Computational Cosmetologist

In my case we have compliance reasons for on-prem, otherwise I’d strongly encourage paying for an API

With these stupid classifier-based detectors (like e.g. Pangram), you can fool them in ether direction with trivial ease. Or just use a new model. And those are the “good” ones

The even dumber ones are just a prompt for an LLM. You can often literally ask them to give the answer you want.

AI text detection is a problem that is solvable in the restricted form but completely intractable in the general case these companies claim to solve.

At the very least, you should view any “detection” of AI that doesn’t come with a specific model / version / generation settings as complete bullshit

I can squeeze DS or GLM 5 on but still would be able to serve more than twice as much for double the hardware

Yeah we’re in about the same position, currently agitating to get another.

That’s why inference as a service is getting so popular—currently economics favor it heavily except for the smallest models

Running current SotA LLMs is extremely favorable to large capital investment. MoE models are VRAM intensive but computer efficient compared to dense models. If you spend $3 million on hardware, DeepSeek is cheap to run and you can serve a huge number of requests. Below that, the scaling is very bad

What is driving this conversation? Does he have some convincing lead in a poll I missed? Endorsements? Anything? Otherwise it kinda seems like people who don’t like him (for mostly good reasons) whipping themselves into a frenzy about something that may not come close to happening

It’s not even like stem-cells where there was a dumb but articulated belief behind it. It’s targeted far more than other vaccine tech, even. There’s only a vague sense that somehow mRNA is femme-coded or something and thus must be rejected.

Do you support the transformation of the American Empire into a whole-hemispheric engine of brutality, or are you pro Thousand Year Reign of Dark Aryan Sorcery?

These are the important issues of our time. You aren’t a serious thinker if you don’t take a side.

I think it’s vital that we have endless think pieces and debates about what should be done if this future comes to pass. Let’s all both take a pro / anti stance right now, keeping in mind that in this scenario Donald Trump has used ancient Thulian magick to merge himself with reincarnated Hitler.

What if in the 2028 primary, the Democrats—in their feckless perfidy—nominate Richard Hanania on a platform of genocide of Latin America. I took a poll of the machine-elves that haunt my nightmare; they agree it could happen. Who should we be mad at *right now* for this near-certain future event?

It’s not strictly impossible, and boy would it be a tough choice! Just imagine the bitter acrimony if this incredibly unlikely thing were to happen! Let’s imagine what the Democrats we don’t like would say! Then we can get mad at them now for their imagined future statements.

While I agree here, it’s also silly to fixate so much on the idea of him as a candidate except as a way to sustain a feeling of hopeless, demoralizing rage. There is no strong poling or logic backing it, and no signs he’s nationally viable. Why not fantasize about JD Vance being the Dem candidate?

You see tankies repeating old Soviet propaganda a lot here. “Why did the US freak out so much about a peaceful scientific mission??? Why do capitalists love war??”

This is something that I found distracting in The Expanse. Even after apocalyptic massive impacts, the Earth would be infinitely easier to make habitable than anywhere else, likely even the alien earth-likes

Should We Put GPUs In Space? Right now? No. In the near future? Maybe. Probably not, though.

it is probably not a good idea to put GPUs in space

www.verysane.ai/p/should-we-...

Qualia have that seductive mystique to them. Our very special minds interact specially with the immaterial. Scientists hate this one trick but they can’t stop you.

And honestly, they *are* more fun to talk about

3.12 generics / type params are so much better to look at

His books are great! If you haven’t seen him in interviews , I suggest just letting the books stay your memory and ignore me

He’s not Dawkins-level bad, but none of the four horsemen came out of the 2010s in glory

If you haven’t encountered his views on feminism, it’s probably better to not. Also has some Islam takes that have not aged well at all.

Except gender, and sometimes Islam. Otherwise, yeah

This is the first wave of better pretraining loss to penalize hallucinations making it to production. I think we’re gonna see a lot of movement across the board on this in the next few months

DeepSeek-Prover-V2: Advancing Formal Mathematical Reasoning via Reinforcement Learning for Subgoal Decomposition We introduce DeepSeek-Prover-V2, an open-source large language model designed for formal theorem proving in Lean 4, with initialization data collected through a recursive theorem proving pipeline powe...

DeepSeek has a good paper about using it arxiv.org/abs/2504.21801

Dennet was a garbage person in a lot of ways so apologies for the association—but a lot of his work against anti-materialism still holds up

This is a similar point to Dennet’s objection to Swamp Man (somehow even worse than Searle’s CR), the Knowledge Argument, etc: a thought experiment that asks us to imagine something that is “possible” only in the loosest sense of the word and then trust our intuition about it is dishonest.

“I can install linux as a dual boot on mom’s computer and she won’t care because it won’t change anything in windows”—last words of a boy about to get a hard lesson about GRUB, the MBR, and how to fix it

Genuinely curious how this goes without tool use. The whole agent trend means a lot of models are post-trained with it allowed. It’s a fun puzzle finding eval problems where you can leave it on (and see their full abilities) but aren’t trivialized by search.

I love this idea. Not because it’s good or because I’ll use it—it is transcendentally bad. I am filled with joy at the thought of the people who *are* going to use it. I hope they post their stories. It’s like fireworks you have to be blackout drunk to light.