The US Patent and Trademark Office Banned Staff From Using Generative AI 22
An anonymous reader shares a report: The US Patent and Trademark Office banned the use of generative artificial intelligence for any purpose last year, citing security concerns with the technology as well as the propensity of some tools to exhibit "bias, unpredictability, and malicious behavior," according to an April 2023 internal guidance memo obtained by WIRED through a public records request. Jamie Holcombe, the chief information officer of the USPTO, wrote that the office is "committed to pursuing innovation within our agency" but are still "working to bring these capabilities to the office in a responsible way."
Paul Fucito, press secretary for the USPTO, clarified to WIRED that employees can use "state-of-the-art generative AI models" at work -- but only inside the agency's internal testing environment. "Innovators from across the USPTO are now using the AI Lab to better understand generative AI's capabilities and limitations and to prototype AI-powered solutions to critical business needs," Fucito wrote in an email.
Paul Fucito, press secretary for the USPTO, clarified to WIRED that employees can use "state-of-the-art generative AI models" at work -- but only inside the agency's internal testing environment. "Innovators from across the USPTO are now using the AI Lab to better understand generative AI's capabilities and limitations and to prototype AI-powered solutions to critical business needs," Fucito wrote in an email.
That's a shame (Score:5, Interesting)
It actually seems like a great first level search for prior art.
Re: That's a shame (Score:1)
Re: (Score:2)
To do a search for prior art, yea, that's useful, but the decision making should not be using AI. Also, what if the AI is flawed, so that it will not properly return results just because the AI developer wants to see Apple or some other company get more patents, so will ignore the "prior art" if what is being looked at was submitted by one person or company on a list?
Re: (Score:1)
It actually seems like a great first level search for prior art.
The top brass at the patent office realizes that at the end of the day, they get paid to issue patents.
Finding prior art interferes with that goal. It's for the best to proscribe tools that would make finding prior art too easy.
Re: (Score:3)
Spotted someone who knows absolutely nothing about the patent process.
Re: (Score:2)
I know plenty about the patent process, and I assure you I'm on plenty of patents. My old employer would routinely come around asking if anybody had done anything that could conceivably be patented, no matter how stupid or obvious. This was because big companies negotiate patent deals largely based on the count of patents they have vs the other companies.
It may take a few years, but the main thing involved with getting patents issued is the perseverance of your patent attorneys.
Like I said, this whole indus
Re: (Score:2)
GENERATIVE AI, not any AI.
Huh? (Score:2)
They should use it to check for prior art before granting a patent, instead of what they currently use which is nothing.
Re:Huh? (Score:4, Interesting)
Or you could ask it to solve a problem if the AI comes up with a similar solution then you could say the patent is not creative enough.
Re: (Score:2)
That's exactly what patent examiners do. It requires understanding what a "similar solution" would be, in the end you haven't suggested anything.
Re: (Score:2)
Re: (Score:2)
How would you know. What you're telling us is you don't.
On the other hand ... (Score:4, Funny)
Seems just like humans (Score:2)
bias, unpredictability, and malicious behavior,
The more I here about AI the more I believe it behaves like humans
Re: (Score:2)
It's created by humans so why wouldn't it behave like them?
Re: (Score:2)
Generative AI doesn't "behave". Do NOT anthropomorphize computer software.
Re: (Score:2)
What a weird thing to say. My wife made an apple pie the other day, do you expect it to behave like a human?
Bad wording (Score:2)
I can accept that AI at this stage shouldn't be trusted for most things, but the "Malicious Behavior" is something that is completely impossible for an AI, because that implies that there is an intent to do something wrong. Unless those who write/design the AI are specifically coding negative things in for some reason, there can be no malicious intent in anything it does, because it doesn't have desires.
AI doesn't understand concepts, so can not decide if something is actually new, or just an old idea wit
Re: (Score:2)
"...any of the people working in the patent offices clearly don't understand enough to see when something is just an old idea that has a minor change to how it is presented."
No, that's just what ignorant people on /. say. "People working in the patent offices" know a hell of a lot more than the people posting here.
Makes total sense (Score:2)
Patents are the one thing where cloud-based AI is prohibited for good reasons. The thing uses your prompts to train the model. Workers must not check your patent ideas in cloud-based AI, and patent attorneys must not input queries that contain clues to to-be-patented ideas.
1) Submitting a prompt (patent title, abstract or full text) is a form of disclosure of the idea, which can and will invalidate the patent, worldwide, instantly, if someone can prove it was done;
2) If a random person later asks a question
Re: (Score:2)
"1) Submitting a prompt (patent title, abstract or full text) is a form of disclosure of the idea, which can and will invalidate the patent, worldwide, instantly, if someone can prove it was done;"
No, it is not. A "disclosure" matters if it is a public "disclosure", whether it is a "prompt" is ultimately irrelevant because it being a prompt does not impact whether it is a "public" prompt or not. Even worse, you are wrong about what a disclosure means to a potential patent. A disclosure prior to filing ma