Hacker News new | ask | show | jobs
by deckarep 2211 days ago
Isn’t that the whole point of blockchain? A distributed ledger with no central authority so no single entity has the power to manipulate the history...
1 comments

Let's say you can't manipulate history (it's possible btw).

History is only as good as the one telling the story. In this case, who inserted the initial video? The FIRST person who inserted gets to say "this is truth".

With Bitcoin what's being traded existed from the beginning. It was "mined" from the original pool. Here we all collectively said, "we trust that original pool not to be shifty". So because we trust that original source, we can trade these "items" confidently.

Here the question isn't "Proof of work", it's "proof of source" which currently I cannot see a good solution for.

https://en.m.wikipedia.org/wiki/Proof_of_authority

Proof of Authority. What matters is the reputation of the source. You can have varying degrees of trustworthiness based on an source's reputation, which is built up on the chain.

For example, Craig Wright has faced a lot of backlash for claiming to be Satoshi Nakamoto. There's been no way for him to be proven wrong. Recently, someone signed a transaction with an address that hadn't moved Bitcoin since 2009 (implying it was an early adopter or contributor) and they called out Wright's lie (he had claimed to own that block).

So Wright lost credibility. That's what we need. We need the ability to develop reputation on the blockchain and a way to punish that reputation economically for posting invalid data.

EDIT (I'm out of HN replies): Reputation can be established based on a staking transaction. You create a smart contract for say, independent journalist Tim Pool. He signs it and stakes some amount and publicly declares that address to be his. He posts all his videos with a signature embedded that can be validated against that contract. The hardware itself (with algos signed by chip manufacturer) could also sign the video. An open source app could also sign. If you've got 3 signatures on a video HW, SW, and individual (associated with smart contract address) and a public blockchain timestamp, you've got a high degree of trustworthiness in that footage.

Can I reply twice? Replying to the edit:

That's a social proof token[0][1].

But again, think about bad actors gaming the trust token and then inserting their own deep fakes. It's not even far fetch.

Also, who 'gives' those tokens? What if a whole bunch of republicans give a lot of social proof tokens to Alex Jones. At one point in time, that made quite a bit of sense. So now you've got a blockchain backed authority saying Sandy Hook was a democrat red herring.

So then a bunch of democrats come in to push that Alex Jones vote back down. That same system that pushed Alex down is now a system that can push others back down in a weaponized fashion.

Separately, if you just want to say "I created this work" regardless of who I am. PGP does that perfectly well. The whole block chain infrastructure isn't necessary. It doesn't add anything to the validity of the video itself.

[0] https://ght.dtsgroup.co.nz/ [1] https://www.hubtoken.org/images/hub-white-paper.pdf

I'm not looking at is a token per se. I'm mainly interested in a Decentralized ID, which could represent an organization, like FOX or CNN. Nevermind the political biases of those orgs. Each org can establish a DID that they publicly claim is theirs and they sign all their video footage with those DIDs. Theoretically an open source camera and microphone chip could also sign a MP4 or MP3 file in a verifiable way. So you've got different levels of authenticity verification. This is just to prevent someone from taking this AI tech and making it look like an anchor said something that they didn't actually say. The timestamp of signing a transaction on the blockchain is important for proving when something occurred.

PGP depends on a certificate authority, and is quite centralized in that sense. Blockchain allows multiple open standards to exist and to develop organically. AI will continue to evolve and it's important that verification tech evolve along with it in a open source manner. The blockchain community, especially with it's recent work and research on zero knowledge proofs, are more prepared to adapt to those changes. Their whole industry is based on establishing consensus in a permissionless way.

Again a decentralized ID is all well and good. An extension of PGP/GPG which as before, works great for this person signed this video.

The problem of deep fakes still exists. Deep fake isn't just about altering, it's also about creating whole new videos that didn't happen using the likeness of a person like in OP's link.

If I upload that, sign it with my decentralized ID, CNN and Fox picks it up and co-sign it with their decentralized ID. Great! My deep fake is in the wild picked up by millions all while being completely verified.

But that won't happen because no one would believe you irjustin. Right, isn't that the crux? Who do we trust?

(if you're thinking i can't just sign any random video, it's gotta be signed with my phone - don't worry let's go back to direct feed into the camera)

I'm not sure what else to say to help you understand that if you're not willing to take a hard look at your own solution then there's nothing I can do to convince you the block chain is really good at doing some things and preventing deep fakes is not one of them.

You're never gonna solve it with 100% veracity. But look, if irjustin smart contract got created in less than 24 hours and posted the fake photo or video with no prior reputation and also the video lacks hardware signatures then you've got the markers to establish trustworthiness. That should give people enough reason to doubt it.
You've got to look at it one level deeper. Why do you trust the guy who could sign using the original BTC?

That trust is built on 2 things. PGP is trusthworthy and BTC is worth listening to.

I replied to your other post, but it's embedded in "why is BTC worth listening to"?

BTC/Eth is trustworthy because fundamentally it is a self-contained math problem that can be self tested by anyone.

Today, there is no way to make video self-testable in a trustworthy fashion because you will always be asking "Who uploaded/created this?"

Given that when you see a video recorded by someone's phone on twitter, it is probably the first and only time you will see a video from that person. What use if tracking trust if you have to accept things from unproven people all the time?