Be a property owner and not a renter on the internet

> with the right prompts.

that is doing a lot of pull. Just because you could "get the full copies" with the right prompts, doesn't mean the weights and the training is copyright infringement.

I could also get a full copy of any works out of the digits of pi.

The point i would like to emphasize is that the using data to train the model is not copyright infringement in and of itself. If you use the resulting model to output a copy of an existing work, then this act constitutes copyright infringement - in the exact same way that using photoshop to reproduce some works is.

What a lot of anti-ai arguments are trying to achieve is to make the act of training and model making the infringing act, and the claim is that the data is being copied while training is happening.

I think you're right, and I don't think it's just about public content being "exploited" to train AI models and the like. Rather, even before LLMs, there was a growing sense that publishing ideas or essays publicly is "risky" with very little reward for the very real risks.

I wrote about this a little in "The Blog Chill":

https://amontalenti.com/2023/12/28/the-blog-chill

Speaking personally, among my social circle of "normie" college-educated millennials working in fields like finance, sales, hospitality, retail, IT, medicine, civil engineering, and law -- I am one of the few who runs a semi-active personal site. Thinking about it for a moment, out of a group of 50-or-so people like this, spread across several US states, I might be the only one who has a public essay archive or blog. Yet among this same group you'll find Instagram posters, TikTok'ers, and prolific DM authors in more private spaces like WhatsApp and Signal groups. A handful of them have admitted to being lurkers on Reddit or Twitter/X, but not one is a poster.

It isn't just due to a lack of technical ability, although that's a (minor) contributing factor. If that were all, they'd all be publishing to Substack, but they're not. It's that engaging with "the public" via writing is seen as an exhausting proposition at odds with everyday middle class life.

Why? My guesses: a) smartphones aren't designed for writing and editing, hardware-wise; b) long-form writing/editing is hard and most people aren't built for it; c) the dynamics of modern internet aggregation and agglomeration makes it hard to find independent sites/publishers anyway; and d) the risk of your developed view on anything being "out there" (whether professional risk or friendship risk) seems higher than any sort of potential reward.

On the bright side, for people who fancy themselves public intellectuals or public writers, hosting your own censorship-resistant publishing infrastructure has never been easier or cheaper. And for amateur writers like me, I can take advantage of the same.

But I think everyday internet users are falling into a lull of treating the modern internet as little more than a source of short-form video entertainment, streams for music/podcasts, and a personal assistant for the sundries of daily life. Aside from placating boredom, they just use their smartphones to make appointment reminders, send texts to a partner/spouse, place e-commerce orders, and check off family todo lists, etc. I expect LLMs will make this worse as a younger generation may view long-form writing not as a form of expression but instead as a chore to automate away.

> It's less about competition and more about the ethical way to do it. If another artist would learn the same techniques and then managed to produce similar art, do you think there would be just as visceral of a reaction to them publishing their art? Likely not, because it still required skill to achieve what they did.

Now suppose that the other artist studies to learn the techniques -- several of them do -- and then Adobe offers them each two cents and a french fry to train a model on it, which many accept because the alternative is that the model exists anyway and they don't even get the french fry. Is this more ethical somehow? Even if you declined the pittance, you still have to compete with the model. Even if you accept it, it's only a pittance, and you still have to compete with the model. It hasn't improved your situation whatsoever.

> My hunch is that in the near-term we'll see a major devaluing of both written and image material, while a premium will be put on exceptional human skill.

AI slop is in the nature of "80% as good for 20% of the price" except that it's more like 40% as good for 0.0001% of the price. What that's going to do is put any artists below the 40th percentile out of work, make it a lot harder for the ones at the 60th percentile and hardly affect the ones at the 99th percentile at all.

But the other thing it's going to do is cause there to be more "art". A lot of the sites with AI-generated images on them haven't replaced a paid artist, they've replaced a site without images on it. Which isn't necessarily a bad thing.

> Any argument about this is trying to redefine copyright as the right to extract the semantic or cultural value of a document. In reality the definition is already clear - no copying of a document by any means for any purpose without explicit permission.

I've studied copyright for over 20 years as an amateur, and I used to very much think this way.

And then I started reading court decisions about copyright, and suddenly it became extremely clear that it's a very nuanced discussion about whether or not the document can be copied without explicit permission. There are tons of cases where it's perfectly permissible, even if the copyright holder demands that you request permission.

I've covered this in other posts on Hacker News, but it is still my belief that we will ultimately find AI training to be fair use because it does not materially impact the market for the original work. Perhaps someone could bring a case that makes the case that it does, but courts have yet to see a claim that asserts this in a convincing way based on my reading of the cases over the past couple of years.

>The point i would like to emphasize is that the using data to train the model is not copyright infringement in and of itself.

Interesting point - though the law can be strange in some cases - so for example in the UK in court cases where people are effectively being charged for looking at illegal images, the actual crime can be 'making illegal images' - simply because a precedence has been set that because any OS/Browser has to 'copy' the data of any image in order someone to be able to view it - any defendent has been deemed to copied it.

Here's an example. https://www.bbc.com/news/articles/cgm7dvv128ro

So to ingest something your training model ( view ) you have by definition have had to have copied it to your computer.