This market covers videos generated by AI from scratch (for instance from a text prompt), not AI transformations of an input video, which have existed for some time now.
The market will resolve positively when a channel on YouTube or Twitch will stream at least 25 hours of realtime AI-generated video in the span of 48 hours.
Traditional procedurally generated videos do not count. There should be consensus in the community that the video is AI-generated. The channel owner should declare it, and no plausible refutation should be published within a week after the broadcast. If the channel owner remains silent, I might resolve the market based on a poll or the comments.
I will most probably notice if any gen-AI channel gets linked on the front page on HN, or gets viral on Twitter, but in case I miss it, please link it in the comment if you come across the channel that fits the description.
I will generally resolve each answer as NO as soon as the respective time period is over, but I will delay the resolution by at least a week in both potential "YES" and "NO" cases in case I miss an existing channel, or if the claim that a stream is AI-generated gets refuted.
I do not bet on my own questions.
Related questions
@traders If anyone knows of any AI-generated streams in March that would fit the description, please let me know. Otherwise I'll resolve "Q1 2024" as NO.
@Multicore As the question is written I think it will qualify. Do I understand correctly that the model will be trained on some game, then will start with a single screenshot, and will continuously predict the next frame?
I think this is good enough, provided that a) generation is happening in realtime, is not pre-recorded, b) the footage stays recognizable and doesn't devolve into garbage.
I slightly wish that I added some condition for prominence, like for example that the stream has at least 10 viewers on average, but since I didn't include it in the original statement, I think it might be too late. Oh well, it's always possible to create a new question.
@ProjectVictory Possibly. If the probability for the latest date falls below 50-60%, I’ll add options for later dates.
I think it’s possible that somebody will run a model which would be not as good as Sora, but would still fulfill the requirements.
@OlegEterevsky Sure, I was talking about the current state of bets, the market description and options are perfectly fine.