If LLMs did nothing else for science than what they have been doing in math - combining diverse ideas across subfields in novel ways - it would still be revolutionary. Before LLMs, science was stalling under the burden of knowledge, there is too much to absorb & work was ossifying as a result.
“Museum of Cronenbergian Technology” 🤔
ULTRA🏝️BLAST@ultrakillblast.bsky.social · 3d agoinsane ergonomics in THE FLY
If only we had had spatial visualization software to navigate philosophy problems on our SGIs…
Dan Hon@danhon.com · 2d agoif we ever achieve this level of consciousness then I will consider that a win (another example of what happens to my brain when I catch up with @knowtheory.net)
via isaac king "Each randomly chooses one "friend" & "enemy". At each step every dancer moves 0.5% closer to the centre then takes a large step towards their friend & a small step away from their enemy. At random intervals one dancer re-chooses friend & enemy" community.wolfram.com/groups/-/m/t...
I’m now thinking way too much about “meme rides is not a thing I'm fully prepared for” @davidpierce.xyz
the actual solution for "math can't make anything novel" isn't to lean harder into proving LLMs can make novelty, it's just to teach them to write shaders there's soooo much procgen stuff from 30 years ago that already suffices this conversation
Started work on a PCVR port yesterday for a game I’ve been dreaming of playing in VR. Here’s Metal Gear Solid 2 running on an Apple Vision Pro.
At @skeuomorphpress.org strategizing with @drtlwagner.bsky.social about the official launch of the Undead Media Lab this fall & they convinced me we had to test out the new digital clock font—it’s everything I dreamed!
Ryan Cordell@ryancordell.org · last mo.Could not be more charmed by this new font from Skyline, in the style of a vintage digital clock/watch face Can’t wait to make something with it @skeuomorphpress.org
Introducing Qwen-Image-Bench, a creator-centric benchmark for text-to-image generation, co-designed with artists. It sets a new standard by evaluating T2I models on real-world fidelity and creative generation, revealing gaps traditional benchmarks overlook. https://arxiv.org/abs/2605.28091
Qwen-Image-Bench: From Generation to Creation in Text-to-Image Evaluation
ArXiv link for Qwen-Image-Bench: From Generation to Creation in Text-to-Image Evaluation
arxiv.org
Its kinda strange to me that the age of llms has not revived text based MUDs. Seems like an obvious good fit.
I made this post before watching this video, but this is the exact sort of thing I’m talking about! youtu.be/S67z2aekBrI?...
I Gave ChatGPT a Body
YouTube video by Art of the Problem
youtu.be
Grace@gracekind.net · 2mo agoThere’s currently an overhang where LLMs are quite intelligent and creative, but don’t yet interface with humans in non-text ways. Bridging this gap is going to be very uncanny and possibly very beautiful
Turns out we could have been 3D-printing overhangs at 90 degrees all along with the right slicer. Links to paper and code in comments. Video credit: Janis Andersons.
Now published in open access! Your one-stop shop for the philosophy of language models. It's the spiritual descendant of our two-part preprint from 2024, fully updated. This should be particularly useful for anyone looking for an entry point into this rapidly growing field.
The Philosophy of Language Models
The success of large language models (LLMs) across many domains of AI research has generated intense debate. Some attribute their impressive performance on complex tasks to human-like linguistic and ...
compass.onlinelibrary.wiley.com
