bonoboTP

28 ms·

781.
by bonoboTP 9mo ago
Why do you think batching has anything to do with the model getting dumber? Do you know what batching means?
782.
by bonoboTP 9mo ago
It can be made deterministic. It's not trivial and can slow it down a bit (not much) but there are environment variables you can set to make your GPU computations bitwise reproducible. I have done this in training models with Pytorch.
783.
by bonoboTP 9mo ago
How is this related to overloading? The nondeterminism should not be a function of overloading. It should just time out or reply slower. It will only be dumber if it gets rerouted to a dumber, faster model eg quantized.
784.
by bonoboTP 9mo ago
This has nothing to do with overloading. The suspicion is that when there is too much demand (or they just want to save costs), Anthropic sometimes uses a less capable (quantized, distilled, etc) version of the model. People want to measure
785.
by bonoboTP 9mo ago
The commenter I was replying to wanted it to be a prerequisite for a degree, not for a grant. Grant awarding institutions also have to justify their spending to other parts of government and/or parliament (specifically, politicians). B
786.
by bonoboTP 9mo ago
We are on the comment section about an AI conference and up until the last few years material/hardware costs for computer science research was very cheap compare to other sciences like medicine, biology etc. where they use bespoke inst
787.
by bonoboTP 9mo ago
Universities are not really motivated to slow down the research careers of their employees, on the contrary. They are very much interested in their employees making novel, highly cited publications and bringing in grants that those publicat
788.
by bonoboTP 9mo ago
Prerequisite required by who, and why is that entity motivated to design such a requirement? Universities also want more novel breakthrough papers to boast about and to outshine other universities in the rankings. And if one is honest, othe
789.
by bonoboTP 10mo ago
Curl|bash isn't any less safe than installing from random a ppa, or a random npm or pip package. Or a random browser extension or anything. The problem is the random, not the shell script. If you don't trust it, don't install
790.
by bonoboTP 10mo ago
Net is excluding tax, you mean gross.
791.
by bonoboTP 10mo ago
Recently asked Claude Code how to do more thorough tests and described how I imagine it and it set up Hypothesis and mutmut testing. The latter is quite cool, it introduces bugs in the code like swapping values and relational operators and
792.
by bonoboTP 10mo ago
I cannot reproduce this. It says it's GLM by Z.ai.
793.
by bonoboTP 11mo ago
Seems like Claude to me.
794.
by bonoboTP 11mo ago
Smog and immensely useful products, yes.
795.
by bonoboTP 11mo ago
Yeah, he's very old for a person and the US is very young for a country.
796.
by bonoboTP 11mo ago
It's different, more lightweight hardware for inference but can be in the same data center. Training requires beefier GPUs.
797.
by bonoboTP 11mo ago
Related, that in English "must not" means that it's forbidden, while in German the analogous "muss nicht" just means "don't have to". Also "need not" means "don't need", inste
798.
by bonoboTP 11mo ago
"This is just semantics" is a set phrase in English and it means that the issue being discussed is merely about definitions of words, and not about the substance (the object level). And generally the point is that it does not matt
799.
by bonoboTP 11mo ago
Another circle for the Venn diagram is that the schizophrenic sister's name happens to be Carol, the same as the name in the ad shown on the fridge. Obviously made up.
800.
by bonoboTP 11mo ago
Sans Francisco
801.
by bonoboTP 1y ago
> at the cumbersome payment interface and annoyed that enabling Drive is the only way to save. For the general audience, Gemini is the intended product, API and AI studio is for advanced users. Gemini is very easy to pay for. In Gemini,
802.
by bonoboTP 1y ago
You can use it also in Gemini.
803.
by bonoboTP 1y ago
It's similar to how they schedule all the free time of their kids too. Or their own free time for that matter. Must spend every minute being productive, every minute counted towards some labeled activity. Everything that isn't cla
804.
by bonoboTP 1y ago
It's the same with dating. The American rules about first dates, second dates, the exclusivity talk etc. is just not how it works in Europe. Maybe with online dating it moves towards it, but the regular way is much more informal and lo
805.
by bonoboTP 1y ago
Yes, for the vanilla Fourier transform you have to integrate from negative to positive infinity. But more practically you can put put a temporally finite-support window function on it, so you only analyze a part of it. Whenever you see a 2d
806.
by bonoboTP 1y ago
I didn't mean it in a moral sense, but in sheer production, population, the physical energy moved around the mass of materials moved around etc. It's a wholly orthogonal question whether this is better in a moral sense, whether pe
807.
by bonoboTP 1y ago
I get the thrust of what you say but this wasn't always the case. The idea of linear progress upwards is not universal. Many cultures placed the pinnacle age in their past, before some corruption or decay took place -the ultimate expre
808.
by bonoboTP 1y ago
Yeah at this point this is more like a cultural ritual. You know like howin some cultures you have to refuse a gift 5 times with increasingly stubborn facial expressions, and the gift giver has to insist through it all, and then accept and
809.
by bonoboTP 1y ago
That's the old speech recognition stuff that just matches keywords. LLMs can infer what you want pretty flexibly. I mean yeah if it's a bad model then it will be bad. If it's good then it's good. It's not like getti
810.
by bonoboTP 1y ago
Yes. The alternative these days is often a maze of a phone menu. If I can just say my issue in complete sentences and the AI routes it appropriately and otherwise acts on it as I want, that's pretty good.

More ›