Notes from my terminal

Why the AI community should stop whining and realize that they've already won

There are things in life that are certain, among them death, taxes and the lamenting of LISP hackers that everything would be so much better if only the world would follow their path to enlightenment. These "But-LISP-already-does-that"-articles almost form a literary sub genre in computer texts with classics such as Richard P. Gabriel's "Good News, Bad News, How to Win Big" or "Revenge of the Nerds" by Paul Graham.

While it's understandable that our world with its PC architecture and MS Windows operating systems must be a constant source of frustration for those who know that we used to have so much more advanced systems like LISP machines, I think that the survivors from the Golden Age of AI miss an important point: The world already is a better place. Their favourite languages (Prolog and LISP) are in wide use every day. However, most people don't realize that it's really dialects of LISP and Prolog what they use when publishing on the Net or designing web applications. Prolog and LISP managed to sneak into mainstream, thanks to XML. Read on to find out why.

Two of the most exciting technologies in the XML world are XSL, the Extensible Stylesheet Language, and RDF, the Resource Description Framework. If we think of these two as programming languages, they are variants of LISP and Prolog, respectively.

XSL is a simpler and gentler form of a SGML stylesheet language called DSSSL (Document Style Semantics and Specification Language), an extension to the LISP dialect Scheme. Much like SGML, it suffered from the fact that it was simply too complex for average coders. As SGML became popular in its stripped-down, cleaned-up version XML, DSSSL became popular as XSL (the fact that these scary-looking brackets were done away with also helped a lot).

Still, it's easy to see XSL's LISP heritage. XSLT is a functional language, so functional in fact that variables cannot be changed once they are assigned. If you want do get anything complex done in XSLT, you'll eventually come to a point where you realize that, as in LISP, certain tasks are actually easier solved by using recursion. Even XML itself is close enough to LISP notation and a transformation is trivial. It's also probably not a big surprise that many recipes in Sal Mangano's XSLT Cookbook are ported from Common LISP.

RDF, or generally speaking, the whole Semantic Web concept is very hard to grasp. However, the moment when I started "getting it" was when I realized that all the abstraction about RDF triples really boiled down to what is done to describe facts in a Prolog expert system. RDF is really Prolog. This is an obvious thing once you think about it: Describing facts and knowlegde is what meta data is about. I could go on how much of RDF is actually Prolog with angle brackets, but fortunately it has all been written before, by someone who knows more about both RDF and Prolog than me. It's just a logical consequence that SWI-Prolog, the free Prolog implementation, has packages for RDF and the Semantic Web available in its default distribution.

Great ideas don't die, sometimes they just need their time. LISP and Prolog live on, they just look different. I can only speculate on the reasons why they didn't make it in their original form, but maybe it's just because angle brackets look so much sexier. :)