I always find Lisp a bit mind bending and feel like I will be a better programmer once I master it. But it also feels like wood carving, and most of the time I just need to slap some planks together for a bookcase, and don't have the time to scrimshaw it into a beautiful work of art. I'm not seeing here how Lisp is differentiated from an agent with a code sandbox and python, unless we're operating that agent inside of Emacs... That's where the code and the data are swimming together. I don't think there isn't a difference, it's more a matter of my not getting it, if anyone is able to help me understand it'd be appreciated
post
He doesn't appear to be using any features of Lisp that made it "the language of (70s) AI". You could do all of this just as easily in many other languages.
His prof was still wrong.
Yeah the intelligence is still in the model. The promise of symbolic AI is about logic programming/ formal semantics not recursive loops.
To a large extent the idea has failed because it proved too hard to get non-experts to represent systems formally.
I still think there's potential value in a hybrid approach - e.g. get language models to do the representation then let them use formal reasoners/ verification instead of hallucinating.
Sometimes restrictions are features. What if you don't want to hand out shell access? Then you are back to a list of tools again.
Lisp is a cool language. It was many years ago now, but I recall writing a lisp interpreter in lisp for a college project. It was only ~100 lines of code and my recollection is that it worked remarkably well.
all 10 comments