brennan

"Astra autonomously researched and found hundreds of photos of the Palace of Fine Arts, iterated on the Blender scene..." "It even found an old scan of a document from the Library of Congress that described the dimensions for some of the columns." x.com/sharifshamee...

Sharif Shameem (@sharifshameem) on X x.com

Unnamed Simcluster Podcast episode #5! A chat with @scythiamarrow.bsky.social about her research in AI consciousness, the definitions of human consciousness and how perhaps the highest level Buddhists might not even be conscious at all! youtu.be/-1ywJ64B_H0

Unnamed Simcluster Podcast -- 005 scythia youtube.com

having a lot of conversations lately with various friends where AI people are complaining about their non-AI coworkers who are doing things slower and worse than could be done e.g. 20 minute tasks are being coded by hand but it took a week because there's bugs and errors to troubleshoot, etc

it is funny now to see "advanced futuristic AI" in past-made media and it's dumber than haiku

fable ironically being way cheaper on this project opus would do hour-long rips and do all this uselessly agentic testing and checking and running the scene, there'd be a dozen headless chrome threads spinning at all times, and fable is just like "I figured it out" in 30 seconds

increasingly coming across instagram accounts with 1m+ followers that are entirely AI and not a single comment seems to notice or say anything about it like, for the past year you'd see antis post boos and hissing in the comments and I think genuinely no one even realizes now

so if the antis can claim that using AI is against their religion, does this now also imply the reverse: the government is forcibly censoring access to my religion?

okay, but the phone call AI auto-screener thing is actually so handy

so, I've been to a few art and design and tech conferences over the years and this is one of my first in the post-AI era and it's really really interesting mr doob: "oh, yeah, I haven't used an IDE in like a year now"

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

for the past 2-3 years, at the top of the whiteboard for <big yearly creative project> is simply "what can we do to polish beyond what the LLM defaults are" and it's genuinely hard now! the water line has risen so much that some single shot demo posts exactly what we're making and looks the ~same