what do you think of jonathan blow's takes on this stuff?
LLMs will never be capable of making real good software
because they're not like alphazero, they just spit words
once someone does an AI that is actually a native programmer like AlphaZero is a chess player then it will be able to write assembly perfectly for anything and that's when human programmers won't be needed anymore
Login to reply
Replies (2)
I don’t think he’s tried the latest models.
Things are changing rapidly. They are good enough to pair program with now and have just become a really good tool.
Maybe if he used them he could finish his game in less than ten years.
> because they're not like alphazero, they just spit words
This seems to be changing right now: here's a paper on recent, roughly speaking, AlphaZero-like research that specifically uses coding problems. They make it learning on experience rather than on traditional datasets.


arXiv.org
Absolute Zero: Reinforced Self-play Reasoning with Zero Data
Reinforcement learning with verifiable rewards (RLVR) has shown promise in enhancing the reasoning capabilities of large language models by learnin...