3 ms·

AI is the use case blockchain didn't know it was looking for. The exaflod of content and deep fakes, etc. that's coming towards everyone soon will require some

AI is the use case blockchain didn't know it was looking for. The exaflod of content and deep fakes, etc. that's coming towards everyone soon will require some sort of trust protocol, blockchain is great for that. With regards to crypto specifically microscopic payments will be needed to allow for adding cost to a lot of the things that are currently "free" now that AI content creation is going to overwhelm all of us.

quotemstr 4y ago
> The exaflod of content and deep fakes, etc. that's coming towards everyone soon will require some sort of trust protocol, blockchain is great for that. No it isn't. Blockchains have no solution to the oracle problem. A blockchain is useless for distinguishing AI-generated content from other content for the same reason a blockchain is useless for managing mortgages on real property.
- ThomPete 4y ago
  There is no solution to the oracle problem. But there is a solution to the trusted protocol problem by only trusting content created through transformers as you will be able to see the intent it was created with.
anonymouskimmer 4y ago
> "The exaflod of content and deep fakes, etc. that's coming towards everyone soon will require some sort of trust protocol, blockchain is great for that." I don't see how this is the case. Any blockchain verification is one off, and limited to the chain it is on. I, and many other people, would probably more trust a variety of centralized claims to truth or trust, than a decentralized claim. You trust when multiple parties state how they verified, you don't trust when one party writes "this is true" to a decentralized service, without all of the evidence they used to verify as true. Crypto, except transactions on centralized exchanges, is hideous for microscopic payments. It costs too much to verify even minimal transactions on chain (even more for contracts). It's a lot, lot cheaper to use a centralized service for minimal transactions.
- ThomPete 4y ago
  You are not wrong but try and turn it around: AI creates all the content and the output gets put on the blockchain. I can now see what the intent was when it was created and if it wasn't created via an AI. As a user that's the most secure way to interact with any content.
  - anonymouskimmer 4y ago
    Maybe, but either you don't need to buy any tokens to interact with the content. Or you have no guarantee beforehand (minus a prohibitive chain analysis) that what you're interacting with won't rip you off. Also an AI (or a person hacking the AI or its data stream) could just as easily copy something that's non-AI generated and publish it to the chain as the AIs own creation. If this something came from a private source there might not be any way to verify the original creator. So this really only works for full verification if literally every bit of information of a kind (or at least its hash) is placed into the blockchain in real time, and the calculations generating the information and data streams placing the information into the chain are transaction-secured along their entire route to the chain. This becomes a tax on everyone, for the benefit of a few.
    - ThomPete 4y ago
      I don't need to know who the original creator is, I just need the smart contract or smart agent be created via an LLM then I will know exactly with what intent it was created and whether to trust it or not. It's literally written in plain english. It's not obvious I know, but once you accept the idea that LLMs will become the tool we create with (and tokenize) and that smart agents becomes the mediators then you realize that there is in fact a way to allow for this and that blockchain is going to be fundamental. In fact if it's not created via an LLM on the blockchain we know we have to be much more careful.

quotemstr 4y ago

> The exaflod of content and deep fakes, etc. that's coming towards everyone soon will require some sort of trust protocol, blockchain is great for that. No it isn't. Blockchains have no solution to the oracle problem. A blockchain is useless for distinguishing AI-generated content from other content for the same reason a blockchain is useless for managing mortgages on real property.

ThomPete 4y ago
There is no solution to the oracle problem. But there is a solution to the trusted protocol problem by only trusting content created through transformers as you will be able to see the intent it was created with.

anonymouskimmer 4y ago

> "The exaflod of content and deep fakes, etc. that's coming towards everyone soon will require some sort of trust protocol, blockchain is great for that." I don't see how this is the case. Any blockchain verification is one off, and limited to the chain it is on. I, and many other people, would probably more trust a variety of centralized claims to truth or trust, than a decentralized claim. You trust when multiple parties state how they verified, you don't trust when one party writes "this is true" to a decentralized service, without all of the evidence they used to verify as true. Crypto, except transactions on centralized exchanges, is hideous for microscopic payments. It costs too much to verify even minimal transactions on chain (even more for contracts). It's a lot, lot cheaper to use a centralized service for minimal transactions.

ThomPete 4y ago
You are not wrong but try and turn it around: AI creates all the content and the output gets put on the blockchain. I can now see what the intent was when it was created and if it wasn't created via an AI. As a user that's the most secure way to interact with any content.
- anonymouskimmer 4y ago
  Maybe, but either you don't need to buy any tokens to interact with the content. Or you have no guarantee beforehand (minus a prohibitive chain analysis) that what you're interacting with won't rip you off. Also an AI (or a person hacking the AI or its data stream) could just as easily copy something that's non-AI generated and publish it to the chain as the AIs own creation. If this something came from a private source there might not be any way to verify the original creator. So this really only works for full verification if literally every bit of information of a kind (or at least its hash) is placed into the blockchain in real time, and the calculations generating the information and data streams placing the information into the chain are transaction-secured along their entire route to the chain. This becomes a tax on everyone, for the benefit of a few.
  - ThomPete 4y ago
    I don't need to know who the original creator is, I just need the smart contract or smart agent be created via an LLM then I will know exactly with what intent it was created and whether to trust it or not. It's literally written in plain english. It's not obvious I know, but once you accept the idea that LLMs will become the tool we create with (and tokenize) and that smart agents becomes the mediators then you realize that there is in fact a way to allow for this and that blockchain is going to be fundamental. In fact if it's not created via an LLM on the blockchain we know we have to be much more careful.