We accidentally solved robotics by watching 1M hours of YouTube

> We evolved with pneumonia for some reason. It could easily be a Chesterton Fence.

It's not, and I don't think you're seriously arguing this point so I'm going to ignore it.

It is, I think, a reasonable observation that had Swartz formed an LLC to pursue advanced analysis of academic papers for, I don't know, trends in the language used in research and slurped bunch of JSTOR for that purpose, the trial would have taken longer and involved more lawyers. That's probably an observation that should give us pause. Or not, because nobody argued that's what he did or that was his intent, including him. So I also think the premise of comparison to the current circumstances is flawed; I don't think the CFAA can be applied in a context where people have access rights and go through Google's front door to scan videos for the purpose of training a machine learning algorithm. It might be a TOS violation. It's not hiding a server in a closet with unauthorized physical access, which is what Swartz was accused of.

Intent matters, and, sadly, we never got to the trial where intent could have been proven out.

> Being hassled to the point where you go through with it is definitely violence.

The government does have the monopoly on violence. But I think what happened to Swartz is a far cry from that, as he never got to sentencing, much less trial. There was some light compulsion (requirement to appear in court), of course. But everyone who's ever wanted to contest a parking ticket has to experience that. Sadly, this train of thought goes into a station of "Swartz should have been under professional care if his condition was this much a danger to him," and I don't know how the government should change its behavior if he wasn't. Prosecutors are not prognosticators of the mental health of defendants, and I've never read anywhere that Swartz wanted to be committed for mental illness.

Our system is much harder for defendants grappling with mental illness; I'll acknowledge and argue for change regarding that. I don't know that such change would conclude with "Swartz should never have been accused of committing a crime that a lot of evidence suggests he committed," however.

We haven't gotten to the part of the trial for Anthropic yet where we determine whether they actually broke the law when they downloaded from pirate sites. Copyright has multiple exceptions. And on the topic at hand here (training on YouTube videos to understand space and relationships in it), I don't think even Google would want to make the case that it's a violation of copyright.

That's the thing about copyright; it's a whole category of law more based in utility than morality. One of the reasons AI is such a fight right now is that nobody was opposing it as an academic project when it was generating, for example, tools that could go from an image to describing the image, or from an image to recognizing the likely artistic style and helping somebody find the original artist. But with just a few tweaks those tools became devices for generating novel images, and now people are upset. Intent matters.

And again, you are drawing equivalence between harvesting data from openly accessible sources online and hiding a server in a closet with unauthorized physical access to a network. Swartz's prosecution wasn't accusing him of copyright violation; it was accusing him of compromising a network. A far more serious charge; if the researchers in the story here had collected those YouTube videos by wiretapping the fiber optics between two of Google's data centers I suspect they would have concerns.