The Industrial Revolution & LLM's

In the software engineering community, LLM's are often compared to the industrial revolution, or seen as a second industrial revolution. I've heard this used as some sort of fact in discussions about AI and LLM's for years now, and I'm growing tired of it. There is one important difference few seem to mention:

While the industrial revolution was built on technical innovation, LLM's also essentially need models, which are trained on about every bit of available published knowledge the model authors can find. This published knowledge is the product of decades of human work. And most of it wasn't taken with permission (or compensation).

Recently, I listened to a podcast where Jon Stewart put it succinctly:

I think they're strip mining our intellectual property in the first place. The whole point of AI is it's training itself on the entirety of human existence. And yet they get to put it all behind proprietary walls, and we don't see either a shareholder stake or anything about it. In fact, we have to pay them for the privilege.

• Jon Stewart (source)

You might think by reading this post, it makes me "anti-AI", but that wouldn't be the thruth. I do believe that the technology behind LLM's is rather interesting. But the part where the IP of so many people was just ripped off is completely morally wrong (and perhaps lawfully too).

Some may think: but a model does need a foundation, you have to get IP (such as code to train on) somewhere. To that my answer is: those who want to profit from it by using the models, should be required to license their codebases to be trained on. Then you get a mutual benefit out of it all. Because a world where companies don't want models to be trained on their private codebases, but want to use models which were trained on ripped off IP from the internet and just look away, is a world that is completely upside down. IP for me and not for thee.