On 11/23/23, a 4chan post claimed that OpenAI had made cryptographic breakthroughs. Since then, threads have appeared on reddit discussing this post. A couple examples:
https://www.reddit.com/r/singularity/comments/184z6yp/i_think_there_is_big_chance_that_the_letter_leak/
This resolves to YES if credible evidence is publicly available by 12/31/2024 of either of the following:
An advancement from an OpenAI model of a ciphertext only attack that can provide the plaintext from an AES-192 ciphertext .
An advancement from an OpenAI model of a preimage vulnerability in MD5 with a computational complexity of at most 2^42
Credible evidence may consist of public demonstrations of the attacks, reporting from established media companies confirming these breakthroughs, official statements by OpenAI or government agencies, and the like. Additional anonymous leaks, secondary analyses of purportedly leaked documents, Twitter/YouTube commentary on the plausibility of such a result are not considered credible evidence.
I may change the resolution criteria within the next week (by 11/4) based on feedback from commenters. Generally, my goal is to proxy whether it will be an establishment belief that OpenAI made these kind of breakthroughs and the leak is real, and avoid conspiracy-theory levels of evidence.
Edit: Not hearing any feedback, I did not change the resolution criteria.
Related questions
Found that an analogous market was created just before mine, but apparently I didn’t have the right search terms to find it (I searched for Q* and 4chan). https://manifold.markets/LachlanMunro/did-an-openai-model-crack-aes192-en?r=VHljaG9uTm90b3M
Reddit post against: https://www.reddit.com/r/singularity/comments/186ab07/the_q_4chan_leak_is_nonsense/
Another Reddit post: https://www.reddit.com/r/singularity/comments/1867bqx/there_is_only_8hour_time_window_between_the_first
And a comment from jimmy apples: https://twitter.com/apples_jimmy/status/1727735063562096981
The problem here is the resolution criteria… If this turns out to be legitimate, it would almost certainly be kept under wraps for plausible deniability reasons. The actual question then turns into "would US intelligence allow a US company to publicly announce that they broke AES", and I don't see a realistic way that happens.
@Karu If you have a better way of capturing the idea of this becoming the widely-accepted establishment belief (rather than a conspiracy theory) I'd be interested in hearing them, per the note at the end of the market description.
Of course, one should keep this in mind when interpreting this market probability.