Yes, good test coverage will be even more important. But tests rarely consider every possible eventuality, unless techniques are developed for AIs to prove substantial programs are correct, while not hallucinating said correctness.

Yes, source has to be a triple of prompt, AI model release, and its random seed. A patch may change all three.

Yes. Source-available commercial software can even be fully publicly forkable.

For most software and users the libre aspect of open source is more important than the gratis aspect.

Another alternative is my browser extension that emulates the Live Connect aspect of Java Applets. Live Connect let Java call JavaScript and JavaScript call Java using normal fuction calls. My extension doesn't do Applet Graphics, but they sucked anyway.

It's a good way to use a browser as the UI of a Java app. The Java app has more speed, permissions, and type-safe object-orientation, while the browser side can leverage the many browser UI libraries.

The competition between jurisdictions I spoke of will push each country's politics (and courts, especially in countries without the separation of powers) toward not being left behind on the money and military power that lightly-regulated AI can bring. Copyright holders will be expendable, especially now it's less risky to anger a weakened MSM.

I think, on balance, copyright protects the little guy more than the big guy. Without it I can see even more concentration of power in shameless marketing conglomerates.

Courts and legislators will soon decide whether AI training is fair use because it's like a human learning, or is not fair use because it often regurgitates close variants of specific training data, and, unlike humans (but like publications), can be replicated. Both damage the viability of its data sources, so I agree that the latter is closer to the truth. But competition between jurisdictions will pressure accommodation.

No original research, but summarising the research of others has been a business model long before Reader's Digest and Slashdot. At least Slashdot links.

User comments are now a profit-centre, because AI companies will buy the freely-donated writing.

YouTube does have the thumb down, but it's used less since they hid its counter. Plus there's no clarity on whether it reduces algorithmic ranking, or whether that's just based on views, where clickbait is king.

X's pay-to-win approach (buy Premium or shout into the void) is coming back to bite them.

A counter to the buggy and hard to maintain problems with AI code is that it allows entrepreneurs to quickly bootstrap their business with a good-enough solution, then use the revenue or investment that it attracts to hire humans to redo the code properly. So freelancer sites can profit from this second wave. But they do lose all the business writing software that commercially fails.

Are humans going to fight back by reducing their source-available code, putting it behind bot-proof walls, or poisoning the well by sprinkling in some incorrect code?

Rare earths also make drone electric motors much more efficient. Plus there's China's dominance in lithium batteries. So China could really hurt Ukraine by shutting off their supply of drone motors and batteries. They might threaten this if they look like getting hit with sanctions for buying Russian oil. So they may escape these sanctions while India still gets hit.

One each of course, unless you believe consciousness is a quantum entanglement from when the blastocyst split.

I agree with all that (but can't up-vote you).

And how did Slashdot respond? Stagnation. I guess their poorly-received and reversed UI overhaul, and a smugness from being the first with a voting system, led them to believe that their system is perfect and should not be experimented with.