Last Modified 5
Simulations aren't evil
In the old days of `New AI', one of the
things that helped define the field was the claim that simulations were
and that true AI could only be achieved on robots.
Why? Two reasons:
- Real intelligence isn't as complex as it looks. The
complexity of behavior is a reflection of the complexity of the natural
environment in which the behavior is expressed. If you take
away nature and construct a simple environment, then you will either
wind up putting
too much complexity in your agent or you will fail to make it look very
- In most of engineering, the hardest part of solving a problem is
understanding it. If you build a simulation, then you are
claiming you understand
the problem well enough to replicate that problem. Naturally you
be able to solve the problem you've created yourself! Simulations
you to either deliberately or accidentally over-simplify the problems
So if I know all that, why don't I think simulations are evil?
Well, basically, I've realized two more things:
I've learned a lot from working with robots (and I probably wouldn't
have developed BOD without having worked on
one robot platform for a year
and a half). But one of the things I learned was not to be a
- Nobody builds their own simulations anymore. `Simulations'
often real problem environments that are developed by a large number of
like the robocup simulation
league or unreal
tournament. So no one can cheat; these are real problems.
- Robots are also simulations. They don't have anything like
sensing or actuation of real animals, and they seldom run in realistic
environments. Arguably, a software agent that has to swim
in a fluid dynamics simulation with simulated vision may be more like a
real animal than a robot with wheels and a laser range finder.
I talk about what I learned from robots more in my PhD
dissertation, and why they aren't critical to AI more in a paper
Testing for Complex Agents". Both of these are available on
page author: Joanna Bryson
back to my links page