By Wayne Gillam / UW ECE News
UW ECE Professor Maryam Fazel is a program co-chair for the 2025 International Conference on Machine Learning, which will be held from July 13 to 19 in Vancouver, Canada. She is one of four faculty members from universities across the United States and Canada, who together are overseeing all aspects of peer-review of the paper submissions and of producing the event. Photo by Ryan Hoover / UW ECE
Artificial intelligence is all over the news these days. This powerful technology comes with a bright promise to usher humanity into a new era of better health, connectivity, and prosperity. But it also holds the dark potential to disrupt economies, damage social systems, and perhaps even plunge our world into information chaos. With so much at stake, it might be encouraging to know that scientists and engineers recognize these serious issues with artificial intelligence and are tackling them right now, from every perspective imaginable.
A case in point resides within a subset of artificial intelligence that doesn’t get as much public attention — machine learning, a field of study aimed at developing statistical algorithms that can learn from data and perform tasks without explicit instructions. Machine learning is at the core of artificial intelligence. For example, machine learning enables large language models like ChatGPT, computer vision in self-driving cars like those at Waymo, and algorithms that underpin popular social media platforms like TikTok.
The upcoming 2025 International Conference on Machine Learning, which will be held from July 13 to 19 in Vancouver, Canada, is dedicated to the advancement and improvement of this branch of artificial intelligence. With well over 15,000 attendees expected this year, the ICML is the oldest, second-largest and fastest-growing conference of its kind in the world. Over 12,000 research papers focused on machine learning have been submitted to the conference as well as 350 “position papers,” which are designed to bring attention to urgent issues in machine learning, such as privacy, safety, algorithmic biases, and intellectual property concerns. Conference attendees will examine and discuss these topics in detail, a process that builds groundwork for solutions to some of the most urgent and complex problems that artificial intelligence and machine learning present today.
Register to attend the 2025 International Conference on Machine Learning
UW ECE Professor Maryam Fazel is a program co-chair for this year’s Conference. She is one of four faculty members from universities across the United States and Canada, who together are overseeing all aspects of peer-review of the paper submissions and of producing the event. Fazel holds the Moorthy Family Career Inspiration Development Professorship, is the UW ECE Lytle Lectureship chair, and is director of the Institute for Foundations of Data Science at the UW, which brings together data science experts and tools from the mathematical, statistical, and algorithmic foundations of machine learning to address contemporary data science challenges.
“It is a privilege for me to be a chair of this Conference. I’m trying very hard to make ICML the best it can be, serve all the communities that are involved, and contribute to the growth of the field,” Fazel said. “I’m also looking forward to the productive discussions we will have about issues and challenges related to artificial intelligence and machine learning.”
I sat down with Fazel to learn more about this year’s ICML and how it will be contributing to the development of machine learning and artificial intelligence.
Why is the ICML important?
We are living in a very exciting time for machine learning and artificial intelligence. The field is evolving rapidly and the progress is fast paced. In my experience, this is the first time that this field has gotten so much of society’s attention, and the impact is so visible and tangible. The advancement of these tools is beginning to have a very broad effect on society, and there are new, impactful developments almost every day.
Artificial intelligence and machine learning are opening new ways to address some of the biggest challenges in science and engineering. For example, several of the Nobel prizes last year in scientific discovery went to researchers in machine learning and artificial intelligence. They used these tools for groundbreaking discoveries in physics, chemistry, and other sciences. That’s one very big and promising direction for these technologies.
So, the pace of progress is really exciting. But as a researcher, this also gets me thinking about how to address some of the pressing issues arising with this technology. A lot of challenges are becoming more apparent over time. In general, AI systems have issues with reliability, interpretability, security and safety, privacy, and things like energy efficiency. All of these problems are unresolved. These are things that are being worked on now, and it’s exciting to try to address these issues. This conference provides us with an important forum to do that.
Who attends the ICML?
A very broad range of academics — students, faculty, postdocs, researchers, teachers — but also industry researchers and non-academics. Attendees come from very different fields. There are people who work on algorithms, theoretical computer science, statistics, applied math, to people who do systems and hardware engineering and people who work on application areas for machine learning and artificial intelligence. It’s a broad mix of very different sets of people. The interaction between academia and industry in the Conference is also pretty strong. There is an industry expo, for example, where companies introduce their work, and they have interaction with Conference attendees. There’s also talks given by industry leaders and researchers.
What does the ICML offer those who attend?
There are many different sessions, tutorials, and workshops. There are also talks by experts in the field, which includes six invited speakers who will give keynote talks. There are oral presentation sessions, in which there are short, 15-minute talks presented from selected papers in the Conference. There are also large, poster sessions throughout the Conference, in which many of the accepted papers are presented by their authors as posters. This is a very nice model that works well in these large-scale conferences. This way, authors and their audience can interact individually, and we can fit in many papers.
Those who attend will get exposure to the most recent, most active research and development and progress in the field. So, it’s very educational. Also, given the breadth of the Conference, it helps to foster collaboration between different subfields and even outside the technical field of machine learning. I think people will gain a lot by learning directly, making new connections, networking, and being stimulated by new ideas for their own work.
How would you describe research featured at the ICML?
The ICML has a broad focus on machine learning, so it touches upon many different aspects of the field. You can say it starts with general machine learning methods and tools and theory, but it also more specifically focuses on deep learning, evaluation of AI systems, things like meta-learning, human-AI interactions, learning theory that touches on statistical, mathematical, and algorithmic theory, optimization of machine learning models, and reinforcement learning.
I’d also like to say that my co-chairs and I care a lot about the quality of the work that gets presented at the ICML. We also care about the quality and integrity of the complex peer-review process that selects research papers featured at the Conference. While there are challenges when the process has to scale up so fast from year to year (we had more than a 30% increase in the number of submissions this year over last), we have been working hard to put measures in place to ensure quality, integrity, and ability to scale in the future.
What do you think will be the hot topics this year?
I think fundamental questions about how these systems work and how we can make them more robust, reliable, secure, safe, and interpretable will be hot topics as well as how scientific discovery will be revolutionized using artificial intelligence. These are some of the driving questions for current research. Plus, a topic that I think will be addressed in many of our position papers is the broader picture of how society should even think about these issues as artificial intelligence is integrated more and more in society. Things like interactions with the law, privacy, legality of accessing data, copyright, and intellectual property will be covered. These are really important issues that go outside of the technical field of machine learning, but the position papers as well as some of the invited keynote talks make that connection.
Can you describe the different ways people can attend?
We are aiming to make the ICML as inclusive as possible. To that end, we are ensuring the venue is accessible for people with disabilities. We’re also providing on-site childcare and support for nursing mothers, so parents with small children can attend. All presenters will be live and in-person, but there is an option of registering for virtual attendance for those who cannot attend in-person. And all talks will be broadcast online and streamed, so the audience can participate remotely and virtually. A few weeks after the Conference has concluded, all the papers and posters presented will also be available on the ICML website.
Is there anything else you would like people to know?
This Conference is open to everyone. Those who would like to attend should register early! There is a huge demand, and the venue size is limited, so if people are interested, they should act quickly.
Visit the 2025 International Conference on Machine Learning website to learn more about the event. More information about Professor Maryam Fazel is available on her UW ECE bio page.