Hacker News Reader

Comment by naet | Hacker News Reader

AI feels different to me because it's "retroactively" extractive.

Social media may be extractive if I give it my attention and data to resell. Crypto is extractive to the people buying into it and losing money. But I can avoid doing those things if I want.

AI these days takes anything I've written (from literature to code) or illustrated, and retroactively extracts value from it to be repackaged and redistributed. Even if I was proactive about securing the trademark or copyright, they can seemingly just ignore it without consequence, without me ever participating or opting in to it.

It feels massively different to have social media extract value from and image I uploaded willingly, vs AI extracting value (eg making things "in the style" of me) after training on my work against my will.

View on HN

You mean like how you learned from others? Against their will?

Scale has a quality all its own.

You create a highly useful website full of dense hard earned knowledge. Some people will just use that knowledge, but a few create their own sites, using what they learned.

A company sweeps up hard earned knowledge from millions of sites, with automation but not permission, and goes a long way to making those millions of informational sites redundant without any compensation.

Simplifying things obviously. But these two scenarios are not equivalent.

Why can't we agree there are differences between other human beings learning from other people and a multinational corporation learning from everyone at industrial scale, with industrial means?

I can't po$$ibly imagine.

Truly, your reply is ridiculous in its comparison. How you could fail to see the difference between these two things, their scales and also their qualitative differences is impressive.

Still there is a similarity between humans reading stuff, thinking about it and then writng something and AI doing the same.

I think if the ai were a free agent or something nobody would have an issue, but “the ai” isn’t really the best way to summarize the thing learning. It’s just a tool or service offered by a corporation which said corporation solely profits from.

This is not an argument. This is essentially you saying "Why doesn't everyone agree with me, because I am self-evidently right?".

It's basically a store-brand argumentum ad verecundiam, casting yourself in the position of authority.

It should be obvious by now that there are many, many people who disagree with you without their arguments necessarily being "ridiculous".

I mean, I too would enjoy it if everyone agreed with all my opinions all the time, but I don't expect that to actually happen.

In this case, the argument is indeed self evidently right. Trying to equate the two things after being able to observe both in the real world is like seriously claiming that a stove fire and a nuclear blast are pretty much the same thing because they both produce light and heat. Why would anyone bother convincing anyone out of such an obviously absurd claim in a practical context?

> In this case, the argument is indeed self evidently right.

No, it absolutely is not.

> obviously absurd claim

You are obviously and absurdly wrong.

It's clear that nothing productive is going to come of this conversation.

There's a clear qualitative difference. If I post a lecture on mine on YouTube and a million people watch it, great! If a training company takes my lecture without my permission and re-sells it to a million people, that's a clear distinction.

(Based on a true story, it took me 3 years and 2 law firms to recuperate royalties for IP theft)

Did you ask the people who wrote the book and articles you read to get knowledge on a subject how they feel about a YouTuber earning ad money by repackaging their content ?

I'm a CS professor not a YouTuber. It was a lecture from a chapter of a book I wrote.

Yeah, I'm unsure why learning without author consent is judged differently between humans & AI. Some folk are justifying the distinction "because scale", as if that explains anything. Why would scale matter? And what's the threshold for scale that turns freely learning into theft? I'm not being facetious; I genuinely don't see the connection.

I would say being “derivative” or plagiarism are examples of where it is judged negatively especially if nothing novel/original is added.

Reply on news.ycombinator.com