Fascinating read. None of it is even the slightest bit surprising. Towards the end of the article, a poll is discussed in which a majority of respondents think what AdVon did should be illegal....
Fascinating read. None of it is even the slightest bit surprising. Towards the end of the article, a poll is discussed in which a majority of respondents think what AdVon did should be illegal. Interesting thought.
I guess I'd like to see provenance on text articles posted online. A standard in which the summarized track changes can be reviewed.
Document created by User A.
Document outline pasted by User A, generated by LLM <model>
420 words typed by User B.
...
Document published.
Obviously way more in depth. But the same idea of Gen AI images tracking metadata in an invisible layer.
It'll never happen because our governments are spineless and completely ill-informed unless we're talking about technology to kill other people.
You’re underestimating the difficulty of making this happen. It’s like making DRM work for text, which won’t happen because text is too easy to copy. Even in a locked-down system where you removed...
You’re underestimating the difficulty of making this happen. It’s like making DRM work for text, which won’t happen because text is too easy to copy. Even in a locked-down system where you removed cut-and-paste altogether, a user could retype it from a photo.
Some copyright holders might be pleased, though, if it could be made to work.
A more practical goal might be some kind of opt-in system for authors who want to prove authenticity?
I’m reminded of how some musicians record everything that happens in a recording session, so there aren’t disputes about who gets credit for what. (I don’t remember where I read that.)
I feel like it should be possible for any publication above X readership to keep track of this kind of stuff, and have labels like how packaged food you buy has to have nutrition labels and...
I feel like it should be possible for any publication above X readership to keep track of this kind of stuff, and have labels like how packaged food you buy has to have nutrition labels and ingredients listed. It requires the entire company to be on board from the top down and from the inside out, but it’s certainly not impossible.
One could argue that it’s an enormous burden to change the rules, but once the rules are in place, you just don’t design processes that can’t be in compliance. Just like packaged food doesn’t have a chef wandering through the factory adding extra salt or herbs to a batch as it’s being made. The end result is required to be labelled, therefore you change the process from the very first step to arrive where you need to be.
Sure, many publications probably use systems that have some kind of version control already. I haven’t heard of a system that automatically keeps track of where someone pasted text from, though.
Sure, many publications probably use systems that have some kind of version control already. I haven’t heard of a system that automatically keeps track of where someone pasted text from, though.
You could probably implement this using version control or revision control. I assume you mean people could "cheat" though? It would be nice of search engines started enforcing something like this...
You could probably implement this using version control or revision control. I assume you mean people could "cheat" though? It would be nice of search engines started enforcing something like this in the search results and delisting sites that cheat.
Unfortunately it's very difficult for a search engine to distinguish between the original, a legitimate copy (syndication, etc) and one that's illegitimate. They all look the same. Lyrics and...
Unfortunately it's very difficult for a search engine to distinguish between the original, a legitimate copy (syndication, etc) and one that's illegitimate. They all look the same. Lyrics and recipe sites are kind of an extreme case of that. Dates can't be trusted.
There are unreliable signals, but delisting the wrong site would be an injustice. Consider how much criticism YouTube gets over bad copyright takedowns.
A publisher can enforce some things. On Tildes, the username and date on a comment can be trusted. When looking at a random website from the outside, though, it could all be fake.
Certificate transparency is an example of how to how to detect certain kinds of publisher shenanigans from the outside, so perhaps something like that could be done to get a backed-up, trusted log of what a website publisher published. (And blockchains too, but certificate transparency shows that cryptocurrency isn't necessary for this.)
This idea was touted before. Sources of info backed by a blockchain to make them traceable. But these are information and not the creative style. That said, it'll be a very hard pill to swallow...
This idea was touted before. Sources of info backed by a blockchain to make them traceable. But these are information and not the creative style.
That said, it'll be a very hard pill to swallow for many for such a thing as ownership of creative style, which is to say that humanity is doomed. I mean, a lot of online content write almost the same way nowadays anyway. AI is just a symptom of something else. People pretend as if there's no template being followed when articles are written, but, really, we're just kidding ourselves.
+1 for version control of a sort. Toss it into the marketplace, ya know? Maybe the first media company that really lands a transparent version control system on articles that people trust wins in...
+1 for version control of a sort. Toss it into the marketplace, ya know? Maybe the first media company that really lands a transparent version control system on articles that people trust wins in the marketplace. Maybe that creates pressure on other firms that choose to never share their version control.
Oh, yeah, I was absolutely oversimplfying intentionally. Writing a comment on mobile will do that to ya :) I agree with you. Provenance of articles won't happen for a bevy of reasons, only one of...
Oh, yeah, I was absolutely oversimplfying intentionally. Writing a comment on mobile will do that to ya :)
I agree with you. Provenance of articles won't happen for a bevy of reasons, only one of which is our unwillingness to collaborate and write legislation. There is absolutely a pile of technical reasons it's nigh-impossible.
Having talked with people that implement DRM on systems I can tell you they all laugh when talking about it because they know how unsolvable the problem is.
Having talked with people that implement DRM on systems I can tell you they all laugh when talking about it because they know how unsolvable the problem is.
Fascinating read. None of it is even the slightest bit surprising. Towards the end of the article, a poll is discussed in which a majority of respondents think what AdVon did should be illegal. Interesting thought.
I guess I'd like to see provenance on text articles posted online. A standard in which the summarized track changes can be reviewed.
Obviously way more in depth. But the same idea of Gen AI images tracking metadata in an invisible layer.
It'll never happen because our governments are spineless and completely ill-informed unless we're talking about technology to kill other people.
You’re underestimating the difficulty of making this happen. It’s like making DRM work for text, which won’t happen because text is too easy to copy. Even in a locked-down system where you removed cut-and-paste altogether, a user could retype it from a photo.
Some copyright holders might be pleased, though, if it could be made to work.
A more practical goal might be some kind of opt-in system for authors who want to prove authenticity?
I’m reminded of how some musicians record everything that happens in a recording session, so there aren’t disputes about who gets credit for what. (I don’t remember where I read that.)
I feel like it should be possible for any publication above X readership to keep track of this kind of stuff, and have labels like how packaged food you buy has to have nutrition labels and ingredients listed. It requires the entire company to be on board from the top down and from the inside out, but it’s certainly not impossible.
One could argue that it’s an enormous burden to change the rules, but once the rules are in place, you just don’t design processes that can’t be in compliance. Just like packaged food doesn’t have a chef wandering through the factory adding extra salt or herbs to a batch as it’s being made. The end result is required to be labelled, therefore you change the process from the very first step to arrive where you need to be.
Sure, many publications probably use systems that have some kind of version control already. I haven’t heard of a system that automatically keeps track of where someone pasted text from, though.
You could probably implement this using version control or revision control. I assume you mean people could "cheat" though? It would be nice of search engines started enforcing something like this in the search results and delisting sites that cheat.
Unfortunately it's very difficult for a search engine to distinguish between the original, a legitimate copy (syndication, etc) and one that's illegitimate. They all look the same. Lyrics and recipe sites are kind of an extreme case of that. Dates can't be trusted.
There are unreliable signals, but delisting the wrong site would be an injustice. Consider how much criticism YouTube gets over bad copyright takedowns.
A publisher can enforce some things. On Tildes, the username and date on a comment can be trusted. When looking at a random website from the outside, though, it could all be fake.
Certificate transparency is an example of how to how to detect certain kinds of publisher shenanigans from the outside, so perhaps something like that could be done to get a backed-up, trusted log of what a website publisher published. (And blockchains too, but certificate transparency shows that cryptocurrency isn't necessary for this.)
This idea was touted before. Sources of info backed by a blockchain to make them traceable. But these are information and not the creative style.
That said, it'll be a very hard pill to swallow for many for such a thing as ownership of creative style, which is to say that humanity is doomed. I mean, a lot of online content write almost the same way nowadays anyway. AI is just a symptom of something else. People pretend as if there's no template being followed when articles are written, but, really, we're just kidding ourselves.
+1 for version control of a sort. Toss it into the marketplace, ya know? Maybe the first media company that really lands a transparent version control system on articles that people trust wins in the marketplace. Maybe that creates pressure on other firms that choose to never share their version control.
Oh, yeah, I was absolutely oversimplfying intentionally. Writing a comment on mobile will do that to ya :)
I agree with you. Provenance of articles won't happen for a bevy of reasons, only one of which is our unwillingness to collaborate and write legislation. There is absolutely a pile of technical reasons it's nigh-impossible.
Having talked with people that implement DRM on systems I can tell you they all laugh when talking about it because they know how unsolvable the problem is.
A polished turd is still a turd