Mahlzeit

I understand. The idea would be to hold AI makers liable for contributory infringement, reminiscent of the Betamax case.

I don't think that would work in court. The argument is much weaker here than in the Betamax case, and even then it didn't convince. But yes, it's prudent to get the explicit permission, just in case of a case.

I think a big picture view makes the problem clearer.

Licensing material means that you must pay the owner of some intellectual property. If we expand copyright to require licensing for AI training, then that means that the owners can demand more money for no additional work.

Where does the wealth come from that flows to the owners? It comes from the people who work. There is nowhere else it could possibly come from.

That has some implications.

Research and development progress slower because, not only do we have to work on improving things, but also to pay off property owners who contribute nothing. If you zoom in from the big picture view, you find that this is where small devs and open source suffer. They have to pay or create their own, new datasets; extra work for no extra benefit.

It also means that inequality increases. The extra cash flow means that more income goes to certain property owners.

That shouldn't be an issue. If you look at an unauthorized image copy, you're not usually on the hook (unless you are intentionally pirating). It's unlikely that they needed to get explicit "consent" (ie license the images) in the first place.

It's possible to run small AIs on gaming PCs. For Stable Diffusion and small LLMs (7, maybe 13B), a GPU with 4GB (or even 2GB?) VRAM is sufficient. A high-end gaming PC can also be used to modify them (ie make LoRas, etc.). Cloud computing is quite affordable, too.

Stable Diffusion, which had such an impact, reportedly cost only 600k USD to train. It should be possible to make a new one for a fraction of that today. Training MPT-7B cost MosaicML reportedly 200k USD. Far from hobbyist money, but not big business, either.

The models are deliberately engineered to create "good" images, just like cameras get autofocus, anti-shake and stuff. There are many tools that will auto-prettify people, not so many for the reverse.

There are enough imperfect images around for the model to know what that looks like.

Tja, dann muss man sich wohl auf neue Flüchtlingswellen einstellen. Das kommt, wenn Putin regelt, so wie Syrien 2015 und nicht erst mit dem letzten Ukraineüberfall.

Aber vielleicht gibt das jetzt Aufwind für den Klimaschutz in Staaten wie Ungarn. Orban wird sicher viele Alleen anlegen wollen. Viele Bäume. Bindet dann CO2. Und die Russen können schön im Schatten marschieren.

That ought to satisfy all those who wanted "consent" for training data.

It just seems that Google should have been able to move faster. Yes, they did publish a lot of important stuff, but seeing the splash that came from Stability and OpenAI, they seem to have done so little with it. What their researchers published was important but I can't help thinking, that a public university would have disseminated such research more openly and widely. Well, I may be wrong. I don't have inside knowledge.

For the fine-tuning stage at the end, where you turn it into a chatbot, you need specific training data (eg OpenOrca). People have used ChatGPT to generate such data. Come to think of it, if you use Mechanical Turk, then you almost certainly include text from ChatGPT.

Quite possible. Whatever the case, they apparently saw no pressure to innovate. It implies that tech development is being slowed down by the Big Tech monopolies.

I could download my file and be done with it.

That's true, but that's kinda delivering a physical copy via the net, and you pay the storage medium. I understood OP as talking specifically about online "property".

Took em long enough.

I wonder if they used ChatGPT to create any of the training data.