[JohnFen]

> For example, LLMs could attempt to cite sources, or share ad revenue fractionally with all the sources of that inform the response they're presenting.

Neither of which address my problem: how do I share with people generally without sharing with AI?

[tjpnz]

At the risk of sounding flippant you might print your articles out onto sheets of paper and send them to interested parties by mail.

I'm sure a standard not unlike robots.txt will emerge. That might give some comfort, although I would remain sceptical given that many crawlers refuse to honour it.

[JohnFen]

> you might print your articles out onto sheets of paper and send them to interested parties by mail.

Or, better, do what I've actually done and make the websites private, invite-only.

> I would remain sceptical given that many crawlers refuse to honour it.

yeah, a robots.txt-like solution isn't adequate for just that reason. I don't rely on robots.txt alone to stop crawlers because of your observation here.

[magoghm]

I don’t understand why you wouldn’t want to share with AI. Sharing with AI makes it easier for other people to benefit from what you shared.

[dougb5]

It certainly makes it easy for OpenAI, Microsoft, and Google (etc.) to benefit from what I shared, charging a toll to end users and buffered from any consequences of sharing it incorrectly. If I had some assurance they'd link back to my content so that users could see the primary source material, and if they did all this for free, I'd be keener to share.

[JieJie]

It seems clear that there will be an initial contraction that we're seeing now, with people being distrustful of others benefiting from their work.

I've been doing art for decades, and so much of what I did in the past got merged into culture without much in the way of remuneration, even when I did get paid. Commercial and fine artists who make money off their works are rare, and the main benefactors were large corporations long before OpenAI came to pick the bones clean.

As we circle the event horizon (personally, I'm with the people who argue we passed the point of no return back in the 1930s), it will get more difficult to tell what's going to happen next, but everything only has to be added to the training data once. To a determined attacker, there is no data fortress that can't be raided, and it only has to be raided once.

The old hacker motto, "Information wants to be free", wasn't an ideal to work towards, it was a statement of fact: keeping information locked up is hard and it only has to get loose once.

The problem of how to get paid has always been the main problem facing people who work. I suspect with compensation, like everything else, we'll do the right thing after we've exhausted all other options.

[JohnFen]

I don't want to share with AI because if what proponents of AI are predicting is correct, I think it will result in very bad things. I don't want to have contributed to that, even a little.

[wombatpm]

Rocko's Basilisk will thank you.

[JohnFen]

It's a good thing that Rocko's Basilisk makes about as much sense as Pascal's Wager.

[nsedlet]

Yeah it feels like there should be a legally enforceable ai.txt