To keep up to date with the latest news, our AI Practice Group Have We've put together an overview of the trends we're watching.
- Last week, the U.S. House of Representatives Judiciary Committee's Subcommittee on Courts, Intellectual Property, and the Internet Hearing A hearing was held entitled “Artificial Intelligence and Intellectual Property: Part III – Intellectual Property Protection for AI-Assisted Inventions and Creations.” The purpose of the hearing was to consider the criteria and policy considerations regarding whether intellectual property protection should be afforded to inventions (patents) or creations (copyrights) produced with the assistance of AI. Multiple stakeholders, including policy experts and law school professors, testified before the subcommittee. Claire LaPorte, former head of intellectual property at Ginkgo Bioworks, urged the subcommittee: Prioritize leading the world in AI-assisted biologyFor example, it could dramatically reduce the scale and time it takes to train an AI model to solve a bioengineering problem. Regarding human copyright in AI-generated artwork or music, Professor Sandra Eysters of the Antonin Scalia School of Law said: testified The pertinent inquiry is “Did the human authors use GAI?” [like an artist might utilize] Have other tools or materials been used to realize the artist's creative vision, or has the use of GAI become a substitute for the artist's own creativity?
- The USPTO Guidelines Guidelines for practitioners to use AI tools in patent and trademark prosecution. Finally, the guidance reiterates that the use of AI tools is no different than a practitioner working with a paralegal or junior attorney. Practitioners are responsible for the legal and technical accuracy of all documents submitted to the USPTO. In other words, practitioners should carefully review AI-generated output and should not simply trust that these tools will always get it right.
- There have been concerns for some time that AI companies would run out of training data for large language models (LLMs), but it appears these concerns have not been fully resolved. Hitting a wallSeeking new sources of data, companies are reportedly transcribing millions of hours of audio from YouTube into text, considering acquiring major publishers, or using synthetic data for training – the former raising questions about copyright infringement, and the latter approach not yet proven viable.
- Microsoft and OpenAI $100 Billion Data Center A facility for AI computing is under construction in the US that could potentially run on a nuclear reactor.
- Chip giant NVIDIA has announced a new Blackwell B200 GPU platform, It will contain 208 billion transistors and scale to 20 petaflops of computing power, and is expected to consume significantly less power per calculation compared to current generation GPUs.