They both appear today on the chart archived, the one you say they don’t appear on today.

My paranoid impression was more that the updating and addition of tests — obviously essential in this game — was delayed to provide advertisement for Kimi K3.1. The same happened for K3 — and this is why it was famous — but its position had collapsed by the time of 3.1

What is most charming about this benchmark is that by now they do train on past results, and thus commit all the same errors over and over.

"At a standard Q4_K_M quantization, the model footprint takes about 459.5 GB."

Every human culture thinks that nothing should be done to save the deer from the wolves or interfere in the food chain of intact ecosystems. Those that have them have different ideas for pets which are ‘members of the household’ as Kant put it. There is no unreason in this. And hunter gather cultures have still others for hunting quarry ; pastoral peoples others for eg the sheep they move from hilltop to hilltop as distinct from the dogs who assist. None of these distinctions have a trace of hypocrisy nor do they need rationalization.

The bourgeois factory farm is a menace but because of the usual competitive drive that showed in Dickens time leading to eg child labor and 16 hour days - but which here has no one to rebel and it’s all hidden by the separation of production and consumption. The operative principles, market and competition, are near universally praised.

This quite particular hyper capitalistic system is the origin also of the sentimentalist point of view you think opposes it by erasing distinctions and ignoring structures — it is a point of view which necessarily calls us depraved for leaving the wolves and deer in peace. It is clear that deer are perfectly conscious and suffer but also that the nature in which wolves eat them is inviolable. This involves no rationalization but is simply the truth.

Why didn't they just give mathematicians access, so they could at least understand and write up the results in publishable form?

They're all here https://github.com/openai/math/tree/main/preprints

I think they meant "access to the model" rather than the results.

I'm seeing a lot of this and it makes no sense, the internal model solved these but give one of the papers to Astra and Opus and I'm sure it will have no problem recreating it.

I don't think they mean this from a "verify this paper" perspective.

How valuable it would actually be to share the model with other mathematicians vs just have OpenAI's mathematicians churn out and clean up results isn't very clear to me though as they don't say how much effort it's requiring from their team to prompt and clean these up vs how much it's bound by "time to run the model" or similar.

I recursively self-improve my shoe-tying powers every morning.

I first saw it with Homebrew over a decade ago. It made it more linux, it's true. (Of course now that its for linuxes too, so they get to be more like macs.)

I see archive.today has the charming old ruby version on the bottom of a 2013 brew.sh page in 2013, when I must first have used it https://archive.ph/lCqJ1

    ruby -e "$(curl -fsSL https://raw.github.com/mxcl/homebrew/go)"

But the form of incantation is now pervasive.

Philosophy of technology has existed since Plato and Aristotle. Every other word in their corpus is technē, which is why we call technology 'technology'. In every succeeding epoch the theory of the technological has been developed. It isn't clear there are 'disciplines in philosophy': there are only a few thousand researchers to begin with, on the scale of art history. I asked perhaps doubtful AI but it said that in AI there is an 'elite tier' of AI researchers of 5000 from maybe 100,000 total; the corresponding figure for philosophy was 500 out of 5000. The AI world could finance its own army of philosophers and it could be bigger than real philosophy, and no one would notice the dent on funds

版权