-------------------------------------------------------------------------
Note: this document is not entirely up-to-date.

The AI hasn't changed much recently, since John has stopped working 
on it.  The main change from the information below is that an "easy" 
skill level has been implemented, and made the default.  Also, the
AI now uses diplomats and spies, for defense only. --dwp

Old README.AI from John Stonebraker, from the 1.7.1 release:
-------------------------------------------------------------------------
    I haven't worked on the AI as much for the past three weeks as
I did for the previous three months.  I did substantially improve
its ability to manage invasions and conquer cities, and prevent
misbehaviors that I doubt anyone else would have realized were bugs.
The AI seems capable of winning against experienced players now,
although to my dismay it still hasn't beaten me.

    The AI is getting so complicated that I'm sure there are problems
that I haven't encountered.  If it does something stupid or questionable
that isn't referred to as a Known Problem, send me mail mail describing
the problem and attach a save file so I can take a look at it.  If you
have opinions on which of the below are most important, send me mail
about it.  I'm always happy to get user feedback.

    The AI difficulty levels aren't fully implemented.  I've done all my
work getting the default level to work right, and based on what I've
heard, I defined that level as 'hard'.  The only available alternative
is 'normal', where the AI doesn't know where huts and military targets
are before it sees them, and also must obey government max rates.
It seems to behave sensibly in the few tests I've run in normal mode,
and I haven't gotten any complaints that there are problems with it.
There is no 'easy' mode yet, although I suspect there needs to be.

John Stonebraker
ppaqebb@surfshop.net

Known problems:
-------------------------------------------------------------------------
AI doesn't understand when to become DEMOCRACY
Cities don't realize units are on their way to defend it
AI doesn't understand obsolescence
related: AI likes to build Oracle more than Chapel
AI doesn't know how to use spies and diplomats, even for defense
AI doesn't know how to use planes and missiles
AI doesn't understand how to favor trade when it needs luxury
AI builds cities without regard to danger at that location
city tile values are not cached; wastes CPU time
Food tiles should be less wanted if city can't expand
AI doesn't build infrastructure near where a city is about to be built
ai_calc* routines are slow and perhaps shouldn't regard government
AI won't build cross-country roads outside of city radii
Non-military units need to stop going where they will be killed
GOTO doesn't popup diplomat/caravan dialog when city is reached
Locally_zero_minimap is not implemented when wilderness tiles change
AI doesn't understand 13th city is unlucky
Settlers won't treat about-to-be-built ferryboats as ferryboats
If no path to chosen victim is found, new victim should be chosen
Values for Mysticism, Electricity, and Theology need to be verified
AI doesn't know how to make trade routes or when
AI is vulnerable to well co-ordinated mass attacks
AI doesn't acknowledge effects of MAGELLAN in some of its calculations
Boats sometimes sail away from landlocked would-be passengers
Ferryboats crossing at sea might lead to unwanted behavior
Emergencies in two cities at once aren't handled properly
Auto-settlers never change terrain types

Partially fixed problems:
-------------------------------------------------------------------------
AI handicaps are minimally implemented
AI manages workers one at a time, which is not as dumb as it used to be
Transport paranoia code is mostly working but not 100% reliable
Auto-settlers have a simplistic idea of enemy territory

Idea space:
-------------------------------------------------------------------------
Friendly cities can be used as beachheads
It might be best to only sentry the #1 passenger and its bodyguard
Unit_move_turns should be implemented more thoroughly
Assess_danger should acknowledge positive feedback between multiple attackers
Urgency and grave_danger should probably be ints showing magnitude of danger
It would be nice for bodyguard and charge to meet en-route more elegantly
It may be correct to starve workers instead of allowing disorder to continue
Bodyguards could be used much more often
AI should sell barracks (and other worthless buildings) voluntarily