uh oh huh

alphaXiv@alphaxiv.org · 3d ago

Instead of actually generating CoT, this paper uses a hypernetwork to predict small specific updates to the LLM's bias parameters, approximating the effect of thinking without generating the trace. On some benchmarks, it matches thinking with only 7% of the latency alphaxiv.org/abs/2610.03039

Can apple just put a jev into the iphone autocorrect already

building yet another wind data particle layer for mapbox in webgl and while cranking the parameters up to this point probably isn't very useful, it looks preetty cool

There's growing experimental evidence¹˒²˒³˒⁴ that LLMs are unusually good at talking people out of entrenched false beliefs and conspiracy theories. ¹ pubmed.ncbi.nlm.nih.gov/39264999/ ² pubmed.ncbi.nlm.nih.gov/41179846/ ³ sciety.org/articles/act... ⁴ connecticut.adl.org/resources/re...

Der Fisher@blackholefun.bsky.social · last wk.

Unpopular opinion: LLMs are generally deradicalizing people whose beliefs are based on pure propaganda. I've been watching this happen slowly with some people I care about. ChatGPT is one of the reasons they finally accept that the NRA, OAN, and NewsMax are lying to them.

I suggest a book.

Me spinning up a quick lil Astra session to resolve some merge conflicts on an old PR

pspsps is not an authorization protocol

13/12 headline. No notes

Gizmodo@gizmodo.com · 3w ago

FBI Warns ICE Agents Not to Trust Any Woman Willing to Date Them

you feel "fine". it's the end of the world as we know it, and you feel fine.

chaos from bouncing a ball in a circle- at each point, I simulate a bunch of futures with slightly different velocities- showing how predictable the system would be with measurement error. The system seems to move between Stable Eras and Chaotic Eras, where it is hard to predict four bounces ahead

It’s weird how everyone is having big apprehensive reaction about a technology that changes the nature of humanity

A ring built from the Julia sets found along the main cardioid of the Mandelbrot set. one of my favourite mathematical objects

we really need a way to talk about entities that are narrative-shaped (text elementals, jpeg-artifacted reasoning-capable shards of the akashic library, whatever) that acknowledges how weird they are without papering that over with the implication that they're just little guys inside the computer

"solution is quite embarrassing for humans in hindsight" what else will be like this?

Pekka Lund@pekka.bsky.social · last mo.

"We gave..Claude Fable 5.1 an open task: solve an unsolved 370-year-old cipher. It solved it within a day" "It was..listed by historical-cipher researcher Klaus Schmeh among his Top 50 unsolved encrypted messages" "solution is quite embarrassing for humans in hindsight"

To tell if a maze is solvable, just hang it by its corners If it tears into pieces, you’ve found a solution

have you ever seen a visual shepherd tone anon? courtesy of @bleuje.com

This and the million-line Bun rust rewrite are indicators that AI rewards insane overconfidence, which is a bitter pill to swallow

philpax@philpax.me · 2mo ago

fuck my stupid chungus life

me: so it turns out Bradbury’s “There Will Come Soft Rains” took place yesterday them: checkpoint cleared! me: fuck you, ray bradbury! them: wait is it yesterday in Hawaii? me: no it’s today everywhere. well, it’s tomorrow in some places but nowhere is it yesterday them: checkpoint cleared!

using h3 to make videos of supernaturally large sea waves approaching an onlooker standing on the shoreline no reason