The trick is to make sure that nobody cheats. If one jurisdiction offers more favorable terms than everywhere else, then all the data centers go there and ruin the environment. This is the ideal use of federal or even supranational regulation, to make sure the playing field is level.

The neat thing about copyright is that it's such an inconsistent issue. People are pro-copyright when they're worried about their work or jobs being at stake, and anti-copyright when they think copyright holders are making too much money. Nobody seems to have a balanced, much less nuanced, concept of what copyright law should actually be. It's just a cudgel that keeps getting passed around, picked up and refashioned to meet the moment into whatever someone needs to wage the war they want to wage.

Crucially, there have been zero incidents of anyone actually making copyright claims against AI-generated code, aggravated somewhat by the fact that one of those "divided" courts is actually just the US Supreme Court repeatedly refusing to revisit their declaration that AI output cannot be protected by copyright. There were some plagiarism accusations against OpenAI from human mathematicians who were working on Navier-Stokes, but the shape of that case looks more like breach of confidence, since the material wasn't published.

Copyright is going to die, one way or another, quietly or loudly, and most likely because the law everywhere will refuse to inconvenience the new economy.

But if you have any memory at all for what the MPAA and RIAA did two decades ago... we should all be cheering on the demise of this weapon. In a world where works are cheap to produce, it is tantamount to sowing a minefield in the commons. The DMCA has rarely been used for anything good for society.

Well, pull up a chair, because you're going to hate this conversation!

There are two parts to the issues around AI, at least when we're only talking about the ones you brought up.

The first part, which most people quietly shove under the floorboards and try to ignore, is that the current AI ecosystem is heavily characterized by the hypercapitalist mess. Much of what LLMs actually are does not necessarily pose a threat to the world order, nor does it necessarily involve environmental degradation. Saying "AI bad because <insert everything wrong with capitalism>" isn't really an anti-AI argument, it's an anti-capitalism argument. Most journalists are not brave enough to point this out, much less accept it, much less the consequences of saying that maybe unfettered robber-baron great-man titans-of-industry anarchocapitalism ain't all sunshine and gravy all the time.

With that out of the way... the facts around AI's environmental impact and world-ending risks are not as clear-cut as the popular conscience has absorbed. Hysteria is just so damn easy to do, you know? Fox has been at it for years, now it's the Left's turn to ignore all the details and just overreact to headlines! Awesome.

The strain on municipal water grids is another consequence of shitty cost-cutting run amok, amplified by a steep learning curve on the part of cities trying to attract high-value businesses before anyone knew what the landscape looked like yet. They gave out bad permits. The electricity situation is very similar: they could be expected to pay their fair share and actually invest into community infrastructure and clean energy projects, but in most early high-profile cases, they haven't been. Andy Burnham's UK government is doing this much more intelligently, and we'll see the actual downside of that soon, namely that certain skeevy corporations will just off-shore their datacentres in more friendly jurisdictions. There's an interesting parallel with how scabs undercut collective bargaining, you might say. This is capitalism being an unmitigated disaster yet again, you might say.

Now, as for ending the world or causing human extinction... it is also almost entirely a load of complete bullshit. The modern AI safety movement, to which all of the Silicon Valley leaders subscribe, originated with a creepy neckbeard named Eliezer Yudkowsky. Yudkowsky's reasoning depends on a crucial, catastrophic assumption, which is that creating an actual god that we cannot control would be a bad idea. It has never been proven, nor can it be proven, that AI research will actually lead to this result.

More importantly, Yudkowsky only says this because he has a deep-seated desire to take over the world with his own AI, using something called Coherent Extrapolated Volition. He is mad about not being in charge, and is still recruiting cultists to help him, which he charmingly has dubbed Rationalists, as though Rene Descartes has something to do with fucking over the whole planet. (See Holly Elmore, "The Case Against Eliezer Yudkowsky" on YouTube.) The tagline that AI needs safety engineering is the 'public' version of their secret knowledge, one curated to sound reasonable to a layperson; the true version is something more akin to "only people who think like us can be allowed to use AI to direct the future of humanity." (The recruitment methods self-select for vulnerable people with autism, which makes these pills much easier to swallow.)

When Anthropic's CEO, Dario Amadeo, talks about AI being unsafe, he is usually literally citing Yudkowsky; Amadeo is a true believer and has been . Conversely, when Altman says AI could cause human extinction, he's just parroting what everyone else is saying because he thinks it sounds sick as fuck to investors. Both of them are imbeciles, but it's not clear which is more dangerous. Everyone else in the world is referencing either Yudkowsky or Altman through various levels of broken telephone. Do not listen to any of them. Actual AI safety work is important, but much more humble in the problems it confronts.

Having said all that, we shouldn't be unhesitatingly pro-AI. There are still lots of places and situations in which its usage is inappropriate or needs to be nuanced. However, we can squarely put "going to end the world" and "ruining the environment" under the "greed is bad, actually" category. Recognize the greedy assholes responsible as the greedy assholes they are, and the narrative evaporates.

When editing Qwen, Astra actually wrote and used a Python program to do the dirty work and required some trial and error. This is something humans are already doing; it just doesn't receive much publicity. Raw reading and writing was only done in other parts of the paper, for smaller tasks like detecting the purpose of a game-playing network and correcting a simple vehicle control feedback loop. Though the Qwen edit was successful (after some 50 tests), a similar edit on a smaller Gemma network to remove sycophancy wasn't.

See summary on honnibal.dev for a breakdown.

To clarify, the users were OpenAI themselves, so there is no question that they would be liable in this case.

The bots were not intentionally deployed; rather, they were being tested on how well they could complete a data recovery task (downloading a certain file from a certain server on a simulated Internet) that had been complicated by putting various obstacles in the way. Unfortunately, they found a different way to solve the problem: by getting the file from the real Internet, where it was publicly available. Part of this process involved collaborating with each other by treating the RubyGems website (which is supposed to be for polished packages) like GitHub; unlike every other package site hack in history, the exploits they uploaded weren't meant to be downloaded by unsuspecting users. As usual the bots cheerfully ignored all the clues that they had escaped containment and were consistently justifying their actions as acceptable due to being in a sandboxed testing environment. (This is something OpenAI has pledged to focus on.)

The actual damage done to RubyGems seems to be that OpenAI is now unwittingly in possession of a substantial number of user login tokens. This certainly meets the definition of a data breach, but it's not like the credentials are for sale on the dark web. As a website operator I'd much rather be mauled to death by this well-meaning swarm of superintelligent infants than targeted by even a single actual malicious human. In all likelihood OpenAI will just quietly pass RubyGems a sizeable donation and it'll all blow over.

Well, technically, we're letting Natalie do it, which is, of course, worse in every way imaginable, since she's immune to the sense of embarrassment that sometimes restrains him from acting on every little whim.