On Tue, Sep 11, 2012 at 6:03 PM, Michael Carpenter
On 09/11/2012 04:21 PM, Darren Grant wrote:
This made me think that it could be much more effective to develop AI in a functional language. There's no way I could do this with Haskell presently as I am still struggling to approach all problems from the FP perspective first, but I do think there is the potential.
There is a lot of potential here in my opinion. The two language families that have built their reputations through their use in AI research, Lisp and Prolog, share a lot in common with Haskell. AI, along with a lot of the big problems out there, seem to always boil down to parallel relationships between sets of data in a model rather than sequential object-oriented recipes. I would not be surprised if functional languages like Haskell supersede many of the imperative languages because of these problem sets. Sometimes I think I'm stupid to say that, but then I remember what SQL and RDBMS's did for the database world.
The parallel relationships point is interesting. I think sequential cases could still be helped by FP as well. I look at things like OpenGL wrappers for Haskell, which successfully solve immediate interop problems, but then wonder about revisiting whole approaches by applying new techniques. For instance, what if you cut out the existing OpenGL API and feed an infinite list into the device command buffer? It seems like there should be this potential for cutting latency and improving bandwidth through system comprehension tools, but that this sophistication has not become accessible to application programmers. I know there are plenty of experts hard at work on these problems, but they are interesting.
- Michael
_______________________________________________ Beginners mailing list Beginners@haskell.org http://www.haskell.org/mailman/listinfo/beginners