5

There have been suggestions of proof of work requirements that exclude GPUs or ASICS (or try to) but I was wondering if anyone had suggested a proof of work that is not suited to being solved by computer at all, so that the miners would be humans working on proving their own work done - work that cannot (currently) be performed by a computer in a competitive amount of time.

This would presumably need to be quickly verifiable by computer, but not independently solvable by computer.

If this hasn't yet been proposed, is there any apparent reason this would not work?

2
  • It won't work because we haven't found anything simple enough for the majority of humans to do, while being both computer-verifiable and too difficult for a computer to do (see: all the failed variations on CAPTCHA).
    – Mark
    Oct 28, 2015 at 21:42
  • @Mark there is a very easy solution, but it is hard to realize it. Let humans do the verification, not computers. Pair every person in the world 1-on-1, to do a simultaneous verification event. Do the event monthly, mix all proofs from month to month, making it fully anonymous. Simple enough for the majority of humans to do, difficult (impossible really, but, at least difficult) for a computer to do.
    – BipedalJoe
    Aug 20, 2023 at 22:45

5 Answers 5

3

PoW needs to be hard to do but easy to verify (by a computer). There is very little left that humans are better at than computers (OCR, face and image recognition, IBM Watson [1] etc.). And probably nothing is easily computer verifiable.

One example are captchas. The ones that don't get read by OCR, simply get outsourced to children in poor countries. You might say: still proof of human work, but there is no way for a computer to verify the result. On top of that captchas don't satisfy some other requirements for PoW.

One such requirement is that the work is progress-free. Another: whatever the "puzzle" is, it must not be known beforehand as that would allow people to start the race early. This also implies there may not be one person or group of people that have access to the puzzles in advance as they would have to be trusted to not cheat (i.e. the system would not be trustless).

[1] https://en.m.wikipedia.org/wiki/Watson_(computer)

2

Yes, there is a paper about: "Designing Proof of Human-work Puzzles for Cryptocurrency and Beyond" https://eprint.iacr.org/2016/145.pdf

2
  • The linked paper addresses this exact question and is very interesting. However, it would be helpful to also have a description of the concepts here in the answer. Otherwise this seems more like a comment. Mar 23, 2016 at 14:05
  • The solution in the paper depends on cryptographic computer program obfuscation. Unfortunately, cryptographic code obfuscation is known to be impossible since around 2000 (there is a mathematical proof that these programs cannot be obfuscated). To make matters worse, cryptographic obfuscation goes far beyond any of the cryptography that we have today since a good enough cryptographic code obfuscator will be able to solve all of the problems in cryptography. The technology required to make this proof-of-work puzzle work simply does not exist. May 26, 2017 at 20:15
2

Something like this has been created and it is called Idena.

Idena is a novel way to formalize people on the blockchain. It does not collect or store personally identifiable information. Idena proves the humanness and uniqueness of its participants by running an AI-hard Turing test at the same time for everyone around the globe.

The Idena blockchain is driven by proof-of-person consensus: Every node is linked to a cryptoidentity, one single person with equal voting power.

Website: https://idena.io/

enter image description here

0

I designed one between 2015 and 2018, and had implemented it in full by 2021. It achieves the same goals as "proof of human work" concept. It gained some popularity in 2015 and likely led to some of the other similar projects that you've seen, that started later. I eventually named it after Bitcoin: https://bitpeople.org

The "work" done by "miners" (that grants them right to sign and publish a block), is: video chat Turing test, more or less. The imitation game, as Alan Turing called it. This is the hardest "digital" problem for a computer to solve, and therefore the most secure digital "proof of human work" (proof of unique human, granted that the idea to use "human work" for mining relates, somehow, to one-person-one-vote analogous with one-cpu-one-vote in proof-of-work) that can be generated.

-1

There are many different schemes of Proof of Human Work (PoH or PoHw), though https://www.miftycoin.com is the only one known to be a real implementation publicly available. In the case of MiftyCoin, block validation is performed by humans solving puzzles. Each puzzle is unique to each solver, as it is a function of the block identifier, the solver’s address hash and the transaction hashes.

A whitepaper on MiftyCoin can be found here: https://arxiv.org/pdf/2211.14444

Disclaimer: I am the creator of MiftyCoin.

8
  • Agreed. Done. Thanks for the suggestion! Jan 23, 2023 at 1:45
  • Your paper makes no sense. You assert that human capacity exceeds machine capacity to solve your problem, but it doesn't. Both humans and computers would use trial and error to solve it, and computers can try thousands of times more frequently than a human could. How many attempts per second do you think a human could make? Fewer than one, no? And a computer could easily do many attempts per second. Jan 23, 2023 at 21:32
  • You are correct if the assumption is that attempt is only by trial and error; in which the machine will always win, but only if given enough time. For a 24-tile puzzle, each tile has 4 possible moves, then the total number of trials is 4^24 = 281,474,976,710,656 moves. Even with a very fast single-processor unit that can process 1x10^9 trials per second, this amounts to more than 281,000 seconds = 3 days which is far from the 10-minute limit. Feb 2, 2023 at 15:01
  • So then is the paper complete nonsense? Or do you think there's some method humans could use to solve these problems that computers could not use? Your paper doesn't explain its sole claim -- that humans will be better at solving these problems than computers are. How do you expect humans to solve them? Feb 2, 2023 at 18:55
  • Finding an optimal solution to a problem with a large search space like the 24-tile puzzle is an NP-hard problem; in our case there can be more than one solution, as the goal is not to find a specific arrangement but an arrangement that brings the highest score. This paper is a good start to learn more about the type of problem we are solving: courses.cs.washington.edu/courses/csep573/10wi/korf96.pdf Feb 5, 2023 at 4:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.