OTOH, one strain will always be more popular than the rest, so that will be the most profitable.

Not necessarily. Compare with apples, where many varieties thrive. Some of that is growing area, different apples do well in different environments. But consumer preference is a big part. Some people like sweet apples, some like tart, some like sour. They have varying degrees of crispness. Different recipes call for different varieties.

So it is at least possible. Single-strain monoculture is not the inevitable result of profit / efficiency over all else.

The real debate is if governments have the right to "shape" consumer choices with policy, subsidies, and the force of law in the aggressive ways they've been doing using taxpayer money.

Of course govt has the right. It's in the Constitution, clear as day. What part of "regulate commerce between the states" don't you understand? Article 1 clause 8 section 3.

Should they? That's a legitimate question. You can have a policy debate about whether the govt should (I'm firmly on the yes side). But can they? Settled question since 1789. Find another horse to flog.

Their music is their property and it is protected by copyright (because that is the law against stealing for this kind of property).

You don't understand copyright at all. Let an attorney educate you:

First, bottom line - no one is taking anything from these musicians. They never had any such right in the first place. They're claiming that any song inspired by theirs that sounds kinda similar is infringement. It's not and never has been. Regardless of whether a person or a machine does it.

Copyright is a time-limited (if ~150 years can be called that) set of economic rights (Euros say moral rights too, but that's stupid). Copyright is not "it's my work, I have absolute control of it". Uses outside those economic rights are not governed by copyright. News reporting, educational use, libraries - copyright cannot be used to prevent these things, they are explicitly allowed. Fair use (US) / fair dealing (UK) gives conditions where use does not infringe on copyrights, notably considering how much of the work is used and whether it's done for profit and whether the use is transformational or verbatim. Example: news reporting on a copyrighted work is explicitly allowed, even when news is for profit and verbatim, as long as a limtied excerpt is used that's less than the entire work. That's simplified and with slight inaccuracies, but close enough for a non-legal audience.

The trouble with the artists' argument is that they're essentially complaining about learning. "That there AI learned mah werk and made more kinda like it. That ain't fair!". Except humans do the same thing all the time; that's how creation works. If you have a blackbox that takes existing music in one end and outputs similar music on the other end, that box could be filled with a person or with an AI. The artists are saying "It's infringement when the black box contains AI but not when it contains a person". Perhaps that's their policy wish, but that's not what the law says.

Nothing happens in a vaccuum ex nihilo. Every artist combines influences from previous works when making anything new. There's even a legal test for this: derivative work. If a new work copies too much from an existing work, then it's termed a derivative work and triggers copyright protection on the original work. Derivative works are deemed to borrow too much.

The flip side is, you can borrow a lot from previous works (every work does) and as long as you don't cross that derivative work line, then you're ok. Your work is considered new enough that it doesn't infringe on previous works. Doesn't matter whether it was made by a person or an AI. That's not part of the legal test.

The main factor is: are these new AI-generated songs too similar to prior songs that they constitute a derivative work? From the AI songs that I've seen, any sane person would almost surely say no. They do sound similar and take inspiration from the original, but do not cross the line. Except the music industry is so screwed up that they consider even a 5-second samples of a beat or riff to be infringement. Like when a hip hop song makes a totally new song with new music and new lyrics that incorporates a few seconds of a heavily modified version of a well-known riff from another song - music industry says that's infringement. See Supreme Court case Acuff v Campbell Rose where 2 Live Crew sampled a few beats from Pretty Woman - not infringement. Compare to any other creative industry - movies, literature, painting, etc - and music industry's views are asinine. No one outside the music industry takes musicians' views on copyright seriously.

"Who owns it?" is the wrong question. The right question is: who should own it?

IP rights are premised on a bargain. Creators get protection for a limited time. In exchange, the public eventually gets full rights when it falls into the public domain. Yes public domain has been a mockery since at least Eldred, if not before. The principle stands though: at some point, rights expire.

Who deserves protection here? What incentives do we want to promote? Who invested time / money / effort to achieve the result, and would granting or denying rights change the amount of those investments? That's the answer to who should own the copyright, if anyone.

At first blush, Microsoft has the most invested. They created the AI, trained it, hosted it, and made it available. The guy doing the copliot chat invested very little, just a normal conversation. But if MS gets the whole copyright, then no one would use their system.

Probably the best approach is private agreements. Human author owns the copyright - they had the idea and invested some time on it, however little. With more time, they could have done the same work as the AI to build their idea. But MS doesn't have to let authors use their platform. So MS can draft a Terms of Use Agreement that says any copyrighted works developed with use of their system owe an X% royalty to MS on any earnings from that copyright work. If MS doesn't care then X can be zero. But let MS decide the terms, and users can decide whether to use the tool or do the work themselves.

I'm an IP policy attorney. I do this for a living.

Why do you need a poll? Just publish slashdot traffic stats.

But that would show how much readership has fallen off a cliff...

The alternatives are a drunken stepdad who throws up all over you at random intervals and hit on your friends (Windows), or a pervy uncle who lurks in bushes outside of bathrooms and places spy cameras all over the house (Android). No thanks.

Apple is the chain-smoking control freak mom. Yeah she's a bit too strict, and sometimes she lashes out at the neighbors. But at least she's not trying to ruin your life. It's the only sane choice.

Then you've got linux, the fussy hyperactive 5 year old nephew. He'll say yes to anything you wanna do. But he's got no focus and constantly bugs you with a million tangential questions ("you wanna run amarok? cool! did you check /proc/audio/SOME_OBSCURE_FLAG? what about /etc/xyzlib/wtf_is_this.config? are they compatible with amarok -F undocumented-gcc-flag? no? then it's tantrum time!!!"). Technically capable, but a huge pain to deal with.

And Microsoft fails utterly. It's not like W11 won't work. It's just that it gets in the way.

Let's test that theory...

• [performs totally mundane operation with no destructive side effects whatsoever]
• [gets assaulted with 20,000 "Are you sure?" dialogs slapping me in the face like flying dildos]

Yep, checks out.

Says the guy confusing terrible hack with good idea.

Should block demarcators be visible? Why yes - python has them. It's called whitespace, and it's perfectly visible. Different blocks start in different columns. Very easy to spot.

Should a language have two visible block demarcators? No, that's silly. Redundancy creates confusion. Brace and whitespace should not both be allowed to demarcate blocks. Likewise, spaces and tabs should not both be allowed. Python avoids this with -tt to prevent mixing spaces and tabs, though why it's not enabled by default is a facepalm mystery.

Should a language use different block demarcators for humans and machines (compiler / parser / interpreter)? Absolutely not! That's the worst of all worlds. Machine will interpret blocks one way, while humans could interpret them completely different. Talk about a recipe for disaster. Hey guess which languages do exactly this? Hint: not python.

The only knock against whitespace is silly programmers expecting to cut and paste code directly from whitespace-munching systems like the web. Get over that silly notion and the whole damn problem goes away.

It's math not maths. "Maths" is based on faulty grammar and only came into use around 1911. Sorry British people, you're wrong.

The castigation usually goes: "Mathematics is plural, so maths needs its -s." It's a logic based on a false (AmE) premise/(BrE often) premiss. Just because there's an -s at the end of mathematics doesn't mean it's plural. The suffix -s is homonymous.

Mathematics doesn't work with numbers because it's not a countable noun, it's a mass noun. That is, it does not take plural marking because it is not the kind of thing one can or does count. Similar examples (without the confusing -s) on the end are cinnamon and boredom. Note that you don't talk of putting *cinnamons in your food (unless you're making the point that they are different types of cinnamon--which is a different matter), nor does one suffer *boredoms if the boredom happened at different times. Cinnamon and boredom are treated as masses with undistinguishable (or at least not-worth-distinguishing), and therefore uncountable, parts. If we want to make such words countable, we have to use another noun to do so: two teaspoons of cinnamon, three episodes of boredom. Similarly, you can have three theories of mathematics or three mathematics classes, but not *three mathematics.

Now, some of you will say that Mathematics please me is what you'd say. This is the effect of the folk-belief that mathematics is plural; it has started to change how people use the word. We see the same kind of language-change due to misapprehension of the -s suffix in the short form maths. Math is the older form--the OED has examples back to 1847, but examples of maths only from 1911.

Faster / more efficient / higher quality always wins out, right? Look how Betamax beat VHS. Blu ray beat streaming. OS/2 beat Win95. Digital cameras beat cell phones. Oops.

Modern web is way past "good enough". 5ms latency instead of 500ms, who cares? Machine gets a little slow from browser load, put more RAM in the sucker. Much cheaper to throw hardware at the problem than develop a completely new "optimized" software stack full of bugs, security holes, unknown failure modes, lack of documentation, no support base, etc, etc.

This is a solution in search of a problem. Never going to displace the current "good enough" infrastructure. Why would anyone want a whole set of unknown problems? Only for very very niche markets like high-frequency trading*.

* illustrative example only. HFT doesn't use web browsers. Please route all other hypertechnical, smarty-pants objections to /dev/null where they will be promptly handled with the care they deserve.

Racism requires the belief that one race is better than others. Can an algorithm believe something?

False. Racism is not a belief. It's discrimination based on race. Discrimination measures outcomes, not beliefs.

Unequal outcomes for different races is prima facie evidence of discrimination. This can be rebutted if other causal factors explain the outcomes - for instance, maybe the algorithm really has trouble with blue bonnets, and blue bonnets are more often worn by some races than others in the sample set.

Once disparate treatment of different races is identified, the algorithm is immediately suspect and deemed racist. The burden then shifts to the algorithm provider to give a credible alternate explanation, if there is one. Until an alternate explanation has been tested and validated, the algorithm should be considered inappropriate and removed from service.

This is how allegations of racism work in the legal world. It applies to algorithms the same as any other actions.