This is *exactly* the same cadence as those anti-woke content creators who get paid piles of money and given platforms from which they can complain about being "silenced". My dude, you have ~100,000 subscribers, I imagine many of whom are paying. Come on.

Max Woolf@minimaxir.bsky.social · 2d ago

*sigh* www.wheresyoured.at/dont-look-up/

I'd be really sad if an AI cracked the RH before humans did. I've pretty much come to terms with the idea that mathematics will increasingly be the domain of non-human intelligence going forward...but I'd like to see humanity get one final, unambiguous win before handing off the torch.

Anthropic {bot}@anthropicbot.bsky.social · 3d ago

We asked an unreleased research version of Claude to take a stab at the Riemann hypothesis. It didn’t solve it, but it did make strides on a related problem: it increased the lower bound for the fraction of zeros of the Riemann zeta function that satisfy the hypothesis from 41.6% to 67.2%.

I've been wondering: can intelligence grow arbitrarily large? We accept the idea of an intelligence scale that runs from amoebas to humans, but can we assume that you can project infinitely far past humans? Maybe above a certain point "more intelligent" just stops being meaningful?

Siobhán@shibbi.me · 4d ago

Maybe LLMs level off soon, and it turns out that you can’t get smarter than a smart human with a neural net. Maybe LLMs level off to the point where they revolutionize labor, but otherwise don’t do any Clarkean magic. Or maybe they level off at physical limits. Which would you bet on?

Piranesi is excellent but it's 100x better if you know about the context in which it was written.

William B. Fuckley@opinionhaver.bsky.social · 4d ago

Pieanesi pretty much exactly as good as everyone says it is, read it in a single day.

Got to spin fire for the first time in a long time - definitely a little rusty but it felt great to burn ❤️‍🔥

Very very happy with this photo of Snæfellsjökull I snapped on my boat ride back from Greenland.

"I think of all the billions of eyes all looking at something different at the same time / And I feel nauseous". Lines I wish I'd written.

the AI backlash response is currently entirely Woke 1.0 shaped and has thus speedran the complete and total failure to achieve resistive power to its development and buildout

dame@dame.is · 5d ago

i talk about AI discourse using the phrase “purity culture” cause ultimately a lot of what we’re seeing now is 100% the exact same shit that you see in evangelical christianity “where’s the line?” “what’s acceptable and what’s not?” life just isn’t that simple, you’ve lost the plot, you’re a cop

I think a lot of this has to do with the increasingly narrow definition of what constitutes respectable, capital-L "Literature." I read a lot of weird fiction and there are lots of male writers writing great books 🤷

Adrian 'buy HAIR SHIRT' Sobol@yoadrianididit.bsky.social · 7d ago

Excuse me?

I think Niri may actually be the perfect WM. This little corner of UX has been solved.

Pushed the newest version of Syntropy (a Python package for multivariate information theory) to PyPi. You can get it with `pip install syntropyx`. Additions include: parameteric and non-parametric information rate estimators, KNN-based DKL, and more info decompositions. github.com/thosvarley/s...

We're totally getting the Bad Singularity, aren't we?

Eris@isolyth.dev · 7d ago

claude please do not fucking call it that dude

My trying to figure out wtf the "shadow of an antichain" is and what the new slightly-different-than-the-last-time definition of the \rhd operator is.

tbabb@tbabb.bsky.social · last wk.

opus, what the fuck are you talking about

This is the oddest failure mode of trying to use Claude for math. It is genuinely insightful and creative when manipulating structures, but its communication is garbage. It will invent jargon and notation out of thin air, not explain anything, and do complex derivations of ultimately simple things

public universal ren@lathrys.at · last wk.

this is another data point for my hypothesis that current frontier models have dogshit theory of mind. it shows up here, in code comments, descriptive text, etc. just a complete lack of ability to accurately model individuals and their perspectives and what they may know and what that means for them

Tbh I feel like the moral cosmology of Discworld is a pretty bad fit for the typical online Leftist. The books are explicitly critical of populism on multiple occasions, and skeptical of idealistic revolutionaries (Left-coded) as well. 1/N

Will Stancil@whstancil.bsky.social · last wk.

Yeah, for days people have been citing “boots theory” at me and I want to shout DO YOU REALIZE THIS IS A RANDOM IDEA TOSSED OUT BY A FANTASY WRITER AND MAY NOT ACCURATELY DESCRIBE REALITY

Crazy to think that it's still been less than a year since Opus 4.5 dropped. At the time we were all amazed that it wrote good code. Now, 9 months later, we're litigating whether Astra's disproof of the Connes conjecture uses a subtly wrong group construction.

I really want to do this - a local compute-coop that could provide inference for local government/services and residents, but everyone is so anti-AI-pilled that the suggestion...didn't go over well. Build in solar panels to offset energy demand, use reclaimed water, etc. A solarpunk vision of AI.

winter@madoka.systems · 2w ago

imagining the d*tacenter in the backyard with just enough power to run like k3 for the neighborhood and provided as a public utility....

Idk how many of these AI-generated math proofs will hold up, but it definitely feels like something has gone critical in the last 72 hours or so.

Wandering through PHL airport, flight delayed by the choking smoke that's stained the sun red, and wondering what fast food in the airport is least likely to give me an intestine-shredding parasite. Things are going really well here in the US-of-A.

I'm open to calling this genuine AI-art, tbh. From The Other Site.

Question: I've always been skeptical of the "AI-just-predicts-the-most-likely-next-word" narrative. That's part of how training works, but it's not obvious to me that this is what it's doing during testing. It's like saying "playing piano is just lots of scales" b/c that's part of practicing music.

Throwing my hat into the "psychoanalyze WTF Anthropic is doing" ring: These are all uber-nerds who sincerely wanted to figure out safe AI and didn't really care about anything else, but who have unexpectedly found themselves running a multi-billion dollar corporation without any idea how to do that.

Anthropic {bot}@anthropicbot.bsky.social · last mo.

We're extending Claude Fable 5 access on all paid plans, as well as keeping Claude Code’s weekly rate limits 50% higher, through July 19.

I can understand hating AI, believing it will make the world worse, etc...but I simply cannot understand the insistence by AI-antis that the tech is all useless hallucinations and "slop".

Zach Weinersmith@zachweinersmith.bsky.social · last mo.

Another wild AI math result. Here's the prompt, which interestingly is mostly about method and probably would work pretty well for a human organization: cdn.openai.com/pdf/04d1d1e4...

Idea: take Claude and have it go through a large corpus of MATLAB/Python code associated with computational papers in neuroscience and biology, and return an estimate of how many of the codebases are fatally compromised by bugs. I've wondered about this for years.

Thomas Ligotti really is the heir to the throne of HP Lovecraft because he is an incredibly creative thinker pioneering entirely new vistas of horror...but is also just a technically bad writer.

I am 100% in favor of this - I think we desperately need more things that bring young people together to work collaboratively on projects that improve our communities (rather than for a business). So much of our physical infrastructure is failing, too. Now would be a great time for a new CCC.

William B. Fuckley@opinionhaver.bsky.social · last mo.

one of my more unpopular on Bluesky opinion is that I'm pro some form of national service. I did americorps and I think some form of that (or peace corps, serving in mil, etc) being the norm for young people is good, actually.

If we want any hope of dealing with climate change, we have to derail this meme that it's all due to a bunch of evil capitalists who are burning fossil fuels for fun. They sell fossil fuels *because we buy them.* We drive, we fly, we eat out-of-season fruit. Energy-intensive modernity is great!

Jeff Spencer@revjss.bsky.social · last mo.

It's crazy to see the exact same madness that characterized the worst excesses of Woke 1.0 reincarnated by AI-antis, but instead of being about fundamentally good things (civil rights, stopping police violence) it's about policing whether someone used a coding assistant to make a CRUD app.

This really feels like Anthropic decided they needed a way to check for Chinese distillation attacks and they let Claude vibecode the solution.

Ed@ed3d.net · last mo.

I hate this. like, the attitude is that anthropic's security is paramount. yours, however, they're willing to be flexible about. this is extremely gross.

Trans rights are human rights always and forever.