Yann LeCun and others created https://lush.sourceforge.net, which was a Lisp for ML.
It was probably the best option during the 2000s, and only superseded when Theano (Python) and Torch (Lua) emerged.
A Lisp comeback in this niche would be great. Python has fantastic libraries, but the language leaves a lot to be desired.
LLMs have trouble counting parentheses accurately. ;)
We need a lisp that is written in JSON. That should fix it
It would REBOLutionize everything.
First version of Torch was in LISP. Python is where people are, and in turn a data for LLMs.
Optimal ML lang is half-way from LISP to APL.
Because it got burned in the last AI hype cycle, and many devs still think Emacs + SBCL is the only way, never even bothered to open the websites of LispWorks, Allegro Common Lisp or Clozure (not the JVM one).
If I make the assumption that future ML code will be written by ML algorithms (or at least 'transpiled' from natural language), and Lisp S-Expressions are basically the AST, it would make sense that it would be the most natural language for a machine to write in. As a side benefit, it's already the standard to have a very feature complete REPL, so as a replacement for Python it seems it would fit well