Would-be strawberry man Riley Coyote commented on Sam Altman's post, asking "can I tell them about the thing?" Sam replied "which thing?" (possibly implying that Sam knows Riley and that Riley knows some internal information)
Further down the thread, Riley made this claim about an OpenAI model creating "a new type of an existing drug" that is "better than anything any human had ever made:" https://x.com/RileyRalmuto/status/1880768187475620167
This market is meant to resolve YES if that claim is true, with some wiggle room in case we don't actually get direct confirmation about this specific event.
Specifically, this market resolves positively if we get confirmation that an OpenAI model created an improved design of an existing drug, whether internally to OpenAI or prompted by some external party. As long as we hear about it sometime in 2025, it counts, regardless of when it happened.
An "improved drug" may not get through all the FDA testing to make sure it's actually safe by the end of 2025, but should show a clear improvement at accomplishing its primary purpose during preliminary evaluations. Even if it looks like the new drug has more side effects, is more expensive, etc. it still counts as an "improvement."
The model has to do a significant majority of the work in coming up with the new design. Synthesizing the drug and running tests doesn't count as part of this "work." It's okay if the model has a pretty low success rate, e.g. it comes up with 10 full-fledged designs, all of them get tested, and only one of them is an "improvement."
I will not bet in this market.
Update 2025-05-06 (PST) (AI summary of creator comment): The creator has provided additional details on how the market will be resolved:
Regarding improvements to the drug's primary purpose: An improvement will be considered valid if the drug is clearly better than existing versions in at least one way related to its primary purpose, even if it is clearly worse in another way related to that same primary purpose (e.g., for a weight loss drug, an improvement could be losing more weight, even if it's at a slower rate).
The creator will not be conducting an exhaustive search of all individual medical papers published in 2025 to find instances of such improvements. Resolution will likely depend on information that gains broader public or scientific attention.
1,000People are also trading
@QuimLast Yeah, it's a good point that there might be multiple axes of improvement, even if they must be improvements to the "primary purpose" as I mentioned. For example, if the drug is supposed to help you lose weight, the improvement could make you lose more weight but lose it slower, making resolution unclear.
I would say that if the "primary purpose" is clearly better than any of the existing versions of the drug in at least one way, it counts, even if it's clearly worse in another way.
I think the market is likely to resolve NO unless there's an improvement that's "major" enough to be widely reported, so I think you should still feel pretty comfortable betting NO. I'm not going to dig into every single medical paper written in 2025 to see if they mention using ChatGPT and attempt to judge whether it did a "significant majority of the work."
@vluzko I was thinking it would probably be tested in rats or something first. Maybe we'd hear about those experiments going well
Quoting Riley Coyote's tweet:
> the drug was sent off for analysis thru the appropriate channels (bc ya know, ai researches aren't exactly qualified to assess the efficacy of a drug..) the other day the results or assessment or whatever came in and essentially stated that the drug was better than anything any human had ever made. for that specific drug.
I don't know what he means by "results or assessment or whatever," but I asked in a reply