Supreme Court Rejects Computer Scientist's Lawsuit Over AI-Generated Inventions (reuters.com) 69
The U.S. Supreme Court on Monday declined to hear a challenge by computer scientist Stephen Thaler to the U.S. Patent and Trademark Office's refusal to issue patents for inventions his AI system created. From a report: The justices turned away Thaler's appeal of a lower court's ruling that patents can be issued only to human inventors and that his AI system could not be considered the legal creator of two inventions that he has said it generated. Thaler founded Imagination Engines Inc, an advanced artificial neural network technology company based in Saint Charles, Missouri. According to Thaler, his DABUS system, short for Device for the Autonomous Bootstrapping of Unified Sentience, created unique prototypes for a beverage holder and emergency light beacon entirely on its own.
The U.S. Patent and Trademark Office and a federal judge in Virginia rejected his patent applications for the inventions on the grounds that DABUS is not a person. The patent-focused U.S. Court of Appeals for the Federal Circuit upheld those decisions last year and said U.S. patent law unambiguously requires inventors to be human beings. Thaler told the Supreme Court that AI is being used to innovate in fields ranging from medicine to energy, and that rejecting AI-generated patents "curtails our patent system's ability - and thwarts Congress's intent - to optimally stimulate innovation and technological progress."
The U.S. Patent and Trademark Office and a federal judge in Virginia rejected his patent applications for the inventions on the grounds that DABUS is not a person. The patent-focused U.S. Court of Appeals for the Federal Circuit upheld those decisions last year and said U.S. patent law unambiguously requires inventors to be human beings. Thaler told the Supreme Court that AI is being used to innovate in fields ranging from medicine to energy, and that rejecting AI-generated patents "curtails our patent system's ability - and thwarts Congress's intent - to optimally stimulate innovation and technological progress."
Not wrong... (Score:2)
AI should not qualify as an inventor because there is no driving creative consciousness in the platforms. It is just a very fancy mechanical tool with operation complicated enough to seem like magic to people.
But they've also gone too far in blocking creations in which that tool has been used. I'm not entirely convinced that IP as it sits is a thing that should happen at all but within the current framework using AI should be treated no differently than having used CAD and simulation tools or for copyright
Just don't say that AI did the inventing! (Score:5, Insightful)
Just don't say that AI did the inventing! No one would know.
Re: (Score:2)
Re:Just don't say that AI did the inventing! (Score:5, Insightful)
Re: (Score:2)
Yup, he is playing games with public systems for kicks and they did what they did the right thing tossing his ass. He's lucky they didn't penalize him for a frivolous suit.
Re: (Score:1)
Technically this but the reason they want AI to count is for the mega patent troll, because it would become obvious if things of not particular high or natural conclusion of patents based on the work their company does to try and patent practically everything. They literally want to generate an AI that spits out things they can patent and just have that count.
That's why we can't do this, and what if tons of people build different AIs, ask for it to generate a solution, and it's the same solution, are you go
Re: (Score:3)
"That's why we can't do this, and what if tons of people build different AIs, ask for it to generate a solution, and it's the same solution, are you going to argue you can't use it because a similar AI generated that first? No, dark path, do not allow."
I don't see any difference between that and humans chomping at the bit to patent the next iterative improvement. People commonly come up with the same solutions and first to file gets the patent.
If you and I are trying to solve a problem and have similar back
Re: (Score:1)
someone has to feed them enough parameters and train them to solve a problem... they aren't self-driving.
Self driving / self driving cars exist because they were fed data on the world. So yes, they can easily be 'self driving' in that context.
You design a program to consume the world (Say, everything on the internet) and "invent" everything, since you fed it the world it now owns the world?
It's just lazyness, oh I made something that invents things for me and now you need to give me money. That's not a benefit to mankind. The other problem you tried to snuff is the extreme difference. If you're an AI you didn'
Re: (Score:2)
"Self driving / self driving cars exist because they were fed data on the world."
First they don't really exist and second they are a solution to a complicated but discrete mechanical task.
"You design a program to consume the world (Say, everything on the internet) and "invent" everything"
That is something you can easily say but that nobody can easily do. There is actually a huge amount of effort that goes into not only deciding what to 'feed' these models but into formatting the information in a way they ca
Re: (Score:1)
So you don't think people looking to make a profit by acquiring patents any way they can won't try to leverage AI to acquire more patents for the sake of litigation and licensing those patents to make a profit? Because people do that right now, we call those companies patent trolls, and I doubt they would find it unethical to use AI to generate as many valid patents very quickly as possible if the law let them.
I might not respond because I very much feel like I'm talking to someone who has a vested interest
Re: (Score:2)
Actually we don't call companies that try to acquire patents through invention patent trolls. Why is it a problem for someone to use a tool to generate valid patents more quickly? The faster they are created the faster the covered technologies revert to the public domain and the sooner they can be taken advantage of in industry. I'm sure they wouldn't find that unethical, I certainly wouldn't. Inventors who simply plan to license the technology to others rather than attempt to reinvent the manufacturing wh
Re: (Score:1)
There's one really big misconception in the response.
I think there has the potential for a lot for AI to offer to humanity, and AI systems to be patented and available for people. The one specific part here, where you say,
Are you just a communist who is opposed to people profiting from adding value to the world?
One, trying to imply I'm a communist would be like if I said "Are you an alt right, are you into dictatorship? Are you an alt left trying to cancel everything?" Like this is just name calling, the discussion is going fine we don't need to try to assign and assault someone with their potent
Re: (Score:2)
"And before it gets thrown at me, a vested interested is a business decision, not a political agenda. We're just talking business."
You were very much implying there'd be something evil about having made such a business decision. I don't think it is entirely unfair to query into why. Getting past the notion that you've somehow taken advantage of someone if you made a profit transacting with them is surprisingly tough and pervasive. Getting past the notion someone else has taken advantage when they do so is e
Re: (Score:1)
Re: (Score:1)
Yeah, that's possible. I would imagine though that he could still do that by being the author of the AI / tool that assisted in developing the product and get the rights. I think the big risk they want to prevent is someone developing an AI for the purpose of taking existing patents, feeding that to it plus the patent laws and having it churn out 500,000 patents a month that are technically 'valid' so you can basically patent everything between the existing patents.
STFU 101 (Score:3)
> AI should not qualify as an inventor because there is no driving creative consciousness in the platforms. It is just a very fancy mechanical tool with operation complicated enough to seem like magic to people.
Maybe the human brain is the same way, we don't really know. We cannot even define "creative consciousness" in an objective way. I know some really stupid humans who appear to run on social reflexes and habits instead of any coherent logic, and they have regular jobs and vote (cringe).
But here's s
Re: (Score:3)
"Maybe the human brain is the same way, we don't really know."
Maybe my fairy godmother is watching over me, we don't really know. The default is the negative. But we do know human consciousness doesn't work by the mechanisms we've implemented in AI systems and we do not currently have the understanding that portions of the brain we've loosely modeled them after do either. Many believe the consciousness itself while obviously impacted by many of the systems we have well modeled must exist within other spaces
Re: (Score:1)
> The default is the negative.
No, the default is "unknown".
Re: AI is a tool (Score:2)
Because that sort of AI that explores and takes initiative without human guidance is only a short step from where generative AI is today, and will be here very soon if not already here.
Re: (Score:2)
If I select an option on my roomba to vacuum my place once a week, then go on with my life, am I the housekeeper?
There comes a point when you have to say the automated system is the agent of the task.
The AI with some "seeming" randomness (due to complexity in general, and inscrutable associative reasoning over a very large knowledge base) or "
Re: (Score:2)
Driving is a mechanical process. It is just a complicated enough process we don't know how to encode it. The agency is in choosing the when, where, and why to go. But it goes beyond that because you might have another machine which incorporates the need for travel into its task and can summon the car to drive it. Perhaps a shopping system. The agency will then come from the person who decides what foods, stores, and preferences they have. You'll simply have removed agency from part of the problem through au
Re: (Score:2)
I would say that you can have an AI run forward-looking simulations, "playing" a self or "ego" symbol (that's what Tesla FSD calls the car in its software) against various external things in hypothetical scenarios, judging likely outcomes, selecting a preferred action which is most likely to advance some higher level g
Re: (Score:2)
"The one critique I would have is that it seems you regard the qualia of consciousness (the feeling of self awareness, awareness of flow of time etc) as necessary to agent-like decision making."
On the contrary, I view it as necessary to be an agent or ego, as opposed to agent-like and attempting to substitute for an ego. Agent-like behavior is not only possible but trivial, any branching decision tree does that much. But some thinking and conscious being populates the tree, chooses its purpose, judges the o
Re: (Score:2)
I'm interested in us building such a thing for the very reason that successfully building it would help us explore which parts of the thinking process (the general-purpose think-act cycle) are NOT dependent on qualia stuff that we have no theoretical insight into.
My hypothesis is that MOST (if not all) of real AGI behaviour (human-like albeit without direct body-experiences) can be implemented without appeal to a conscious presence being inside the machine pulling the ultimate
Re: (Score:2)
"My hypothesis is that MOST (if not all) of real AGI behaviour (human-like albeit without direct body-experiences) can be implemented without appeal to a conscious presence being inside the machine pulling the ultimate decision levers."
I agree but think about what you are building... about what happens when it encounters a bug or novel situation, misclassified something, etc. People put a lot of work into such systems for financial trading and have spectacular gains only to have catastrophic losses when the
Re: (Score:2)
1. Integrating emotion-model to prioritize attention, recall, and inference direction
Yes. I believe that an emotion analog/model is required for a true AGI.
Simply put the evolving (sensed and neural-net represented) world is to complex to contemplate and act on all of it (or even all of what's going on locally) at once, no matter how fast/parallel your inference.
Therefore you need attention management.
There needs to be a Reaso
Re: (Score:2)
That is a very solid distillation of the points I've made and very insightful and solid reasoning on some of the components and how they might relate and be implemented.
Do that and you have an ego.
"Of course, allowing a perceiving, reasoning, learning agent to emotion-tag new situations it attends to on its own (recursive) initiative, is indeed a risky proposition. There needs to be at least the guard-rails of had-coded albeit fairly general "moral" rules that the AI should follow in its actions and communi
Re: (Score:2)
Re: Not wrong... (Score:1)
Agree. It is a tool. People are attributing things to it that, it simply isn't. That a computer scientist proposed this is kinda embarassing.
No mod points, post instead.
Cross the bridge when you get to it (Score:2)
I see no reason why a real AI, if there were such a thing, should be excluded. But that is definitely a bridge to cross when we get to it, and this particular question will probably the least of our problems when that happens.
The stuff we have at the moment is at best a clever and efficient way of chewing the intellectual cud of humans. Not AI.
Re: (Score:2)
Re: (Score:2)
AI not a person ... (Score:2)
I see no reason why a real AI, if there were such a thing, should be excluded
Well the law says the inventor must be a person.
Re: (Score:3, Informative)
First off, what is a patent if it's not an invention? It's in fact an exception to the fundamental rule of capitalism that there must be competition to benefit so
Oh (Score:3)
So I can't just make a company that spits out thousands of patents and then litigate when someone else uses something somewhat like it?
Re: (Score:2)
You absolutely can. You just have to take credit for the invention when you use a tool to help you develop it.
Re: (Score:2)
Also, good luck getting the AI's name on the court filings if the patent were issued in the AI's name.
Re: (Score:2)
The AI doesn't have rights, so even if the patent office recognized it as the owner, it doesn't have legal standing to litigate for damages.
Unified sentience (Score:5, Interesting)
Just ugh.
Oh god.
Another futurist that thinks that the world is just around the corner from a sci-fi novel, and he’s got fantasies about being the next Hiro Protagonist. Down here in the real world, human laws still apply, the world is MUCH more conservative than he thinks, and it takes 75 years to develop a halfway-usable internet enabled watch, let alone the nano-tech that futurist CS majors think is imminent.
I actually think that we’re at the very starting edge of the leap forward that Drexler predicts. But we’re not going to be creating a self-improving AI computer that ascends to godhood within 24 hours and pulls all forward to the next level of existence in an instant. If that was possible, biogical systems would have done it. Each step is going to require massive amounts of time, energy, money, effort, and technical advances. Thousands of years.
Re: (Score:1)
this. AI = Artificial Incompetence.
and will be so until we painfully make progress towards real AI. by then hopefully the cringeworthy hype will have died.
Re: (Score:1)
Re: (Score:3)
Smart people are actually more susceptible to believing ridiculous fallacies outside their area of expertise than average people. "AI" are currently overfitted, non-reproducible models. We can't even do much meaningful research until the reproduce-ability issue is addressed.
I didn't read TFA, nanotech is a pipe dream. A 3-d printing revolution is much more likely.
Re: (Score:2)
If that was possible, biogical systems would have done it.
Biological systems only do things that aid in propagating genes. If accelerating super intelligence doesn't do that, then it won't be selected for. Artificial systems do not necessarily work the same way.
Re: (Score:2)
Not saying it’s impossible, and maybe I lack vis
Re: (Score:2)
For the past few billion years, biological intelligence was absolutely correlated with survival and reproduction.
Not generally it wasn't. Highly intelligent species are very much the exception and not the rule.
Oh, wait, there is this ONE obscure species that with brains that leapt forward and their population exploded.
Yes, and there's no evidence that our intelligence is continuing to explode, because that would no longer be beneficial to reproduction. If anything I've heard there's some evidence we're starting to get dumber.
I’m extremely dubious about true artificial intelligence anytime soon.
I don't believe any predictions about strong AGI, whether they be "right around the corner" or "never" or somewhere in between. I think nobody knows what they're talking about.
ChatGPT is basically distilled internet with a bit of randomization thrown in.
Perhaps an oversimplifica
Patents are monopoly capitalism (Score:1)
This points out the absurdity of patents.
All inventions are merely advances on the "state of the art". They all build on the work of others... from basic science to implementation technology.
AI patents are merely the next step. By automating the advance on the state of the art AI exposes the absurdity of patents.
Re: (Score:3)
Its not a perfect system, but its the best we've come up with.
It costs money to develop things. Building the things after that is cheap. If we let anyone just build a doodad after someone else develops it then there's no incentive to invent things since the inventor will go out of business while the copiers rake in money (because they don't have any development debt to repay so they can price their copies much cheaper).
All in all too patents really aren't that bad - they last approximately 18 years here i
Re: (Score:2, Interesting)
It costs money to develop things. Building the things after that is cheap. If we let anyone just build a doodad after someone else develops it then there's no incentive to invent things since the inventor will go out of business while the copiers rake in money (because they don't have any development debt to repay so they can price their copies much cheaper).
Patents are actively harmful to society. It isn't novel new elements that are the product of R&D. It's the long boring hard work invested into product design and preparation of production lines that gets product to market. Patents are simply a means of weaponizing the states legal regimes against your competition.
All in all too patents really aren't that bad - they last approximately 18 years here in the US which really isn't that long to own exclusivity.
The patent system is a massive drag on the economy.
AI based patents trivial (Score:2)
Re: (Score:2)
I think many patents could be classified as trivial... but they get patented anyway and are used to threaten competition. (i.e. "on a computer", "round corners")
So it did it all by itself? (Score:2)
CD/DVD/Blue-ray tray ... (Score:2)
So he was using it to check the verbiage of a legal document or some other task, and it suddenly got bored and produced the design of a cupholder?
I think the AI just imagined an alternative use for a CD/DVD/Blue-ray tray. The AI was trying to identify an image of a drive with the tray extended. Cupholder was one of the rejected AI computer vision hypothesis. A secondary AI process reviews failed hypothesis for patentability and found the hypothesis credible enough for a patent.
Unpatentable? (Score:2)
Does an AI generated invention become prior art and unpatentable?
Re: (Score:2)
Absolutely it does, so long as it is published in some way (the failed patent filing counts!)
Re: (Score:2)
Not sure that it does, because I think "prior art" has to also be developed by humans. Interesting question though.
I think if buddy refiled and claimed himself as the inventor he'd get his patent.
Good for Ghostwriter (Score:1)
If a human can't take credit for an AI invention, then humans shouldn't be liable for AI creations. So ghostwriter shouldn't have a problem with being sued by the music industry. If the argument is that the AI was trained on copyrighted works, take the AI to court and collect damages from it.
J/K. I'm sure the elites will try to have it both ways.
You can already do this. (Score:2)
Re: (Score:3)
>>Registering patents to a corporation you own
Not in the US. Only natural persons may be listed as inventors on a patent filing. You may ASSIGN patents to a corporation, which is just about legal ownership.
Re: (Score:2)
What is the Motivation for this? (Score:2)
Re: (Score:2)
He doesn't care about the patents as inventions: he only cares that he could be the first to have an A.I. listed as a patent author. It's 100% marketing.
These trolling "scientists".... (Score:1)