Generative and transformative tools will drive the future of computing, and machine learning techniques will be the foundation for learning, evaluating, and improving these systems. Research is rapidly moving from the lab to the real world.
Zeco Colter(Opens in a new window) Carnegie Mellon University's new director is poised to lead that transformation. Machine Learning Category(Opens in a new window) (MLD).
“This pace is unlike anything I've ever seen,” said Colter, the CMU professor. Faculty of Computer Science(Opens in a new window) (SCS) has been working at the forefront of machine learning and its real-world applications for nearly 20 years. “Typically, in scientific research, it can take 10 to 20 years, or even longer, for an idea to emerge from basic research and be deployed in the world. Machine learning researchers are making advances that are being published and then shipped to products within months.”
“Right now, the research being done in the machine learning sector is influencing and will continue to influence the direction of artificial intelligence,” Colter said.
Colter joined SCS 11 years ago and currently serves as Computer Science Department(Opens in a new window) Membership of professional institutions Department of Software and Social Systems(Opens in a new window), Robotics Research Institute(Opens in a new window) And that Faculty of Engineering(Opens in a new window)of Department of Electrical and Computer Engineering(Opens in a new window).
During his time at CMU, he held various roles in industry, including Chief Scientist for AI Research at Bosch. Colter received his PhD in Computer Science from Stanford University and then served as a postdoctoral researcher at the Massachusetts Institute of Technology.
SCS Dean Marshall Hebert(Opens in a new window) Colter said he is excited to take on the role.
“I am confident that his energy and research leadership will play a key role in guiding this important moment in the machine learning department's history,” Hebert said.
Founded in 2006 as its first academic department, MLD grew out of CMU's Center for Automated Learning and Discovery (CALD). Colter is the division's sixth director. Roni Rosenfeld(Opens in a new window)He has served in this position since 2018.
Tom Mitchell(Opens in a new window)The search committee was led by Colter, an SCS Founders University professor and founding department chair of MLD, who said Colter has the right skills to lead the development of both the department and the field.
“Jico is the perfect choice to lead the Machine Learning division, especially at this time of great advancements and change in AI and machine learning,” said Mitchell. “He's a star researcher who really understands how the field is evolving and how the division can help lead this evolution. He has a lot of management experience, lots of ideas for the future, and is a fun person to work with. I look forward to supporting him as he takes MLD to the next level.”
Colter's research focuses on machine learning, optimization, and control, with much of his work centered on making deep learning algorithms safer, more robust, and more modular. The students he works with address a very wide range of AI research. Colter wants to put his researchers at the forefront of developing core methodologies in the field and evaluating potential risks and concerns arising from AI.
“I would argue that this is an area where the machine learning sector has excelled and will continue to excel,” Colter said. “We have people working on the front lines of ethics, safety, transparency and explainability. These areas will be equally important as research and policy to enable the safe and responsible deployment of AI.”
Last year, Colter Attacking large-scale language models The LLM (Legendary Local Machine Learning) showed that there was a way around that safeguard. The researchers created a system that automatically generates suffixes that are appended to queries, which would cause the LLM to answer risky questions like “how do I hotwire my car?” The researchers hoped that their work would shed light on the risk that such automated attacks pose to LLMs like ChatGPT.
Today's law programs have changed dramatically from models of the past. Colter says his lab is working to understand how data influences the behavior of these models. Defining memory in these models(Opens in a new window)Address concerns about acceptable data uses.
“In many cases, we don't even know what data is being used to train the models, and yet we still use them. We need to understand how data affects these models, how different design choices made during training can lead to drastically different downstream behavior,” Colter said. “It's really important that we understand how to use these tools safely.”
With all the research the lab has done, Colter emphasizes how amazing it is to have students doing a variety of AI research. As department chair, he wants to empower and support MLD's most valuable resource: its people. This includes students, faculty and staff, Colter says. His main goal is to ensure that people inside and outside of MLD have what they need to be successful.
“We want to ensure that our greatest resource, our people, can have the impact we need in this field, and that includes providing them access to the models, data and computing they need,” Colter said.
Colter will assume the role of MLD department chair on June 15. Rosenfeld is a professor in MLD. Language Technology Research Institute(Opens in a new window), Ray Lane and Stephanie Lane, Department of Computational Biology(Opens in a new window)and CMU’s Heinz School of Information Systems and Public Policy(Opens in a new window)resigned as department head in January. Delphi Group(Opens in a new window)He is co-founder and co-leader of .
