things are different when you build systems from agents instead of just coding programs:
- instead of algorithms you get (concurrent) protocols
- time (and timing) becomes an issue
- instead of sequential execution model, I need to think in terms of swarms and game theory
- users (and programmers) are part of the system, they are agents too
- how to create open, extendable systems
- possibly chaos theory would be needed to analyze stable behaviors
- from Frankenstein (building dead code, then run it) to vets (fix running code)
- who cares about termination
- when an agent is fully configured (to find me the best porn sites there are) why should update it and start from scratch its configuration? How about my current (old) agent tells all its tricks to the new agent (rookie)?
- (programming) language support for all this?

Some links:
- Game theory vs. multiple agents: The iterated prisoner's dilemma (here)
- Agent-Based Modeling: A New Approach for Theory Building in Social Psychology (here)
- Agents that solve complex games (here)

No comments:

Post a Comment