This is all moot because it's false that if you use AI you cannot gain copyright on the work. I literally hold a copyright registered with the US copyright office (I registered with them, despite not being a US citizen, because they're precedent-setting), fully disclosing the use of AI in the process, and directly speaking about the matter with a copyright examiner.

What the standard is is that raw, unfiltered outputs are not copyrightable (and even then, there's wiggle room; the argument put forward by the copyright office was "based on their current understanding" (as of several years ago) and presumed very minimal control by the user over the outputs (which was actually already obsolete by the time they put it out, but that's nitpicking)). But the human creative action you take with those outputs can cause the resultant product to be copyrightable. This can be things like sample mixing, mastering, adding vocals, etc, but can also even just be the selection process itself. Yes, assemblage, selection, curation, etc are all potentially copyrightable acts.

In my case, while the samples were (overwhelmingly) AI, the whole project was weeks of work, involving hundreds of generations, track splitting and reassembling, things down to the level of editing individual phonemes (or for example in one part, stripping the vocals and instruments from a sample down to just the residual "other" channel, throwing it into reverse, and then piling on progressively more reverb and volume up to a sudden cutoff to create a tension-point transition). No issues whatoever registering the copyright on it. I'm sure you could do far less and still have a copyrightable work.

I think people need to keep in mind that marketability (aka enjoyability), talent, and effort are three independent axes. You can be talented at one or more aspects of music production, a genuine skill (play a guitar, sing, etc), but put forth either high or low effort, and make good or terrible music, all independently. You can put forth a lot of effort, but you may or may not be talented, and it may or may not be good. You may make a great enjoyable marketable track, but may or may not have notable talents, and may or may not have put in a lot of effort.

In general: I judge how much I want to listen to it by its enjoyability; I attribute how much the author deserves credit / how much it's "art" by how much non-rote effort they put in; and I value the artist's skill based on their demonstrable talent. All of this independent of what specific tools were used. That said, knowledge of certain facts may colour one's enjoyability of a track. If you hate AI and know / suspect it's AI, that'll strongly colour your enjoyability of it. Or if you know the artist is a bad person or whatnot, that again can colour your enjoyability of a track that you'd otherwise enjoy in a vacuum.

One thing I've been thinking recently is being in the AI trade but only with high seniority, not common stock. So if there is a "pop", but the underlying business remains sustainable, the high seniority investors end up with the assets and make a mint when the market returns to balance, whereas the stockholders are the ones who suffer. This could mean a mix of convertible bonds / convertible ETFs, preferred shares, senior corporate debt, physical infrastructure (power suppliers, REITs, etc), BDCs and private credit vehicles, etc. Or possibly even common stock in some of the diversified giants (Google, Microsoft, etc), who - while they'd take quite a temporary hit in a crash - would survive and then buy up all the distressed players.

Oh, shoot, I forgot my "Claude-isms" ;)

I mean, Leopold was literally fired from OpenAI after one year.

(Though I would hardly call Altman "a genius". He's a rather sociopathic VC guy who happened to be in the right place at the right time)

I'm holding out for so many things.

New fabs are finally being built but they take like 3-5 years, so that's not going to solve anything any time soon :P

Situational Awareness is a "hedge" fund that did the opposite of hedging - they made a bunch of leveraged bets that were all linked to each other in typical market movements. It was headed by Leopold Aschenbrenner, a guy just a couple years out of college, whose employment career had been very brief stints at the FTX crypto exchange firm (until it collapsed) and at OpenAI (until he was fired a year later over an alleged information leak). No financial management experience whatsoever. But he was into Effective Altruism, writing AI whitepapers, all that sort of stuff that Silicon Valley tech bros like, and so when he started a hedge fund, $45B was quickly pumped into it.

Nah, this isn't a sort of bubble-popping event. It's a tangential player in the stock market, not some key participant in operations or loans in the AI ecosystem. My main concern: inference IS profitable. Very. But the heavy leverage of the industry is also very real, and can very much still spread like a contagion. And a likely trigger for that is inflation, triggered by trade wars and the ongoing Hormuz and Ukraine conflicts (or worse, new ones added to the list!). One, inflation directly increases their costs, but two, it causes central banks to raise rates. This starves companies of capital (both loans and equity), incl. to refinance existing liabilities, while also hindering income (e.g. new orders get put off or cancelled). It can easily flip a leveraged company into insolvency, and then that can ripple if there's nothing to stop it.

An investment opportunity can be brilliant, a massive world-changing field with huge margins, but still be a terrible investment. One, because the companies you invest in need to survive continuously and not get blipped out by an adverse market event until they have more ability to withstand them. And two, also, because there's a problem where investors don't merely value the whole market as if it has high odds of success, but value the specific player they're invested in with high confidence as if it will dominate said market. Which, obviously, all players combined cannot do.

Certainly! Your point raises a crucial, multifaceted issue that merits deeper examination. It's not just about detection accuracy, it's about the evolving interplay between human intention and algorithmic assistance — a dynamic that resists simple binary classification.

When we delve into the mechanics of modern AI tools, it becomes evident that the boundary between "correction" and "generation" is increasingly porous. Consider the following nuances:

Gradual authorship spectra: Text exists on a continuum — from fully human-crafted prose, to human writing with AI-suggested punctuation, to passages where the AI has restructured entire paragraphs. Detection tools, however, often operate on categorical assumptions that fail to capture this complexity.

The feedback loop phenomenon: Furthermore, as writers internalize AI-suggested patterns — Oxford commas, em dashes, and certain syntactic rhythms — their "human" baseline shifts. This creates a fascinating, almost paradoxical scenario: the more one learns from AI feedback, the more one's unassisted writing may trigger detection algorithms.

Contextual markers vs. stylistic choices: Notably, certain signals — like the increased use of "delve," "testament," or negative parallelism — may reflect genuine stylistic evolution rather than automated generation. It's a testament to how AI influences language culture broadly, not just individual documents.

In conclusion, the challenge you identify is not merely technical but philosophical — forcing us to reconsider what we mean by "authentic" voice in an era of human-AI collaboration. The question, ultimately, is not whether a tool assisted in the writing, but whether the ideas, arguments, and intellectual labor remain genuinely human.

I hope proves helpful! Please let me know if you would like to delve deeper into any specific aspect of these findings, or if you would like me to explore the broader implications for content strategy.

"do a little research" - you do realize that your post would have had just the same educational value but not made you come across as a jerk if you had just left off that first sentence?

I think the obvious counter to an influx in bug reports / pull requests is to have the first layer of review of bug reports / pull requests to be AI as well.

Final remains humans.

I dunno, if that's how people vote, that's how it should be.

Though there is a big question of how they plan to run a vote with multiple options. I hope it's a Condorcet system to avoid the risk of watering down one side with similar competing options...

Oh, but surely they'll win, they indicted [strike]a ham sandwich[/strike] this man! And the Trump administration has such a spectacular record with high-profile prosecutions thusfar...

That is not what they found. What they found is that if you spend 60 minutes doing the homework with AI, you learn approximately as much as someone who spends 60 minutes doing it without AI. If you spend 20 minutes doing it with AI, you learn as much as someone who spends 20 minutes doing it without AI. Etc. The only real difference is that the AI people get way better homework scores, and as a result, tend to stop spending as much time, and thus tend to learn less.

That's what this preprint study found.

100% this. Even a short 5-minute quiz with just a couple questions at the start of each day will do the trick. But you simply can't rely on homework scores to be a proxy of learning or time spent studying anymore. Brief daily quizzes on the previous day's material will force them to.

Also, It's not like students have the excuse of, "I couldn't study, I didn't have anyone to explain X to me" anymore. The very AI that they might otherwise use to cheat on their homework can also explain to them the things they missed in last year's algebra class that are blocking them from learning this year's calculus, or whatnot. Congrats - you now have an infinitely patient personal tutor who will spend hours dumbing things down for you until you get it. Now you have no excuse - learn the material. It becomes all about actually putting the time in.

Yeah. There's been widespread misrepresentation of this (preprint!) in the media. What they actually found: test scores correspond to how much time you spend doing homework, regardless of whether you use AI or not. Which should not be a surprising result! Of course, the AI people, esp. those who had been using AI for more than a few months, stopped spending much time on their homework, and then their test scores fell to match the non-AI users who didn't spend much time in their homework either.

The TL/DR is, in today's age, you cannot use test scores as a proxy of how much time a user is studying or knows the material.

There are of course some caveats, beyond the fact that it's just a preprint - notably that there's nothing to separate why the user started using AI to do their homework, so there's potential confounders in there. For example, if a student is in a poor family, maybe they had to get a job to help pay the bills. But working a job itself can lower test scores - for example, having to work late the night before an exam. So there are some confounder problems with the study. That said, I'm sure the general conclusion is correct: time spent doing the thing = amount of learning the thing (relatively independent of what tools you're using in the process).