There are numerous tests that work. A battery of all of them would pretty reliably differentiate an AI from a human. The maker said no background checks or data brokers, but questions about politics, background questions (asked via comments), threat responses, and especially changes of subject, are all still good ways to tell a bot from a human in normal usage. Of course, one couldn't distinguish correctly 100% of the time, but the market doesn't seem to be asking that (and would be nondisprovable if it did). I might have put up a bigger order, but since the maker has the biggest Yes position and resolution is subjective, I have to be careful.

and next up to cause more confusion for the everyday user

Aren’t there a couple of questions that will reveal any bot? Like the copyright notice chatGPT gives at the moment, or questions about its infancy?

Somewhat related market:

It’s impossible to do this now, stuff like botometer is highly controversial.

How do you plan to resolve this? It seems that actors capable of training state of the art models are being very careful to make the AI identify themselves as AIs, at least for their public facing models. Any large scale injection of AI comments on social media would likely be a covert operation.

@mariopasquato based on the best determination I'm able to make at the time, I was thinking only about models available to the public, do you have any suggestions on how to improve the description?

Any human or just the most bot-like humans?

@Fion a human one would want to interact with, in other words if it becomes impossible to safely disregard bot comments based on available information.

Just based on their comments, or also looking at their profile and other info?

@NathanScott readily available information, so comment history, consistency, profile yes, background checks and data brokers no.