zugi

> The CO has taken the (entirely correct) position that copyright law exists to incentivize human acts of creative expression.

The US Constitution grants Congress the power to grant copyrights "to promote the progress of Science and the useful Arts." Creative expression that is not useful therefore is not constitutionally copyrightable.

But that aside, I agree with the author that the Copyright Office refusing to further expand the already vastly over-expanded scope of copyrights is a rare positive ruling.

> rules for thee but not for me

Exactly, the same copyright rules should and do apply to both parties.

I'm glad copyright law didn't prevent the creation of LLMs. I'm also glad it doesn't prevent others from creating LLMs. I think that benefits us all.

Without government intervention, I think distillation will prevent these giant AI companies from recouping their investment. But that's not our problem.

Copying the exact weights? Yes.

But asking the trained model lots of questions, and creating a similar but different open weights model? No, that's fair use.

The current copyright rules cover both cases well and benefit everyone, I think. (I think copyrights are waaay too long, but that's another topic....) Whether trillion dollar companies can ever make a profit once open weights competitors distill them is, frankly, not our concern.

I agree on this point!

And also, training an open weights AI on the outputs of an AI trained using "fair use", is itself "fair use" that benefits everyone.

Like many I was a taker, almost never asking or answering questions, but simply finding answers that helped me.

Even so, at some point I made sure to answer a few questions, to get enough reputation to upvote and occasionally comment. I wanted at least to be able to reward / thank those who had posted answers that I found helpful. Even for 5 or 10 year-old answers, if an answer helped me I'd upvote it.

Now we're all "taking" via AI. We don't know who posted a helpful answer 10 years ago that the AI learned from in order to help us. That saddens me.

But in a way that's just the way of the world. We know to credit Pythagorus and Fourier and a few others whose names are attached to their methods. But for the most part we're all constantly standing on the shoulders of anonymous giants whose names are long lost.

> Software engineers who use LLMs should save a curated prose document describing the software, alongside the code that was generated from it.

That sounds great! Though software engineers are all about being lazy - typically in a good way. So I expect in practice they'll use LLMs to generate some code, iterate a few times until it works well enough for them, and then have the LLM generate "a curated prose document describing the software..."

Management surveys have shown that 90% of managers rate themselves as being in the top 10% of management.

I have no evidence but suspect most of the actual top 10% of managers are in the other 10%.

> Jevons paradox

Thanks! I've heard numerous examples of this. E.g. paint rollers, and later paint sprayers, made painting so much more efficient and cheaper that people wanted their houses and apartments painted more often, ultimately increasing house painter employment. But I didn't realize the phenomenon had a name.

> Why is this so overwhelmingly complex? Why would I spend time learning this ugly, convoluted and verbose UI?

My guess is that's the point - now you understand how San Francisco property developers feel.

Corollary: I would like the value of my house to rise, and I would like housing costs to fall to make housing more affordable.

> I was kinda shocked when he was let go, to be honest.

Uh, Mike Tomlin resigned / reitired from the Steelers after 18 seasons without ever having a losing season. He wasn't let go.

Yeah I've evidently been staying at the wrong hostels.

It's perfectly safe for YOU, the CUSTOMER.

It's just so powerful that, like, the WORLD can't handle it, man!

Downloading and redistributing companies' exact AI models would violate copyright. Querying multiple AI models to form your own model with a unique set of weights does not.

Exactly. AI companies distilled the whole internet into AI weights, now other AI companies are distilling their work to generate new AI weights. Nothing wrong with that, and the world gets open weights AI models.