“Where are we getting closer and closer … we want to use these tools, and society wants to use these tools. But if you’re going to throw things out at the conceptual stage, the U.S. is at a disadvantage.” – Corey Salsberg
USPTO East Coast Listening Session on AI Inventorship
The United States Patent and Trademark Office (USPTO) today held an East Coast listening session on AI inventorship, and both USPTO staff and patent stakeholder speakers shared their thoughts on how generative artificial intelligence (AI) has almost creeped out. We pondered possible approaches to patenting in the visible world. everyone. Many speakers cautioned against acting too quickly to change the rules of AI-generated inventions, while others said that if nothing was done he would wreak havoc on the USPTO. , warned that it could result in significant economic and evolutionary losses for the country.
The hearing comes shortly after the US Supreme Court yesterday denied a petition by Dr. Stephen Thaler asking the court to consider the following questions: With the High Court sidestepping the issue and Congress being known to be very slow to respond to changes in legislation to accommodate new technology, Ball said his It sits squarely in the USPTO courtroom.
USPTO Director Kathi Vidal recently said this year would be her “year of action,” and in a short pre-recorded message to attendees today, the USPTO is experiencing a “surge” in more than half of its patent filing technology centers. He said he does. Take advantage of AI. She also said she feels her February call for views on AI inventors “gets to the heart of the matter.” Her comment deadline is May 15th.
Salsberg: Clarify Concepts
Listening sessions ran from 10:30 am to 3:00 pm and featured a variety of speakers. IBM representatives said the status quo shouldn’t change, but Novartis’ Corey Salsberg said AI is being used in the biopharmaceutical field in ways that are already testing the limits of current systems. explained.
“By using AI machine learning, the time to screen compounds and analyze them for desirable properties can be reduced to minutes or hours instead of months or years. We can screen it to determine if the candidate has the desired effect,” said Salsberg. .
But even more troubling, he explained, is the use of an AI known as JAEGER in the field of generative chemistry. The researcher said he trained JAEGER on his 21,000 molecules in the Novartis library that had been previously tested for antimalarial properties, and the AI came up with 282 “realistic, brand new virtual molecules with antimalarial properties.” said Salsberg. Humans then used other AI tools to test his two most promising tools, ultimately proving them effective. This shows that appropriate policies need to be put in place now, Salsberg added, but so far it’s too early to focus solely on when AI can become an inventor. , the problem is too narrow. According to Salsberg, “For now, the more relevant question is whether and how human use of AI affects the status of humans as inventors.”
This analysis suggests a US-specific requirement for “conception” in the invention process raised by some speakers. Conception has to do with forming a “clear and lasting idea of a complete and valid invention” in the mind of the inventor. But the criterion is to use AI to comb through millions of compounds to narrow down to promising structures, or to use AI to propose entirely new molecules that did not exist before. does not match. “When these activities and connections occur in the mind, it is indeed called part of the concept. So the key question is whether the actual connections, correlations, or designs exist outside the human mind.” Whether the concept is still satisfied when it happens ‘in silico’. , with AI,” Salsberg said.
Novartis believes current legislation is flexible enough to account for this, but “but there is also enough uncertainty and ambiguity that the Office should consider clarifying its guidance. For example, the MPEP does not include an inventor’s head of information, even though older versions said that inventors could consider external sources of information in the invention process. There is an entire section devoted to the concepts inside. Another section also contains case law on the concept of chemical compounds, which states that simply telling his AI to define or design a compound is not enough to make an AI an inventor or co-inventor. It can be interpreted as suggesting that it is not enough. Salsberg sums it up:
“The main point is that we are getting closer and closer to where we will be more efficient in R&D. We want to use these tools, and society wants to use these tools. Because it makes the process faster, more effective, and makes better medicine, especially if other countries don’t have fertility requirements like we do.”
Ultimately, Salsberg said he would probably choose to do away with the conception requirement, but the Office doesn’t have to go that far just yet. “It just needs to be properly interpreted to explain technological advances that have made it possible to do outside the mind what used to be inside the mind.”
Feldman: Fools rush
On the other hand, Professor Robin Feldman of UC College of Law in San Francisco says that “only fools rush” into such matters. Feldman said human inventors are incentivized by having their name on patents, but AI is not. “Putting AI in patents is neither socially desirable nor coherent,” Feldman said. In addition, there are too many questions to answer, such as whether the AI constitutes his POSITA or when it will compromise. Feldman recommended potentially using the pharmaceutical data rights model for AI and shortening the protection period. But for now, she said, AI is still a tool and should go “slowly.”
Crouch: move quickly
Professor Dennis Crouch of Patently O also participated, calling for clear guidance from the USPTO urgently due to the “unprecedented pace” of AI. If AI were a human, there are already some cases where it would be credited as the sole inventor, Crouch said, so the problem is faced head-on. “It is imperative that the USPTO act quickly,” he added, as the USPTO has produced some technical improvements.
Crouch’s solution is a special category of patent for inventions that do not involve direct human contribution. This could include, for example, shortening the term of a patent or limiting claims. Crouch likens generative AI to the following situation: amgen vs sanofiawaiting Supreme Court ruling.of Amgen‘humanized mice’ are AI-innovative outputs in that mice were prompted in the form of PCSK9 injections and in response created antibodies that were stored in the spleen for later recovery. is equivalent to generating
But analogies aside, it’s time for guidance, Crouch said:
“Patent attorneys face considerable uncertainty as to the appropriate course of action. Innovators clients seek to protect valuable inventions, but lack clear guidelines and ethical concerns for patent attorneys. A dilemma arises: The USPTO has provided guidance to show that patent applications can properly list human contributors to a concept as the sole inventors, even when AI and other tools provide a significant element of discovery. We need to deliver quickly.”
Crouch also warned the USPTO not to fall into the trap the Copyright Office recently made in its guidance on AI. Crouch, in fact, called for the resurrection of the idea of establishing a U.S. Intellectual Property Office to enable intellectual property policymakers to consider issues holistically across patents, copyrights and trade secrets. . “I don’t want to miss out on new treatments because of some regulation we’re enforcing,” Crouch said. It is essential to provide guidance on how to handle
The USPTO will hold a West Coast Hearing Session on AI Inventions on May 8th.
