Anima Anandkumar, the Bren Professor of Computing and Mathematical Sciences, has been appointed to the United Nations (UN) Secretary-General's Scientific Advisory Board, which brings together distinguished researchers from around the globe to advise UN leaders on breakthroughs in science and technology. Established in August 2023, the 15-person board provides "independent insights on trends at the intersection of science, technology, ethics, governance and sustainable development," according to the organization's website.
Known for her pioneering work in AI, Anandkumar has developed methods to model multiscale processes that are common in scientific domains. She invented neural operators, an AI method which can handle inputs and outputs at multiple scales or resolutions and learn mappings between function spaces. Such AI methods have proven effective in a variety of applications, including AI-based weather forecasting, work that provided a better understanding of COVID-19, and a system to land drones smoothly in complex turbulence.
We talked to Anandkumar about her new role on the UN Scientific Advisory Board and the importance of including scientists when seeking to leverage advancements that have the potential to benefit people and the planet.
How did it feel to learn that you had been appointed to this board of eminent scientists from around the world, including Yoshua Bengio, an A.M. Turing Award Laureate in AI, Thomas Südhof, winner of the 2013 Nobel Prize in Physiology or Medicine, and leadership from the World Health Organization and the UN Environment Programme, among others?
I was pleasantly surprised because other members of the board have been so pivotal in advancing their fields of expertise. I'm confident that the decisions we make will be very impactful. This is not just about talking; this is about doing. Key areas of our work will involve AI, including AI with respect to weather, climate, food, and diseases. At Caltech, I am leading the AI+Science initiative that integrates AI deeply into scientific domains. Standard AI approaches think mostly about language and maybe a bit about computer vision, but not in a broader sense for science. This interdisciplinary lens is precisely where the board has a lot of its focus in thinking about impacts in major scientific domains. It makes me very excited that it's very close to my own interests and my own approach to AI.
What are some of the biggest promises that AI holds for communities around the globe?
It is important that we take a global approach to large language models, which are a part of many AI systems. English is not the main language spoken in a lot of the countries, and training good local language models that can reach out to all people—not just the population that is educated—raises the knowledge base and, in turn, the literacy levels and education of people around the world. We need to think in terms of that more inclusive approach.
Beyond language, I trained the first AI-based high-resolution weather model, FourCastNet, back in 2021. Since then, the field has grown so much and now we open-sourced our model. There's been a lot of adoption of these models, not only by weather agencies here in the US and Europe, but also in so many other countries. For instance, in India there's been adoption of AI-based weather models to better predict the monsoons, which has a huge impact on farmers as so many crops are dependent on the timing of the monsoon. AI can help us understand the physical world and systems like weather, climate, flooding; all of these will have a huge impact on populations that are vulnerable and affected by climate change.
What are some of the challenges that you hope to help address through your participation on the board?
We need to take a balanced approach when it comes to AI. There's a lot of fear about AI and sometimes, especially in the public media, it's not necessarily framed in scientific terms. We have a lot of fearmongering, a lot of science fiction almost of what AI can be. So, we need to ground that in proper science and make sure that regulation is not something that only benefits big companies, and that regulation is not something that gets in the way of open-source innovation.
What are you excited about bringing to this new role?
I'm excited to bring my background in treating AI and science together. At Caltech, I co-founded the AI4Science Initiative back in 2018, and since then, we've reframed it as AI plus Science, because it doesn't just go one way. It's truly integrated together, and we need that understanding of the scientific domains—like the laws of physics—deeply embedded into AI for it to be trustworthy, for it to be able to make new discoveries like we are doing in my group at Caltech and so many other places. I'm very excited about thinking of AI not just as language models, but of models that understand the law of physics that can simulate the physical world.
I really hope that I can contribute in a positive way given a lot of challenges we see in the world. I do think having scientists in the room when all these important decisions are going to be made about AI, about technology, is crucial. When I look at my peers there, they are experts in their fields from economics to different areas of sciences, medicine, weather, and food. I'm hopeful and look forward to connecting with and learning more from so many amazing people on the board.
Why is it important that governmental organizations like the UN involve scientists in educating leadership and informing decision-making?
They absolutely have amazing scientists within the UN, but I think they are wanting to have that external perspective as well, especially on AI, to bring diverse opinions. That is so important because whenever you have diversity of opinions, that's where a more balanced perspective emerges. I've also been recently appointed to an AI advisory board for the Estonian government. They have an ambitious digitalization and AI agenda for the whole country, and they want to engage with scientists. I'm very encouraged to see all these organizations put scientists in the room.
Anima Anandkumar, the Bren Professor of Computing and Mathematical Sciences
