Why is/was LISP good for AI programming?

[Stephen Tashi]

TL;DR

Why is/was LISP thought to be good for AI programming?

Why is/was the LISP programming language thought to be good for writing programs that implement artificial intelligence?

As I imagine it, early AI programs were oriented toward verbal (i.e. text) inputs and outputs. So perhaps a list processing language is/was well suited for that?

 

[jedishrfu]

I think the symbolic manipulation feature was why. You could write a lisp program that wrote and ran a lisp program ie you could dynamically change the programs processing steps.

 

[FactChecker]

Lisp is a very old language, contemporary to languages like FORTRAN. When AI started becoming popular there were few alternatives that could handle non-arithmetic problems. As @jedishrfu says, the ability to perform symbolic and non-arithmetic operations was considered important for AI. Also, as @jedishrfu says, the ability for the program to piece together code on the run and evaluate it was useful. That ability also made FORTH a language to consider.

 

[jedishrfu]

I always felt Forth was Lisp written RPN.

Lisp was also known as ‘Lots of InSipid Parentheses’

 

[anorlunda]

I always felt Forth was Lisp written RPN.

apt descriptor. But Forth seemed so deceptively simple.

 

[jedishrfu]

Forth gave new meaning to the Devils in the details.

 

[FactChecker]

Lisp was also known as ‘Lots of InSipid Parentheses’

And anyone who ever had to work with Lisp code in a regular, non-Lisp-oriented, text editor, has to agree.

 

[jedishrfu]

Well they did introduce the close bracket feature.

 

[jtbell]

Lisp was also known as ‘Lots of InSipid Parentheses’

The way I always saw it was "Lots of Stupid Parentheses" or "Lots of Silly Parentheses".

 

[jedishrfu]

I did an acronym search and found others:
- list processing (formal acronym for LISP)
- lost in a sea of parentheses
- lots of { irritating | isolating | infuriating | insipid } { stupid | superfluous } parentheses
...

Lisp lives on in Clojure a mashup of lisp and java running on a jvm.

https://en.wikipedia.org/wiki/Clojure

and then there's TRAC programming

https://en.wikipedia.org/wiki/TRAC_(programming_language)

 

[gleem]

A reader of this thread suggests LISP is a dead language. However, it was developed for particular applications and as such is a better language than the popular C's, Java's, or Python.

Here is a good review of its utility. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5952920/

 

[Mark44]

Here is a good review of its utility. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5952920/

Interesting article...

 

[jedishrfu]

I had a friend pull up a similar article years ago on Clojure aka lisp on JVM I tried it but felt my team would never consider it even though there were some great benefits. Sometimes procedural / OO folks are a hard crowd to please.

 

[elcaro]

Summary:: Why is/was LISP thought to be good for AI programming?

Why is/was the LISP programming language thought to be good for writing programs that implement artificial intelligence?

As I imagine it, early AI programs were oriented toward verbal (i.e. text) inputs and outputs. So perhaps a list processing language is/was well suited for that?

No. Recursion and meta-progamming and functional features were the pro's of LISP.

 

[FactChecker]

A reader of this thread suggests LISP is a dead language. However, it was developed for particular applications and as such is a better language than the popular C's, Java's, or Python.

Here is a good review of its utility. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5952920/

Considering that Perl was a significant programming language of the Human Genome Project, I can't help but notice that the referenced article is very vague on any advantage that Lisp has to Perl. As a Perl programmer, I came close (although overcome by events) to joining that effort.

See How Perl Saved the Human Genome Project and BioPerl for more details.

 

[jedishrfu]

I remember reading a Dummies book on Bioinformatics and it mentioned Perl a lot. It struck me as funny because Perl was on the outs and python was ascending for general programming use.

Often academic projects latch onto programming languages based on what the grad student or principal investigator was familiar with instead of what’s the best overall language to use. Bioinformatics was very text search intensive making Perl an ideal choice but later as other tools gain traction Perl loses to python or some other more performant language. Sometimes choices are made on the free tools and libraries the project latches onto.

I never understood why Clojure/Lisp was gaining adherents when it seemed easy to really get lost in the details of the implementation of an application. Small programs were elegant by scaling up presented developer issues for maintainability. The article my friend had provided said that this startup team had switched from Scala to Clojure and saw huge benefits in maintainability and performance but again I never fully understood why they just didn’t use Java or something similar.

 

[jtbell]

The way I always saw it was "Lots of Stupid Parentheses" or "Lots of Silly Parentheses".

Argh, I left out a word. Both should begin, “Lots of Insane…”

 

[gleem]

For an interesting exploration into the mystical characterization of LISP read

How Lisp Became God's Own Programming Language​

It's not too long.

 

[Vanadium 50]

I haven't touched LISP in years, so I will only make two comments:

1. The fact that LISP ignores trailing parentheses is a mistake. That encourages the programmer to stick a zillion at the end, thus covering up a logic error.
2. There is AI and there is AI. Modern AI is not so concerned with what LISP does well - what I would call machine deduction" - as 1980's AI was.