Thaler’s lawyers complain that if the Copyright Office’s reign of terror continues unchecked, it “will have irreversibly and negatively impacted AI development and use in the creative industry during critically important years.”
I for one welcome their doomsday scenario.
I’d love to read the exact reasoning behind this. To me it’s completely unclear who should get copyright in this instance. The guy who put in 3 sentences to execute it and generate something? Sam Altman for commissioning some AI? Or the scientists? The model weights itself as some sort of legal entity? The people who created the art that got mashed up to be the thing the AI consists of? Do they all get some share of the end result, like with music where someone writes the lyrics, someone the music and it’s sampled from some song who might also get licensed? How do we even tell with AI being somewhat of a black box?
I think the answer is and should be: no one. No human authorship means it is in the public domain.
Yes. In my opinion that’s the only sane option. And we still got a million of unanswered questions with that approach.
US patent and copyright law is pretty clear that it exists to “promote the Progress of Science and useful Arts.” The idea is to give people some legal protection to encourage people to share what they create.
“AI” is not a person. It can’t receive copyright protections. It’s essentially a word and image producing machine. No one gets copyrights for what it produces because no person is controlling what comes out. Sure, someone writes a prompt. But, they can’t, with any certainty—make changes that yield intentional changes in the output.
Contrast with a camera. The camera is a machine that produces an image. The photographer has control over the exposure, lighting, subject, scene, etc. They can use their creativity and ingenuity to produce specific effects.
Think of any machine that produces something. Does the operator have creative control over what comes out? You can’t copyright a shot of espresso, a potato pulled up from the dirt or a particular car rolling off the assembly line no matter how you prompt those machines.
You could use a backhoe, a machine gun, a supersoaker, a whisk, a rotary phone or an oven to make art but it matters what the intention is, what control you have and how transformative it is.
“AI” prompting simply doesn’t meet standard for creative works.
Now, you can take the output from that and create something. That, depending on how transformative your changes are, may be eligible for copyright protection. An image made into a poster with manually added text, image adjustments, edits, etc. could get copyright protection while the original image still does not.
Yes. I’m pretty sure that’s 100% correct with the person who creates a picture with generative AI. I think the far more interesting question (and most discussed one, apart from this court case) is, what happens to the transformed works. All the questions regarding the datasets, their licensing etc.
And copyright, IP and patents are really weird. I think extending copyright to 70 years after death isn’t exactly done to foster innovation. We get all the third parties, labels, publishers, industry, web platforms earn probably way more than they should and the artists themselves not enough.
And then with every example there’s more weirdness. You can definitely patent the process of making an espresso. Maybe copyright a recipe in some jurisdictions. Patent a potato, and the machine that digs it up.
With the camera, I have some additional weird rules. Once I snap a picture of a building, somehow the intellectual rights of the architect can get in the mix. Sometimes they don’t. I think some rules governing immaterial and intellectual goods are a bit complex. And they don’t always achieve what they’re trying to do.
If I remember correctly from a different post about this same thing, Thaler (the person suing to obtain copyright) developed the AI himself. I imagine that any sane copyright body globally would stand by the position that Thaler has the right to copyright/patent(/?) the AI itself, but that a generative machine learning program isn’t able to produce anything eligible for copyright due to the very nature of how they’re trained. (Thaler does in fact hold patent on the AI.)
Actually…I remembered something about what he named it, and in searching for more information I found that he was denied, and the decision was upheld by the Court of Appeal and then the Supreme Court following a suit by Thaler, to have his AI named on a patent filing in the UK. Original filing was 2019, and final judgment was 2023.
Seems the “esteemed” Dr Thaler is trying as many patent and copyright bodies as he’s able to in an effort to get his program recognised as eligible for legal protection on it’s output.
He’s named it DABUS (Device for the Autonomous Bootstrapping of Unified Sentience)… I may be jumping to conclusions…but he seems like a bit of a nutter…
He’s attempted patent and copyright filings in:
- Australia
- Europe (through the EU’s European Patent Office)
- UK
- USA
- New Zealand
- South Africa
- Switzerland
And he’s been denied in all instances…is what I’d like to say…but both South Africa and Australia allowed DABUS to be named as the inventor on patents. South Africa being an administrative decision, but with Australian courts handing down judgement that there’s no requirement for an inventor to be human…
He seems pretty convinced that his program and it’s associated machinery is actual, legitimate, bonafide AGI. Hard to find a description of the actual setup that isn’t directly from Thaler’s site and journal submissions, but it does sound like it’s unique…but it’s hard for me to figure out as a layman. There’s reference in several places to
extensive artificial neural systems that emulate the limbo-thalamo-cortical loop within the mammalian brain
electro-optical attention window scans the entire array of neural modules in search of so-called “hot buttons,”
And there’s several mentions of a “swarm of neural networks” but depending on what the source is this swarm is either entirely disconnected, or highly interconnected.
Regarding the “electro-optical attention window”…
The underlying difference between DABUS’s structure and other common forms of AI is that DABUS is not directed what to invent. When DABUS discovers a new concept chain, its internal networks appreciate the concept’s novelty.80 To communicate its appreciation of a novel concept chain, DABUS rings these “bells” to alert Dr. Thaler of the idea. DABUS can then convey its specific, novel idea to Dr. Thaler through images flashed upon a screen or through text in the form of pidgin language. Even though DABUS primarily functions without a directional course, Dr. Thaler can also confine its knowledge to be within specific conceptual spaces, such as medical information, so DABUS will concentrate its thinking within that conceptual space to solve a particular problem dealing with medical information.
Thank you very much for the long explanation and links.
I guess we’ll achieve legal clarity this way or the other. I mean doesn’t really matter what he thinks. Once there’s an agreement by most institutions/courts to rule the same way, it’ll achieve certainty.
And seems I was right and there’s more to the story. I was originally under the assumption this was about AI involvement in some invention process and the implications thereof. But seems it’s a bit different and this is about whether AI can act as some legal entity. Guess that question has very simple and straightforward answers 😃
originally under the assumption this was about AI involvement in some invention process and the implications thereof.
I mean…this assumption is actually correct, at least from my perspective and understanding. That Dr Thaler’s patented program/system is purportedly unique enough to have it’s own patent is kinda immaterial?
I got to read the reasoning behind the ruling. But what courts will usually do is not bother with details, once an application is wrong on technicalities. Which is probably the case here. We have a clear legal definition of what a legal entity is. And if someone puts in some nonsense, like AI being some kind of entity. Or like in the Simpsons episode when it turns out the nuclear power plant is actually not owned by Mr Burns, but by a parrot… Most judges will immediately dismiss the case and not bother with the rest of it because they have other stuff to do… I’ll try to look up if it’s like that once I have some time to spare.
It changes things, though. It’ll be a slightly different question we get legal certainty on.
Cool, ping me when you’re done reading up, I’m interested in hearing/seeing what you find.
It needs to be no one for IP and copyright law to continue to function.
If you take music as an example, one could use machine learning tools to make every possible combination of notes, then try to use IP law to go after every person who made a song after them.
Yes. We habe copyright, IP… But it’s an entire can of worms. We have personal rights as well. All the nuance with what’s transformative to what degree. People can’t just commit crimes and then blame responsibility on their AI for doing it… And then current copyright and IP isn’t even good or clever or works out particularly well.
Should you get a copyright for uttering the words make an art to a machine? No.
