4 ms·

I consider CS the most important part of this project and the real open problem. I thought about doing this project in LISP, because dealing with custom-made de

by zatara 16y ago

I consider CS the most important part of this project and the real open problem. I thought about doing this project in LISP, because dealing with custom-made devices will require code generation and lots of AI. I'll have to translate lots of surgical experience and rules of thumb into the code, so Python doesn't seem to be enough (or maybe I'm just unexperienced). Besides that, I really control all the environment so deployment is a non-issue. Does my rationalization makes sense?

ableal 16y ago
To me, I'm sorry to say, no. From experience, I've become a believer in "Amateurs talk about strategy, professionals talk about logistics". I'd suggest sticking to Python, a clean and pleasant language that 'fits my brain', to solve the problems you do have, until you bump into something that needs more. If/when what you can get for Python (from constraint solvers to whatever) proves insufficient. Also, lookup "Beware of the Turing tar-pit". And see what Peter Norvig ( http://norvig.com http://norvig.com ) has to say. On the other hand, you may consider rule systems (see, e.g. http://news.ycombinator.com/item?id=1363680 http://news.ycombinator.com/item?id=1363680 ); but consider also how you'll debug and maintain them.
dkarl 16y ago
Honestly, it did not even cross my mind that you would need to do any sophisticated programming. I assumed that most of the challenge would be in mapping and mastering the problem domain: figuring out what problems you need to solve, researching existing solutions to each problem, finding code and adapting it for your needs, experimenting to see if the existing solutions work well for your particular problems, adapting them if necessary, and then wiring it all together in a simple way using a glue language. Lots of work and sophisticated AI stuff there (probably a few PhDs if you took it far enough) but no fancy programming. That seems to be the sad truth about real world problems: the programming is rarely as hard and interesting as all the other problems you will solve. On the other hand, Lisp is a lot of fun, and you shouldn't pass up any excuse to learn it ;-)

by zatara 16y ago

ableal 16y ago

To me, I'm sorry to say, no. From experience, I've become a believer in "Amateurs talk about strategy, professionals talk about logistics". I'd suggest sticking to Python, a clean and pleasant language that 'fits my brain', to solve the problems you do have, until you bump into something that needs more. If/when what you can get for Python (from constraint solvers to whatever) proves insufficient. Also, lookup "Beware of the Turing tar-pit". And see what Peter Norvig ( http://norvig.com http://norvig.com ) has to say. On the other hand, you may consider rule systems (see, e.g. http://news.ycombinator.com/item?id=1363680 http://news.ycombinator.com/item?id=1363680 ); but consider also how you'll debug and maintain them.

dkarl 16y ago

Honestly, it did not even cross my mind that you would need to do any sophisticated programming. I assumed that most of the challenge would be in mapping and mastering the problem domain: figuring out what problems you need to solve, researching existing solutions to each problem, finding code and adapting it for your needs, experimenting to see if the existing solutions work well for your particular problems, adapting them if necessary, and then wiring it all together in a simple way using a glue language. Lots of work and sophisticated AI stuff there (probably a few PhDs if you took it far enough) but no fancy programming. That seems to be the sad truth about real world problems: the programming is rarely as hard and interesting as all the other problems you will solve. On the other hand, Lisp is a lot of fun, and you shouldn't pass up any excuse to learn it ;-)