Impressions on B1 Machine

[Ivan Krstić]

Mitch Bradley wrote:
> Which suggests to me that it would be better to have a compelling game
> that people like to play as-is, and make it easy to make simple mods
> without having to first learn a complex framework.

This reminds me...

I don't know if anyone here remembers the Cybug system, (I think now
called 'A. I. wars'). It was a game that had me hooked many years ago,
with a very simple idea: you write a little robot in a scripting
language that lets you sense robots around you, move around, and shoot.
Then you pit your (fully autonomous, pre-programmed) robot against
either other robots that came programmed with the game environment, or
against other human-programmed robots in live tournaments.

This was pretty tremendous fun, because you could run the game out of
the box by pitting pre-programmed robots against one another, then look
at how they work, change a single line, start another match, and so
forth, eventually getting to the point where instead of modifying an
existing robot, you write your own from scratch.

In terms of kid appeal, I taught the Cybug system to 7th and 8th graders
with no previous programming knowledge, and they got incredibly deeply
into it.

I think it'd be *fantastic* if someone wanted to take up an
implementation of a cybug-like system in Python, with a subset of Python
as the robot implementation language. There's an IAP class at MIT,
6.370, that does the same thing under the name 'RoboCraft':

    http://battlecode.mit.edu/2007/info

They even do fancy things like give robots 'fuel' which is used to pay
for executing Java VM instructions, so that robots have to make careful
assessments of when it pays to do more complex calculations. Someone
could really go crazy with a system that does this on the OLPC.

-- 
Ivan Krstić <krstic at solarsail.hcs.harvard.edu> | GPG: 0x147C722D