Technical Contributions under EPO and UKIPO Frameworks

Patent Eligibility and AI: Technical Contributions under EPO and UKIPO Frameworks Speaker: Virginia Driver, CPA, EPA, Page White Farrer; Mr. Hervey Date: 4 December 2024 Patent Eligibility At The EPO - Patent eligibility at the EPO • Mathematical methods are not patentable as such • Machine learning is treated as a subset of mathematical methods • A technical application of a mathematical method is patentable • A technical implementation of a mathematical method is also patentable Patent Eligibility At The UKIΡΟ - Novel features need to make a technical contribution • Applied Al • Core Al Technical Application - Multiple straightforward paths to patent eligibility • Technical application  Claim limited to a technical application o Image or audio processing o Speech recognition o Automated medical diagnosis from physiological measurements • Problem-solution approach  Novel features achieve technical effect o Further technical effect not relevant - Eligible effects • Improved performance (better accuracy, precision...) • Improved efficiency (faster, reduced memory...) • Improved training (faster convergence, less training data...) Gen Al And 'Skill Convergence' - Gen Al • Computer vision • Signal processing • Language modelling • Cybersecurity • NLP • Mathematics • Business analytics • Computer programming • Robotics - Who is the skilled person/team? - What is the 'activity' and is it technical? Sufficiency - No need to include full training data set - Do need to enable any new models/how existing models applied in new use case Don't Forget To Think About - Data structures - User Interfaces - Physical inputs (e.g. sensors) - Virtual models (simulations) Patent Law - Fit For An (Emotional) Al Age? - What activities does patent law need to incentivize, and where should the balance be struck if aiming to motivate technical innovation by applicants and to encourage public disclosure of inventions (to stimulate innovation by others)? • E Mishan & Sons v Hozelock [2020] EWCA Civ 871, per Lord Justice Arnold  "... the patent system aims to incentivize technical innovation, and investment in and disclosure of such innovation, by conferring limited monopolies. Monopolies are generally contrary to the public interest, however, because they prevent competition. Patent law contains a number of mechanisms which are designed to strike a balance between these conflicting considerations. - How does the requirement of public disclosure work if an invention is made by machine learning (perhaps using non-public data), and how important is it? • Geofabrics v Fiberweb, [2020] EWHC 444 (Pat)  "The law does not require a patentee to explain the physical theory of the invention. Once a product is made according to the patent, it can be tested without undue difficulty to see whether it works." - If we try to fit AI-generated inventions into the existing system, how will patent requirements, such as inventive step and sufficiency be applied? • Threshold for human contribution • Rectification s.13(3) • Not a patentable invention s.72(1)(a) • Grant to a person not entitled s.72(1)(b) • Abuse of a dominant position • Obviousness