Resolves to the next major competitive sport where AI beat the existing world champion
I use a very broad definition of "sports" here, to include all sports in mind sport Olympics, math olympics, programming competition, etc.
If AI's strength has already surpassed top human player on Feb 24,2024 the answer would not count (i.e. it will resolve to 0)
Current sports in which AI has managed to beat the world champion:
Chess: Perhaps the most famous example is IBM's Deep Blue defeating World Chess Champion Garry Kasparov in 1997. This was a landmark event, marking the first time a computer chess program beat a reigning world champion under standard chess tournament conditions.
Go: Google's AlphaGo made headlines in 2016 when it beat Lee Sedol, one of the world's top Go players, in a five-game match. This was particularly significant because Go is a complex game with an astronomical number of possible positions, and it was believed that AI would need at least another decade to beat human champions.
Poker: In 2017, Libratus, an AI developed by Carnegie Mellon University, beat four of the world’s top professional poker players in a game of no-limit Texas Hold'em. This was a significant achievement because poker is a game of incomplete information and requires bluffing, making it a complex challenge for AI.
StarCraft II: AlphaStar, developed by DeepMind, beat professional StarCraft II players in 2019. StarCraft II is a real-time strategy game that requires long-term strategic planning and quick tactical decisions, showcasing the AI's ability to handle both strategic depth and real-time analysis.
1,000
3.00@AmmonLam I imagine it would be something like this https://www.youtube.com/watch?v=ts5lPDV--cU beating the current world champion in Geoguessr
@ii hmmm I wonder if this might be a little bit of a low hanging fruit, given the nature of Geoguessr AI do have an advantage over human player because AI have access to pictures on Google map. I think I'm going to NA Geoguessr and not include it in this MC market just to keep things interesting and not easy to predict.
I encourage you to create a market about when AI would beat top human Geoguessr (if you are not going to create such market, I will)
well, they do have access, but it's not like these AIs are looking at every location on Google Maps. Right now the top AI is probably stronger than the top human player (In the above video Rainbolt (not the world champion, but still a "pro") won 1 out of 8 games), but I don't think there has been a direct game between them.
@ii I think what would be interesting is, if an AI is only given a limted number of images to train, and set it such that it is about the same as the number of images around the world a human player would have seen in their lifetime (so something like, 36500 images?) and then see how good the AI could do against a human player