Re: [sl4] How much do languages matter for AGI programming?

From: J. Andrew Rogers (
Date: Sat Nov 29 2008 - 13:07:55 MST

On Nov 29, 2008, at 2:29 AM, Philip Hunt wrote:
> 2008/11/28 J. Andrew Rogers <>:
>> How do you know what paradigms are needed for AI? And what about the
>> paradigms LISP does not do well? Why not use a more paradigm-agile
>> language
>> (e.g. Python)?
> Surely Lisp is more paradigm-agile than Python, because if Lisp
> doesn't support a paradigm, you can extend it until it does, using
> macros.

This is essentially true of most languages, as they are all capable of
supporting whatever paradigm or idiom one wants to use with thick
enough abstraction layer -- most programming languages have been
implemented in C, after all. In the specific case of Python, while it
does not have macros it *does* have other first-class facilities that
can accomplish similar things easily.

By "paradigm-agile" I really meant "idiom-agile", observing that the
efficient idiomatic expressiveness of Python is broader than most
common languages, and most simple, elegant languages (LISP, FORTH,
etc) can be implemented in a page of Python code as a result even
though, for example, some FORTH-like idioms are ugly in LISP-like
languages and vice versa. Useful, if a problem is so poorly defined
that the most efficient idiomatic expression is not obvious.

Which is not to say that LISP is not ideal, only that we do not have
much evidence that the natural idioms of AGI are not suboptimal or
ugly if implemented in LISP, mostly because the AGI computational
model has not been defined.


J. Andrew Rogers

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:03 MDT