jhrmnn

I've personally experienced three big jumps in LLM capability: o3, opus 4.6, and opus 5.5. I could go back to opus 4.6, but it did require way more involvement from my end.

Are there simply two cultures among programmers? Along similar lines as the two cultures of mathematics? One motivated by the process itself, by esthetics, by abstractions, ideas. The other by solving problems. And then obviously these two cultures react very differently to AI.

It all hinges on the definition of “progress”. The debate of the past month is all about questioning whether formally proving outstanding unproven theorems without human understanding constitutes progress. This is quite different from solving diseases.

Cancer is the price we pay for complex multicellular life. It’s not a failsafe, it’s the failure of all failsafes to regulate cellular growth within an organism.

No clue what’s the consensus on this but my internal mental model is absolutely that LLM AI is pure fast mode, no slow mode. The “reasoning” loops are an attempt to mimic the slow mode but ultimately it doesn’t really work. I’m curious about the recent maths advances though, they seem to possibly challenge this.

Take the tax system and the state out of the picture. How does a mathematician convince his fellow citizens that they should fund their work? Newton didn’t have to do that, he had income from other sources and did most of his mathematical research as a hobby

It seems only natural for nature to have it like this. It’d be hard for evolution to optimize things if the genotype to phenotype mapping was sharp, with phase transitions and such

It’s written from a perspective of decades, not years

It’s actually a fairly precise use of the word, just not in the very specific and narrow sense in which it has been used for AI.

Working with AI on science (not LLMs though), couldn't agree more.

I think it’s more likely that that’s because no one tried to solve such problems with them before (OpenAI apparently started working in Navier-Stokes after a rumour that someone seriously advanced the problem with AI) plus improvements in orchestration. Fair, the latter could be as dangerous as stronger models.

I can almost see two branches of mathematics developing. One which is human-understandable, the other formally verified. I assume the latter is a strict superset of the former?

The halting problem on steroids?

This starts to feel like chess engines. It’s obvious their play is superior but it’s impossible for humans to understand the moves.