Topic: Connecting MTurk, distributed oracles and Bitcoin - idea (Read 2148 times)

As for the Sybil attack - MTurk workers' identities are verified by Amazon, so it's not that easy to perform such an attack. But yeah, it's the question of whether it's cheaper to have 10 super-verified super-secure oracles doing manual labour, or 1000 loosely verified people doing much a much simpler/dumber variant of labour.
Probably sometimes it will be one, sometimes it will be the other.

As for the debatability of the graphic design.. The program / questionnaire to MTurk workers will be known in advance. So, before contract signing, the designer will see all the rules: "1000 random dudes will have to say that colours are within lines, the picture is not stick figure, and that the picture is funny in scale from 1 to 10. Also, comics based on fart jokes will are to be given 0 points. If I get more than 5.0 in average, I'll get my money".

As for bets vs. contracts... I agree that there's a thin line between them. You could say to a freelancer: "I bet that you won't do this piece of software before a given deadline", but people often prefer to say: "Let's have a contract where I pay you if you do this before a deadline."
When you use "bet", it implies that gambling is involved. When you say "contract", it implies some kind of a deal between two parties, where both of them try to minimise the unknowns and they both expect the contract to pay unless something goes very very wrong.

So I read this yesterday and I still don't get what the practical applications of this are after you consider the risk and cost. Just in theory, how is it possible to prevent any type of Sybil attack when you have humans with registered accounts somewhere, or, if you can through incentives, then how will you do it cheaper than a centralized service? Also, what will prevent subjective outcome determinations - ie; say the graphic designer delivered the work but it's DEBATABLE whether or not that it was what was asked. You would need a centralized discussion about something like that, otherwise it's really risky for both parties to enter into that contract: 'oh what would a stranger think? Is this good enough or no?'

The team involved with BetMoose (launched yesterday, check it out) is working toward the same end-vision but we strongly believe you need a single platform for these discussions and the handling of outcomes. We build reputation and incentives with our "hosts run the bet" strategy. We also use the word 'bet' instead of 'contract' because of our market research on the topic, though change every 'bet' on our page to 'contract' and you got yourself a smart-contract exchange.

I published a rather long article on the subject here:

http://orisi.net/discussion/4/mturk-and-distributed-oracles-two-tiered-arbitration

The general idea is that you could take an Mechanical Turk crowd wisdom and use it to deliver rulings in smart contracts. Thoughts?