Artificial Intelligence and Intellectual Property

Artificial Intelligence & Intellectual Property Presented by: European Intellectual Property Attorneys Date: 26 November 2020 What is AI? - A machine - That receives inputs - That generates outputs - That learns - Roger Penrose defines it as "The Emperor's New Mind" • The inescapable conclusion seems to be: Mathematicians are not using a knowable sound calculation procedure in order to ascertain mathematical truth. We deduce that mathematical understanding – the means whereby mathematicians arrive at their conclusions with respect to mathematical truth – cannot be reduced to blind calculation! In → out - Algorithms take an input and give an output - Conventional computer programming defines the algorithm - Machine learning defines the output (to train the algorithm) Today's Use Cases - Healthcare - diagnosis, smart surgical tools - Transport - self-driving vehicles - Manufacturing - robotics - Smart devices - "Hey Google!" - Finance - fraud detection - Many, many others... Where's The Beef? 1. Al is created 2. AI can create 3. Creativity is valuable 4. The law protects it What Might Be Protected? - Architecture of AI - Parameters of the AI - Use of the AI - Outputs of the AI - Inputs to the AI Copyright - Copyright protection available for source code: computer programs are "literary works" - Huge amounts of work necessary to collate training data etc.: sui generis database rights - “Authorship” • The person who creates the work • In the case of a literary, dramatic, musical or artistic work which is computergenerated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken. Open Source - Copyleft - TensorFlow • Used by Dropbox, eBay, Twitter etc Trade Secrets - Proprietary knowledge - Indefinite duration - For AI: • Note difference between training and inference stages • Where is each deployed? Patents - Disclosure • - Duration • - Breadth of coverage Requirements • - 20 years (in most cases) Strength • - Must give the public an enabling disclosure Novelty & inventive step in a technical field Ownership • Stems from the inventor AI Can Create - ...but is it creative? - All computer programs will provide an output given an input. Machine learning provides an output that was not anticipated/ordained by the designer of the system. Is that creativity? DABUS - Device for the Autonomous Bootstrapping of Unified Sentience - An inventing machine, not created to solve a particular problem - 'Fractal Container' - 'Neural Flame" No Dice So Far - - USPTO, EPO, UKIPO and UK courts all agree • "Inventor” means a person • DABUS can't be one Is this a problem? • Solution there in copyright "authorship" provisions? Patents 101 - Novelty - Inventive Step - Technical subject matter • Computer programs... It's Technical - EPO Guidelines for Examination Non-Technical Features Ignored - Τ 0641/00 (Two identities/COMVIK) What Isn't Technical? - Business methods - Aesthetic creations - Presentations of information - Mental acts - Mathematical methods - Computer programs - Only excluded to the extent that the inventions relate to this subject matter "as such". EPO Guidelines for Examination - A mathematical method may contribute to the technical character of an invention ... by its • Application to a field of technology and/or • By being adapted to a specific technical implementation. Application To A Field Of Technology - EPO Guidelines for Examination • "Use of a neural network in a heart-monitoring apparatus for the purpose of identifying irregular heartbeats makes a technical contribution" - Classification of digital images, videos, audio or speech signals based on low-level features (e.g. edges or pixel attributes for images) - But not classification of text documents Technical Application Examples - Digital audio, image or video enhancement or analysis - Encoding data for reliable and/or efficient transmission - Encrypting/decrypting or signing electronic communications; - Controlling a specific technical system or process, e.g. An x-ray apparatus - Providing a genotype estimate based on an analysis of DNA samples - Providing a medical diagnosis by an automated system processing physiological measurements Specific Technical Implementation - Board of Appeal case T 1358/09 - Example: adaptation of an algorithm to exploit word-size shifts matched to the word size of the computer hardware Where Does This Leave "Core" AI - Advantages across use cases • Better generalization / reducing overfitting? • Improved accuracy in inference? • Faster training of a neural network model? Generative Adversarial Networks - "GAN can be used to detect glaucomatous images helping the early diagnosis which is essential to avoid partial or total loss of vision. - GANs that produce photorealistic images can be used to visualize interior design, industrial design, shoes, bags, and clothing items or items for computer games' scenes. - GANs can reconstruct 3D models of objects from images, and model patterns of motion in video. - GANs can be used to age face photographs to show how an individual's appearance might change with age. - GANs can also be used to transfer map styles in cartograph¹ or augment street view imagery." Batch Normalizations - Standard techniques suffer from "internal covariate shift" making them unstable and difficult to train - Technique resolves this - "Faster training" is technical(?) Other Examples - Stylized photos - Adversarial inputs - Data encoding - Robotics