Understanding IP Matters: AI Needs the Right Balance of Innovation and Regulation to Thrive

Some say overregulation of AI will impede development. One innovation policy expert believes both AI and IP rights need a clearer context for AI to operate productively. On the current episode of Understanding IP Matters (UIPM), Dr. Brandie M. Nonnecke, a policy expert with a background in journalism, discusses what responsible AI should be and says that regulation of AI is necessary and, if done properly, will not stifle innovation. Nonnecke knows the reluctance about regulation is a “knee jerk reaction by companies [and] investors to evade oversight,” she says, adding: “But let’s remember that regulation can actually spur a more competitive environment, a more competitive market for them to operate within.”

Nonnecke is Founding Director of the Center for Information Technology Research in the Interest of Society (CITRIS) Policy Lab, headquartered at UC Berkeley. She is an Associate Adjunct Professor at the Goldman School of Public Policy (GSPP) where she directs the Tech Policy Initiative, a collaboration between CITRIS and GSPP to strengthen tech policy education, research, and impact. Nonnecke is a faculty co-director of the Berkeley Center for Law and Technology at Berkeley Law where she leads the Project on Artificial Intelligence, Platforms, and Society. She also co-directs the UC Berkeley AI Policy Hub,  an interdisciplinary initiative training researchers to develop effective AI governance and policy frameworks.

In this episode of UIPM, Nonnecke and UIPM host Bruce Berman discuss:

• – The current status of AI regulation: “So, the European Union passed the most comprehensive AI related legislation to date, the EU AI Act, which is now in force,” but “the United States is a whole other story.”
• -That Nonnecke gives credit to U.S. Legislators regarding AI. She says: “I believe very strongly that they have put a lot of effort into learning more about the technology and especially having their staffers learn more. I myself have hosted briefings before members of Congress essentially debunking the misunderstandings around what the technology can actually do.”
• -Even though we “do not have a comprehensive law governing AI, let’s not forget that we have established laws that apply. If you develop an AI system that causes harm, as a developer, you could be sued under product liability tort, or if you develop an AI tool for reviewing resumes and it’s discriminatory, well, you can be sued on non- discrimination laws.”
• -That some countries may be favoring open source AI models, as opposed to proprietary closed AI models, because “they won’t be able to keep pace with the United States. So I’m definitely seeing European countries are pushing more for open.”
• – Journalism and Policy are synergistic. “It’s important to communicate to the public, how does the technology actually work? What are the emerging laws and policies that are shaping this technology? So I pride myself on being a good communicator to a diverse audience and the credit goes to having a master’s degree in journalism.”

Key Responses 

Technology evolves quickly, but not always in a straight line. It has never been easy to separate the good tech from the bad. AI makes it even harder. How should we define responsible AI?

Nonnecke: “That’s quite the big question. I think defining responsible AI is very contextual. Also, the technology is dual use, it can have positive impacts on our society, simultaneously causing negative externalities. What’s most important about responsible AI development is taking stock of this careful balance between the benefits and risks and ensuring that we’re putting in place appropriate technical interventions or governance interventions to ensure that the balance leans in favor of the benefits.”

Is open source something that AI policy people are embracing? Are they more comfortable with proprietary code? What is their general thinking?

Nonnecke: “I would say it’s pretty even split, honestly. There’s the benefit of open AI models that can spur other startups, other companies, help the economy. On the other side, there is a concern that these open models can be co opted by nefarious actors who can then scam, scheme, and take advantage of our democracy. So it’s actually a little bit of a hustle right now on which side we should go, but we’ve seen now in the European Union a push for open models, and it’s pretty evident that they want to be able to use those models domestically.”

Much of generative AI is founded on a predatory infringement model. What does it do to innovation? What is it really doing here in the long run?

Nonnecke: “This is the question we’re all facing right now. If you look at how OpenAI built  ChatGPT, they scraped the internet pulling in copyrighted content, right? There’s a lawsuit between the New York Times and OpenAI currently. So we’re seeing some of the other big players who hold copyright pushing back and those cases will provide clarity on whether or not scraping the internet[violates copyright law], even though that does not violate, the computer fraud and abuse act because the content is publicly available.”

More Highlights

Nonnecke says that we’re in a very interesting time, and that “[d]uring the first Trump administration, we did see significant work toward best ensuring that the United States was positioned to adopt this technology in a way that benefits everybody,” and that we should especially “keep an eye out for California” to lead the way.

Listen to the entire episode to learn:

• What fair use could be in a generative AI world: “Could we turn this on its head and also say, should I be citing my generative AI tool for the creation of an image? There’s a lot of art and skill that goes into prompts. There’s a whole burgeoning field of prompt engineers.”
• Nonnecke is “seeing a move away from these very, very big models to smaller models, where they use a method called RAG, which is retrieval augmented generation.”  These smaller models have similarities to ChatGPT, but when you query the smaller models “you’ve only allowed it to be trained on a specific set of data.”
• That both the public and private sectors have a fear of missing out on generative AI, that now everyone is “rushing to be able to adopt this technology or build their own. Even in the state of California, the state of Washington, we have executive orders specifically focused on generative AI tools and their adoption within our government.”
• That valuations for AI companies are wildly high and that “the bubble will burst at some point, just like we had [with] the dot com boom. I guarantee.”