But -- in general -- writing APIs that allow the AI to easily get as much useful information so that the AI programmer can easily put together strategies is crucial.
Actually, this is the part I'm more interested in: how to go about writing an API? I'm about to take on a BIG task. To the tune of writing something like OpenGL. And I don't mean writing IN OpenGL--I mean writing OpenGL. But I've always been more into developing the video card than I have writing the C libraries interfacing to the video card. How do you go about collecting requirements definition from your (potential) customer base? And then organizing/software engineering that into an implementation, with an under-the-hood data model and an interface to the input vectors the hardware is expecting?
Anyone know any good books/white papers on API library development? Particularly, where you're interfacing with assembly code or hardware drivers. thanks!
Writing a good API for something requires a very deep understanding of two things:
- the problem the API solves, the functionality you want to expose,...
- how the clients will use the API, what they need to work, what they don't need,...
That's hard, very hard 