Considering that:
a) data pollution (large scale injection of AI generated data into the information space) and subsequent model collapse have been identified [1] as potential threats [2] for future LLM's and
b) advanced AI models will impact the geopolitical power distribution [3] and therefore be increasingly subject to geostrategic contention [4],
Do you believe that by the end of 2025, there will be evidence of large scale organized data pollution operations by state or non-state actors with the implicit or explicit goal of denigrating the performance of future LLM's taking or having taken place?
Resolution:
This market will resolve as YES if at any point before 01/01/2026 credible information will emerge that a deliberate data pollution operation by any actor (state/non-state) for any reason (geopolitical contestation, ideology, terrorism, lulz) has taken place.
Caveat: the operation must be/have been significant enough to warrant mention by a reputable news source (e.g. the NYT, WSJ, WP, BBC etc.), a government communication, a peer-reviewed scientific publication, a reputable threat intelligence service provider and/or other reputable sources not covered in this list.
I reserve the right to final judgement.
1,000
3.00@ItsMe you can search for Nightshade which is emphasized in the paper as requiring very little sample share to have its intended negative effect.
@JussiVilleHeiskanen No, it's effective on a large scale with a small "dose." I don't think this market is about how efficient the pollution is, and I certainly don't think it's about only inefficient pollution, but I could be wrong!
@Panfilo "large scale organized data pollution operations" -- from description. Single site isn't that organized an operation.
@JussiVilleHeiskanen Then was there a reason you don't think the WaPo Russia story in the further back comments counts?
I bought because of this: https://manifold.markets/post/will-there-be-evidence-of-large-sca?
https://nightshade.cs.uchicago.edu/whatis.html I have seen people using this in the wild.
Is this enough to resolve yes? I'm surprised that story hasn't spread, but I haven't seen any contradicting evidence.
https://www.washingtonpost.com/technology/2025/04/17/llm-poisoning-grooming-chatbots-russia/
@SqrtMinusOne Hey, they’ve been using that for ages: feeding people disinformation in every possible way to fuel division (right–left lies, “democracy” lies, and so on).
They spend billions every year on bot farms, whole armies of bot-feeders, buying freelancers in India and Southeast Asia (they sell oil to India and get it paid back with propaganda spreading).
What do we think about this, chat?