4 ms·
It's the response to Greenspun's tenth rule that mentions Prolog explicitly. My previous comment states that AI programs written in Lisp implemented an interpre
by YeGoblynQueenne 8y ago
It's the response to Greenspun's tenth rule that mentions Prolog explicitly.
My previous comment states that AI programs written in Lisp implemented an
interpreter for a logic programming language, unspecified. Although to be
fair, that usually means "an informally-specified, ad-hoc, bug-ridden, slow
implementation of Prolog" (1).
Now, my knowledge of Eliza, Shrdlu, etc is a little limited (I've never read
the source, say; btw, have you?) but let's see. Wikipedia is my friend, below.
According to wikipedia MYCIN was a backwards chaining expert system. That
means a logic programming loop, very likely a resolution-based theorem prover,
or, really, (1).
ELIZA was first implemented in MAD-SLIP, which, despite the name was not a
Lisp variant (although it was a list-processing language). STUDENT is
actually part of ELIZA- it's one of the two scripts that came with the
original implementation, by Wizenbaum. The other script is DOCTOR, which is
the one more often associated with ELIZA (it's the psychoanalyst). If I
understand correctly, STUDENT solves logic problems, so I'm guessing it
incorporates an automated theorem prover; so basically, (1).
Eurisko was written in a frame-based (think OOP) representation language
called RLL-1, the interpeter for which was written in Lisp.
SHRDLU was written in Lisp and MicroPlanner an early logic programming
language (with forward chaining, if memory serves). In the event, (1) was not
necessary, as MicroPlanner was already all that- and more!
The Automated Mathematician (AM) was indeed written in Lisp, but I don't know
anything about its implementation. However, from the description on its
wikipedia page it appears to have included a very basic inference procedure
and a rule-base of heuristics. Sounds like a case of (1) to me.
- le-mark 8y agoThanks for taking the time to break that down.
- varjag 8y agoMYCIN is a fuzzy logic inference engine, among the first ones. Would love to see that implemented in period-true Prolog, but tbh won't hold my breath. RLL-1 and Microplanner are domain-specific languages, traditional Lisp approach to building complex applications. Fact is both were written in Lisp, that's just sophistry. Neither AM nor EURISKO would map naturally to unification with their very special heuristic-driven task prioritizing and scheduling algorithms.
- YeGoblynQueenne 8y agoI don't know what "period-true Prolog" is. You can still run programs found in Prolog textbooks from the 1970's in modern Prolog interpreters- just like with Lisp. >> Fact is both were written in Lisp, that's just sophistry. I don't see the sophistry. If I implement Lisp in Prolog, the interpreter may be Prolog, but the programs it then evaluates are Lisp. I don't see why unification would be an impediment in implementing any heuristics or scheduling algorithms.
- varjag 8y agoPeriod-true as in from 1970s. There were attempts to build fuzzy Prolog dialects in late 1980s, although it seems not particularly successful. > If I implement Lisp in Prolog, the interpreter may be Prolog, but the programs it then evaluates are Lisp. But it does not implement Lisp in Prolog. It implements a domain specific language in Lisp in a time proven manner. Googling "DSL Lisp" will give you probably hundreds of references. I did it before for my own projects and my "languages" even had no name. I could call it VARLANG-22 or whatever, but unless I told you all you'd see is some Lisp code. > I don't see why unification would be an impediment in implementing any heuristics or scheduling algorithms. It does not bring anything to the table there, because that's not how those systems work. Not every project is a first order predicate logic expression, as incredible as it sounds.
- YeGoblynQueenne 8y agoWe're in a weird situation here. In this thread, you're arguing that a programming language implemented in Lisp, which retains the syntax of Lisp, is a DSL, and so still just Lisp; but in the other thread, you're arguing that Prolog implemented in Lisp, with the syntax of Lisp, is not a DSL, but Prolog. I recognise that the above comes across as an attempt at sophistry, again, but I sincerely think you are not being objective, with this line of thinking.