new vesuvius challenge about to drop? Nat Friedman

it's unfortunate that the ideal time for good old fashioned engineering is 7am-11am, but the ideal time for tensor engineering is 8pm-1am

it only took 2 days for air chat to be taken over by hornyposting

air chat made texting turn-based
user, assistant, user, assistant

i like to think that when Paul Graham’s kids become adults he’ll stop referring to them as “15 yr old” and start saying “child 2”

in some ways competition fixed the 'how to trust negative results papers' problem -- if something was put on arxiv it definitely didn't work :)

i would pay for a service that summarizes all the 6 hour long podcasts into 20-30 min of the highest order bits

the 9gagification of tiktok is nearly complete. expect exodus soon. the pattern strikes again

Now that another LM product is getting flack, I can say this without sounding too self-serving:

Alignment -- controlling a model's behavior and values -- is still a pretty young discipline. Annoying refusals or hyper-wokeness are usually bugs rather than features

it clicked for me today that the way we use llms today is remarkably similar to how computers were once programmed with punched cards - one way dataflow during execution, instructions and data passed in together, results collected at the end

the og ai gf

the computer is evolving

being a vim user is cool because you look/feel like the most powerful hacker every day then one day you install a plugin to do basic linting that any ide does out of the box and suddenly you're in the future

Welcome to Curio, where toys come to life! 🧸

We've partnered with 𝖦𝗋𝗂𝗆𝖾𝗌 ⏳ on our first 3 characters: Grok 🚀, Gabbo🤖, and Grem👽.

Order by Sunday to receive a Beta Program Certificate in the mail before Xmas, toys ship early 2024!

heycurio.com/product/grok

reports of turing test being broken without fanfare are ignoring the fanfare of people on here thinking the model has seasonal depression

“i wish for three more wishes”
—original prompt injection attack

congrats to all at google deepmind on shipping!

It’s rarely discussed that ML researchers have to be strong data scientists. And not like the “tidy data format” kind, but like the kind where you go to great lengths to measure the thing you need, and write terrible code to make baller plots