So what can a person do with one of these?

When my partner mentions the dishes I've left on the side, I stare into the distance, channel Galadriel, and say 'That is no ordinary mess. That is the work of Sauron.' Left my washing in the tumble dryer? Sauron. Pile of books on the floor? Sauron's evil knows no bounds...

I’ve learned a lot helping my family get my dad into hospice but one of the biggest realizations I’ve had is that our country is facing a geriatric financial time bomb. It is absolutely insane how much elder care costs. For my family if we were to have someone 24/7 it’s $450k a

Anyone got a good example of a “reasoning” prompt that fails in GPT-4o but succeeds in the newly launched o1? openai.com/o1/

OpenAI's awkward "o1" AI model branding is kinda strange. "Strawberry" was right there, already christened and used by people to describe it for months

Believe it or not, the name Strawberry does not come from the “How many r’s are in strawberry” meme. We just chose a random word. As far as we know it was a complete coincidence.

Hey Jude Rehearsals, July 30, 1968 The #Beatles via Lucy (in the sky) ♟️

YAY! The CEO of OpenAI just recognized that LLMs generate text-based tokens using randomness and probability! Something objectively true that people have oddly made controversial! Check out our paper on this that introduced the term. dl.acm.org/doi/10.1145/34… It's a good day. 🤗🦜

Last November, several news outlets called OpenAI's new o1 model (then named Q*) a "powerful AI discovery" that insiders said "could threaten humanity" Now that o1-preview is out, I would love to hear if anyone thinks that is the case

OpenAI's o1-preview does pretty well on my "magenta" test. But the first LLM that just answers "no" without any qualifications will probably be AGI.😁 Reading its internal reasoning can be pretty amusing

So that’s why you only get three wishes

Here's everything I've figured out about the new OpenAI o1 family of models so far simonwillison.net/2024/Sep/12/op…

OpenAI’s o1 is also the first specialized frontier model available widely. It doesn’t do everything better than GPT-4o, but it does a few classes of things a lot better. Unless you are doing problems that benefit from planning a solution, you may not see improvement.

It occurred to me today that I am probably not thinking deeply enough to test o1 properly.😁I have fed it various tasks—and no, it's not perfect with every result—but it's clear to me that putting its logical abilities to the test will require a rethink

So excited… #bydhttmwfi

Some things never change! A 2,000 year-old Roman souvenir pen with joke inscription roughly equivalent to: “I went to Rome and all I got you was this cheap pen!" 😂 Dated c. 70 AD, the stylus pen was found in London during excavations by MOLA. The Roman inscription reads: