Rendered at 16:18:48 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
znnajdla 27 minutes ago [-]
One major point of disagreement: LLMs are not necessarily better at writing React because they have seen a lot of React. In fact, some benchmarks which ask the question: “which languages are LLMs better at” often show that LLMs excel at certain rare esoteric languages like Elixir, and can often solve more benchmarks problems when they write Elixir compared to more popular languages like Python. Even though the LLM has been trained on way more Python. So what makes LLMs better at a certain programming language isn’t how many examples it has seen. Frontier models can write very good code in a new programming language it has seen for the very first time, just from a good spec.
The real question here is: what features of a programming language make it a good language for Large Language Models? And how would you design a framework that’s not meant to be written or read by humans, only AI? That is a very interesting question to me. I theorize that LLMs are good at writing Elixir because it doesn’t have OOP concepts that require abstract models “in your head” (the LLM doesn’t have a head) and physical world metaphors that only humans have experience with. Instead Elixir’s primary tool is “pattern matching”, which works on literal text patterns instead of abstractions, and that’s what makes LLMs good at it. The question “what features of a language make it good for AI” is very interesting to me and I foresee a lot more research in this area.
bayarearefugee 51 minutes ago [-]
> The asteroid currently hitting front end web development
...is the same asteroid currently hitting all forms of knowledge work and even more theoretical pursuits (math research, etc).
Different sets of people are just noticing/reacting to the impact at different speeds, but we are all standing in it.
The real question here is: what features of a programming language make it a good language for Large Language Models? And how would you design a framework that’s not meant to be written or read by humans, only AI? That is a very interesting question to me. I theorize that LLMs are good at writing Elixir because it doesn’t have OOP concepts that require abstract models “in your head” (the LLM doesn’t have a head) and physical world metaphors that only humans have experience with. Instead Elixir’s primary tool is “pattern matching”, which works on literal text patterns instead of abstractions, and that’s what makes LLMs good at it. The question “what features of a language make it good for AI” is very interesting to me and I foresee a lot more research in this area.
...is the same asteroid currently hitting all forms of knowledge work and even more theoretical pursuits (math research, etc).
Different sets of people are just noticing/reacting to the impact at different speeds, but we are all standing in it.