HITs (Human Intelligence Tasks) as part of a proof-of-work algorithm

exapted

newbie

Activity: 51

Merit: 0

Quote from: Nagle on March 17, 2014, 01:14:50 PM

Quote from: exapted on March 17, 2014, 04:28:54 AM

If a chess game was integrated into the protocol, such that two players play against each other, and moves are signed with private keys, then I think the winner could be verified automatically. Beating a highly rated player would count more than beating a lowly rated player.
...
I hope I'm on the right track here.

Computers play chess better than humans now. Much better. No human player has won a tournament against a top computer program since the mid-2000s. If you buy any of the top-rated chess programs for PCs such as Fritz or Houdini and run them in full-power mode, you're going to get trounced. There are now chess programs playing at grandmaster level that run on smartphones. If you've been on the cover of Chess Life, you might win against one of them once in a while.

Yes, that was just an example of something that could be automatically verified. Go would be a more appropriate choice of games. And that would only work if: only beating players rated higher than some threshold counted as PoW (otherwise computers could compete), and enough appropriately skilled players participated.

Nagle

legendary

Activity: 1204

Merit: 1002

Quote from: exapted on March 17, 2014, 04:28:54 AM

If a chess game was integrated into the protocol, such that two players play against each other, and moves are signed with private keys, then I think the winner could be verified automatically. Beating a highly rated player would count more than beating a lowly rated player.
...
I hope I'm on the right track here.

Computers play chess better than humans now. Much better. No human player has won a tournament against a top computer program since the mid-2000s. If you buy any of the top-rated chess programs for PCs such as Fritz or Houdini and run them in full-power mode, you're going to get trounced. There are now chess programs playing at grandmaster level that run on smartphones. If you've been on the cover of Chess Life, you might win against one of them once in a while.

exapted

newbie

Activity: 51

Merit: 0

Quote from: dewdeded on March 17, 2014, 06:26:52 AM

100% useless and no intrest of research, big waste of time

Because this would be 1000% be gamed, by buying cheap human ressources from underdeveloped countries or Amazons Mechanical Turk or using botnets forcing the owners of the infected machines to solve these HI-tasks.

I think you have a good point, maybe any HIT PoW protocol could be gamed. But I am not convinced it is a lost cause. Yes, it would be vulnerable to cheap labor. If a game like Go was used, IQ and skill would be very important. POS could be combined with a HIT based PoW.

Maybe the key would be to make HITs as difficult as possible. That means that they would need to be fun. And to prevent the system from being gamed, HITs might need to be a part of a larger mathematical problem solution (so maybe collusion/"sybil" attacks could be detected at a low enough level).

dewdeded

legendary

Activity: 1232

Merit: 1011

Monero Evangelist

100% useless and no intrest of research, big waste of time

Because this would be 1000% be gamed, by buying cheap human ressources from underdeveloped countries or Amazons Mechanical Turk or using botnets forcing the owners of the infected machines to solve these HI-tasks.

exapted

newbie

Activity: 51

Merit: 0

If a chess game was integrated into the protocol, such that two players play against each other, and moves are signed with private keys, then I think the winner could be verified automatically. Beating a highly rated player would count more than beating a lowly rated player. The rating system would need to be secure, and some election process would need to determine a system-wide winner/leader set, or something similar.

Another idea is that an election protocol verifies HITs by randomly polling users, such that users sign their votes with their private keys and the votes need to be nearly unanimous.

I hope I'm on the right track here.

Rannasha

hero member

Activity: 728

Merit: 500

One of the properties of the PoW scheme that Bitcoin uses is that it is (computationally) hard to complete the PoW, but easy to verify it. Every node verifies the PoW of a block to ensure it is correct. This has to be done automatically for obvious reasons. If HIT is a component of the PoW, then nodes can't automatically verify the correctness of a block.

exapted

newbie

Activity: 51

Merit: 0

Has there been any work on designing a proof-of-work algorithm dependent on HITs (Human Intelligence Tasks) ?

An example could be a proof-of-work scheme based on captcha solving. (Assume computers can't solve captchas. Obviously there are captcha-solving systems that work without human intervention.)

The HITs would need to be such that humans are much better than computers at completing them. There might be a mathematical puzzle that reduces to suitable HITs. Those HITs could be presented to humans in a game interface in a client app. Alternatively (maybe no such puzzle is found), there might be some informal HITs that could be verified through a voting system. Another idea that popped into my head is that a challenging game such as Go could be implemented into the protocol of a cryptocurrency, and beating a highly rated player would count as proof-of-work.

EDIT:
Here are some potential reasons such a PoW protocol could be valuable:
1) If enough people were to participate, even money and computational resources would not be enough to do a 51% attack.
2) The HITs could be implemented into games. The games could get a share of mining proceeds.
3) The HITs could be useful, creating a distributed crowdsourcing platform.

Topic: HITs (Human Intelligence Tasks) as part of a proof-of-work algorithm (Read 919 times)