News for hackers. Stuff that matters.
Re:Let's Not Flip Sides on IP Maximalism Because of AI by rjmunro on Thursday November 30, 2023 @08:16AM (#38473089)
I'm not sure it's always true to say "100 GB of weights don't resemble any work".

If I train a model on a famous poem or something, and it turns out if you ask the model to quote the poem verbatim, it can, then the model contains the poem. Have I not "copied" the poem into the model?

Re:Let's Not Flip Sides on IP Maximalism Because of AI by visarga on Thursday November 30, 2023 @08:20AM (#38473123)
You can simply copy the poem in much less space if that's what you want, but incidental replication is not the same with copying.

When you train a model, you use 1T...20T tokens, with deduplication to reduce direct memorization. Then the model is a superposition of gradients from trillions of tokens, no longer just a copy of a poem, it could also write a commentary about it or compose new ones.

GPT4 summarizes this in a haiku:

"Superposition bright,

Gradients blend in the night,

New poems take flight."

Re:Let's Not Flip Sides on IP Maximalism Because of AI by philipwhiuk on Thursday November 30, 2023 @08:40AM (#38473372)
> GPT4 summarizes this in a haiku:

This has never added any value to a comment. Stop doing it.

Re:Let's Not Flip Sides on IP Maximalism Because of AI by sakjur on Thursday November 30, 2023 @09:43AM (#38474099)
It’s also not a haiku afaict.

Doesn’t superposition have five syllables?

Re:Let's Not Flip Sides on IP Maximalism Because of AI by topaz0 on Thursday November 30, 2023 @09:48AM (#38474167)
yep
Re:Let's Not Flip Sides on IP Maximalism Because of AI by visarga on Friday December 01, 2023 @01:50AM (#38483936)
better than I would have been able to make, anyway
Re:Let's Not Flip Sides on IP Maximalism Because of AI by topaz0 on Friday December 01, 2023 @05:44PM (#38493583)
You're selling yourself short. It's not hard to count syllables.
Re:Let's Not Flip Sides on IP Maximalism Because of AI by topaz0 on Thursday November 30, 2023 @09:47AM (#38474158)
Also note, it's not a haiku
Re:Let's Not Flip Sides on IP Maximalism Because of AI by cle on Thursday November 30, 2023 @08:41AM (#38473391)
My understanding is that the law ultimately cares about the market impact on the copyright holder, and balancing that with the interests of the public. I don't think the courts will be swayed by arguments based on technicalities like this.
Re:Let's Not Flip Sides on IP Maximalism Because of AI by fortyseven on Thursday November 30, 2023 @09:05AM (#38473646)
The law doesn't care about the technicalities...? It damned well better.
Re:Let's Not Flip Sides on IP Maximalism Because of AI by cle on Thursday November 30, 2023 @09:18AM (#38473798)
"Technicality" here means an argument that follows a strict/formal interpretation to arrive at a conclusion that is inconsistent with the principles and objectives of the law.
Re:Let's Not Flip Sides on IP Maximalism Because of AI by danaris on Thursday November 30, 2023 @09:25AM (#38473869)
The law doesn't care about technology technicalities.

It cares about legal technicalities.

The law doesn't care, for instance, that the MP3 file of Beyoncé's latest hit is not the same bits in the same order as the original master file created when she recorded it. That holds true no matter how many times you re-encode it into different formats.

As things stand, there has been no definitive ruling on the matter of ML model training and copyright; I don't know how that will end up shaking out, and honestly, I'm not yet sure how I think things should shake out in any detail. I think that people whose works are being used to fuel massive profit-generating engines for the very wealthy should have some right to give or revoke consent, but I'm not at all sure offhand if there's a realistic way to do that without serious unintended consequences.

But I am sure that arguing that the precise technical nature of how the transformations are performed means that nothing is actually being "copied" should not, and is unlikely to, hold much legal water.