Not to mention your phone camera, which uses AI to brighten up areas of the image, work out saturation, focus on faces, etc…
Those techs are a little older so now we call them “algorithms”, because nothing we are familiar with can be true AI. But they operate on many of the same principles as generative AI.
Working with patents, I’ve seen this several times. People trying to solve the same problem coming up with essentially the same idea within days, weeks or months of each other. In such cases, it’s the first inventor who gets the win, and the second place finisher can be (and has been) held liable for patent infringement, even if they never saw the first inventor’s work.
But parents are different than copyright. If I invent a machine but never clothed to patent it, just using it at home, and then a big company patents the same idea - they get the patent, right?
Yes, that’s correct. You have to apply for a patent, so it’s not quite the same thing, but you create the right to apply for and be granted a patent in the moment of invention. But again, it is a human who creates that right, even if they never exercise it.
If I prompt ChatGPT to think up ways to, I don’t know, harvest corn; and it successfully devises a simple machine that has never been created before and effectively harvests corn more efficiently than any existing method; can I patent that?
Not that the others don’t, but this strikes me as a notable question. If I draw a sketch – be it on paper or in Paint – and use it as an img2img prompt in an AI Art program, should I have copyright for the finished result? What if I’m using Photoshop or another image program that now incorporates AI and can draw in a sky or mountain range based on my blue swatch and brown triangles?
If I draw a figure in Stable Diffusion and use ControlNet (a Stable Diffusion plug-in) to pose it; personally choosing how its seated, how its hands are placed, what its holding, etc and then let SD draw in the actual image of a person on top, should that be protected? Or, to combine them, if I draw a sketch of a person, use that in ControlNet to refine the posing, then have SD do the render? We’re well beyond “You just pushed a button” by this point.
Random tangent, but that’s a really cool idea I’d never considered.
When I was a kid, I spent hours drawing stuff in MS Paint. I remember especially one time when I made a giant canvass, so big you had to scroll to see it all, and then drew a massive ocean scene.
It was, of course, all scribbles. But if I could have loaded that into another program and gotten a photorealistic, or even just a well drawn cartoon representation of what I’d drawn, that would have blown my mind.
Like I said, many years ago I asked this question after seeing a talk about using a genetic algorithm to produce a design of a new antenna. It came up with a design unlike anything a human had ever produced, which was apparently significantly better than anything that had come before. Had it been produced by a human, it would have undoubtedly been patentable, but under the law and jurisprudence, the computer could not be considered an inventor.
Did you read the link about patents and slaves? If a slave couldn’t create the right to apply for a patent, why would a computer be able to?
We really need to sort this out as a society, because it’s only going to get worse.
Is something like that really not patentable? I’d think that the person who ran the algorithm, then physically built and tested the antenna should be able to patent it. Can they really not?
The existence of a slave is an abortion of justice. It is a human being with no legal rights. The fact that they have no right to take out patents is part of that injustice.
An AI is not a person, it’s not an entity that deserves consideration as a person, any more than a camera or Adobe Photoshop does. And an AI doesn’t create anything. An AI is a tool used by humans to create things.
I’m not sure if this has ever been adjudicated for patents, but for copyrights, the US Copyright office has ruled against AI creators, and this has been supported by the courts, as discussed above. This doesn’t compel the courts to rule the same way for patents, but I’d be very surprised if they didn’t use this, as the closest precedent available.
Again, we could always change the law. Intellectual Property is entirely a creation of the law, and it has exactly the limits we decide to place on it. If we want this stuff to be protected, we can do that. We can also do the opposite.
For example, here in Canada, “methods of medical treatment” are not patentable. They can be patented in many other countries, but Canada, as a society, decided to go a different way. So we explicitly wrote that into our law.
SkyNet would be AGI with genuine volition, and bear about as much relation to modern generative AI algorithms as a child’s drawing of a stick figure bears to us. The law could and should treat an AGI very differently from Large Language Models.
And there’s a pdf of the USPTO’s current guidance here:
Key takeaways
– AI assisted inventions are not categorically unpatentable for improper inventorship
– Focus of inventorship analysis on human contributions, specifically - significant contribution
(Pannu factors)
The weaseling will start when trying to define a “significant contribution”
Significant contribution
• Each named inventor must contribute in some significant manner to the
invention. That is, each named inventor must satisfy the three Pannu factors:
– contribute in some significant manner to the conception or reduction to practice of the invention,
– make a contribution to the claimed invention that is not insignificant in quality, when that contribution is measured against the dimension of the full invention, and
– do more than merely explain to the real inventors well-known concepts and/or the current state of the art
There’s a lot more in that pdf if anyone is interested enough to read it.
ETA: Lots more stuff to read here, with some examples.
He’s not saying that his art is equal to all other art. He’s saying that his art requires at least as much effort as some other things that the courts have found could be copyrighted. IIRC, there was a court case that found that copyright applied to a nine-word sentence. This guy’s prompts are probably longer than that.
Maybe this is true right now. But eventually, AI will get to the point where it should be considered a person. How will we recognize that point when we get to it? Personally, I support a case-by-case sliding scale, in that, if a computer is “human” enough to be able to create art (even if only on request), then it’s “human” enough to be able to own a copyright.
I’ve created a few bizarre T-shirts using AI designs that I generated.
Are you people saying I can’t copyright my shirts and make fabulous amounts of money selling them online??!?
*the one of a giant leering sheep in outer space wearing a chef’s hat, cooking shish kabobs using an assortment of unconventional items is my favorite, along with another showing giant upright lizards attending to odd-looking but happy infants at a sort of medieval day care.
Right, and a defense is that the accused infringer did not have access to the purported infringed work. AMD had a license to use the Intel instruction set, but not their firmware. Those at AMD who developed their firmware did it in a clean room where they had no access to the Intel code. Something similar was done for IBM compatible mainframes.
If people could copyright AI generated material, the courts would have to decide whether the access of the AI to the same source material counts as access to the generated material. How about someone seeing something they’d like to copy and creates it by twiddling the input to the AI without seeing the input used for the original? Is that a clean room, given that the AI had the same training material?
Beats me.
