Content
@
0 reply
0 recast
0 reaction
Jimmy
@zkp
unless polymarket has a resilient oracle, this will happen again, inevitably on a larger scale https://x.com/ag123321ga/status/1828986766285484364?s=46&t=7FslMF3ptgVZsph362UJJg
2 replies
0 recast
2 reactions
kevin j
@entropybender
i went on a gpt bender a while ago tearing apart Uma piece by piece lol agree that there needs to be a better oracle. we were working on one before but difficult to nail the security aspect
1 reply
0 recast
2 reactions
Jimmy
@zkp
what is the biggest challenge with security?
1 reply
0 recast
0 reaction
kevin j
@entropybender
the usual game of how to define truth by putting in economic incentives like you bond $750 or so to participate in resolution in Uma i believe, then if that ends up getting into a dispute essentially the way it goes (at least from my convo with gpt + uma docs) is a smaller crowd with more money staked reviews the dispute. then if they can't decide, an even smaller crowd with more money staked reviews. and so on...if you recursively do this then the asymptote reached is a final decision maker of N=1 with the most amount of money staked but that doesn't mean even he will decide correctly. oracles are much like the courts/justice system and it's extremely difficult to make the right decision in case of a dispute, people make errors all the time. just introducing incentives doesn't mean it somehow supercharges humans to incorporate all data and make the correct decision. it's not like a stake of $100,000 vs $750 means that person's IQ or reading comprehension or even penchant for evil changes
1 reply
0 recast
0 reaction
Jimmy
@zkp
thats interesting! so i guess your point is even if it gets down to a small group of people with a lot of stake, someone can still just hit the wrong button or something
1 reply
0 recast
0 reaction
kevin j
@entropybender
yes. if uma were to work properly given their current model, there would need to be stringent criteria on how gets hired. they can't scale their recruitment you would want extremely tight, basically ironclad rules when the market is proposed so everyone knows exactly how it's going to go. then all the voters need to be at a certain level of competency and without bias. specifically, they need to at least be competent enough to understand the rules perfectly and able to apply the rules to the given decision say there was a prediction market on something extremely technical like a scientific breakthrough, maybe if lk99/superconductors become real. if 90% of Uma voters are 3rd world country people with no technical education or understanding, like have never read a scientific paper before, can't even answer basic physics/chemistry questions- does it really make sense to have them vote? it just doesn't scale, and simply increasing the amount of Uma bonded/the rewards given doesn't change that
0 reply
0 recast
1 reaction
