🏅 Top traders

Should resolve YES, multiple employee confirmation on twitter.

@DanMan314 mind sharing some links? (or if the other markets resolve to openAI, that'll be sufficient as well)

@Ziddletwix See the discussion here: https://manifold.markets/strutheo/will-the-public-find-out-the-origin#

@DanMan314 Yup resolving this YES—I now see the comments from OpenAI employees, plus other prominent markets are resolving, that's well above the evidence bar I outlined in the description.

@traders To make this explicit, "gpt2-chatbot" is confirmed to be created by OpenAI.

I didn't require official confirmation, but this tweet is basically official confirmation: https://twitter.com/sama/status/1790066003113607626.

I'm also going off community consensus, and strutheo's market resolved YES, without any controversy from what I can tell. Seems clear cut to me.

@Ziddletwix yeahhhhh, that doesn't look at all to me like "basically official confirmation." That's basically Minecraft saying "removed Herobrine" every patch. It's a reference, not an admission. But whatever, this sort of situation was a part of the stated criteria.

To be clear, I am benefiting from this resolution, I bet Yes.

It just feels like describing that tweet as "basically official confirmation" is seeing patterns in clouds and insisting it's "definitely an old man with a pipe in his mouth instead of a horse laying on a beanbag."

I have also looked at the other tweets, and I think I'm still >10% that it turns out in the long run that gpt2-chatbot wasn't from OA.

@NevinWetherill well if your personal probability is >10% then you should place some very profitable bets on the markets at ~99% https://manifold.markets/StephenMalina/is-gpt2chatbot-created-by-openai https://manifold.markets/axr/is-gpt2chatbot-made-by-openai

the description is very clear that official confirmation is not required—just what other markets resolve to would be sufficient

Is gpt2-chatbot created by OpenAI?

98.4% chance. Right now, all the questions about gpt2-chatbot conflate whether it's a new, improved model with whether it's from OpenAI. Let's break this down - this Q will just focus on whether this model was created by OpenAI. The question will resolve "yes" if it's confirmed before EOY by OpenAI officially. I will bet in this market myself as I expect the outcome to be relatively clear cut. If it's not clear cut I'll err on the side of N/A to avoid the appearance of bias though.

@Ziddletwix Yep, I'm not complaining about the way this market resolved, otherwise I'd be saying "mods!"

I was only mildly complaining about the sense people have that this is basically certain when I am not nearly as certain myself. It feels like people are jumping to conclusions when the stuff OA has said about it also seems consistent with them just using this as a conventient opportunity for publicity, since everyone was paying attention to gpt2-chatbot and they could drop references to it.

I'd want something a little more blatant than a cute reference for "basically official confirmation" - like someone from OA saying on a Twitter space "yeah, that was us, dammit, stop DMing me and asking about it."

And I may bet in that first market. They both seem mispriced but the first seems more likely to use the kinda criteria I'd want. Something actually blatant. I don't wanna get burned by a bunch of markets that resolve yes because "everyone basically knows it was OA" and then 2 months later OA is like "lol, no that wasn't us, we just thought it was funny to neither confirm nor deny it."

(Edit: Oh, it turns out I already did bet "no" on that first market back on May 1st.)

only may 4th but worth noting i haven't seen anything remotely like confirmation (nor have i seen any of the related markets resolve). this isn't about "do we have a pretty good guess who is behind it", but whether there's explicit confirmation or "overwhelming community consensus". (if other markets are being actively traded, it probably isn't overwhelming consensus)

Sam Altman basically unofficially confirmed it already...

@ShadowyZephyr Sam loves playing into things like this, I wouldn't take what he says as confirmation. Having said that the model very much feels like an OpenAI model, though we can't rule out that it's someone else proxying an OpenAI model.

@chrisjbillington True, but my confidence is higher than 80% based on the tweet, the initialization prompt, and what people have been saying about it. Who would have the capability to make a model this good that emulates OpenAI? If it were any open-source company or even other company that would certainly flip the AI scene on its head....

@ShadowyZephyr I agree the model looks very much like OpenAI, but it could be someone else behind it - forwarding requests to the OpenAI API and modifying the prompts or responses in between in some way. So it could be an OpenAI model but without OpenAI being involved in it being on lmsys. Not saying that's likely, though.

The case for it being an OpenAI model underneath is strong just based on its output. But I don't trust Sam not to lean into the attention regardless of whether OpenAI is involved or if it comes as just as much a surprise to him to see this show up today. So I treat the tweet as not being much evidence.