OpenAI has just announced the “12 Days of OpenAI”, a series of 12 livestreams with new announcements. What will they announce during this?
https://x.com/openai/status/1864328928267259941?s=46
Notes:
“Released” = available to some random members of the public, or an announcement that it will be available before December 31
“Revealed” = the thing is shown to exist, but not necessarily available to the public
Possible clarification from creator (AI generated):
For "Sora 2" announcements to count, they must show better quality than the original version revealed in early 2024, not just faster processing
The new version can keep the name 'Sora' but must be clearly distinguished as a different version from the original
Update 2024-18-12 (PST) (AI summary of creator comment): - Announcements made on December 23rd and 24th will count as part of the event if OpenAI indicates they are part of the '12 Days of OpenAI' series
A 13th day or additional days will count if OpenAI explicitly connects them to the event
Here are the polls about the most "significant" announcements. I also included a question about Sora vs. o1 for my own curiosity, since those were about neck-and-neck throughout the course of the market.
/CDBiddulph/which-was-more-significant-the-o1-a
/CDBiddulph/which-was-more-significant-the-sora-pLdhQuRyLZ
/CDBiddulph/which-was-more-significant-the-laun
Bonus:
/CDBiddulph/which-was-more-significant-the-sora
I forgot to mention the non-o3 announcements of the 12 days of OpenAI in my poll - oops! Hopefully no one thought 1-800-CHATGPT was more significant than ChatGPT itself 😄
Marcus seems impressed: https://garymarcus.substack.com/p/o3-agi-the-art-of-the-demo-and-what
@dominic this is going to be prompt and interpretation specific. Entire article I can see being negative sentiment (and Chatgpt agrees) but the are positive reactions within it. Chatgpt acknowledges there is "something he reacted positively to" and recommends resolving yes
@Usaar33 I think we disagree on the domain for this question. I didn't bet on this, but I interpreted "something" as an actual product, like o3.
You could say that Gary Marcus reacted positively to "o3's performance on the ARC benchmark", but I wouldn't have counted that as one of the possible "somethings".
“o1 and/or o1-mini fine tuning” is currently trading at 4%, but was revealed waaaay back on December 6: https://x.com/openai/status/1865136373491208674
@mathvc Even if that is a sentence he actually said, I am sure it's not very hard to find a positive sentence in his article. But I don't think it's fair to take it out of context and not analyze his full reaction to each thing.
@mathvc Yeah, this particular tweet might get analyzed as positive. But it's a stretch to take one mildly positive tweet and reasonably say "this is his reaction to o3," unless he hadn't written anything else about it.
@dominic well i would like to disagree. He has a positive reaction to an announcement of model scores on FrontierMath
This is separate from his negative reaction to o3
Should o3 count as 1,000,000+ context length? According to the people who run ARC o3 was able to use almost 10 billion token chains of thought to solve just one problem.
AGI achieved only at 1.6%? 👀
@mathvc there's also this standalone question for the release of o3: https://manifold.markets/JamesGrugett/will-openai-release-o2-as-part-of-t
@MalachiteEagle its 60 and i already moved the market from 90% to 50% lol, do you know what is liquidity? 99% of my mana is locked in another bets, i’ll bet more when get it out of low-return markets
@MalachiteEagle as you wish, i added a little more to the bet and moved the probability from 90 to 28
Btw my guess for the hint “oh oh oh” is 0.00 pricing on some products
@IsaacLiu can be Orion, maybe demo of Operator, but o2/o3 makes no sense lol, they have just released o1
@mathvc maybe announcements/demos of all three? We're meant to think he's teasing o3 but he's actually teasing Orion, Operator AND o3
(three O's)
@MalachiteEagle i agree with you on naming conventions 💯 that’s why i think real probability is non-zero (but way below the market)
@mathvc feel free to ask people (including me) for a loan in https://discord.com/channels/915138780216823849/1194385337510797432 . it can
allow you to make much greater returns if you know something the market doesn't