This market resolves as "Yes" if, by December 31, 2030, artificial intelligence (AI) systems have demonstrated the capability to predict major natural disasters (such as earthquakes, tsunamis, hurricanes, tornadoes, volcanic eruptions, and floods) with at least one week's notice.
For the resolution to be "Yes," these predictions must be consistently accurate, with no unforeseen major natural disasters occurring in the years 2028, 2029, and 2030.
A "major" natural disaster is defined as one resulting in 1,000 or more deaths or causing significant disruption, such as the evacuation of over 100,000 people, substantial destruction of property, or severe economic impacts.
Related questions
@SimoneRomeo say I made a scraper of various daily weather reports and feed that to GPT and GPT can filter out the report that says there will be a storm soon, how much do I attribute GPT to that prediction?
IRL I think it's more complex, where current-day predictions already leverage computing power. It's very human-in-the-loop and I don't expect a sudden shift, so 2030 will probably have more AI helping with the predictions but plausibly there will still a human supervisor, curator or a human doing some kind of preprocessing.
How much should the AI do, for this market to resolve to "yes"?
@Jono3h I wager that computers already play a crucial role in weather prediction accuracy, that's not necessarily DL-based, but people put the label "AI" on a lot of different software.
Also, how accurate is accurate?
@Jono3h If there's a model that accurately predicts natural disasters a week in advance, then you can vote true and point out the evidence. It doesn't have to be based on deep learning. Any kind of artificial intelligence counts.
As for what "accurate" means in this case, I'll leave you with the answer or ChatGPT:
An accurate prediction means that the AI system must correctly forecast the type of natural disaster (e.g., earthquake, hurricane, flood, etc.), the geographical area affected within a 50-kilometer radius for localized events (like earthquakes or tornadoes) or within a 200-kilometer radius for broader events (like hurricanes), and the timing of the event within a window of +/- 12 hours from the predicted start time.