Sara Hooker is a computer scientist specializing in artificial intelligence and machine learning. As of 2025, she serves as the head of Cohere for AI, the research division of the AI company Cohere, where she leads efforts to advance open-source AI research and promote responsible development practices. Hooker is widely recognized for her contributions to model interpretability, particularly her work on understanding how neural networks learn and make decisions, and for advocating for greater transparency in AI systems.
Hooker's research sits at the intersection of deep learning, interpretability, and fairness. She has published influential papers on topics such as feature attribution, the impact of training data on model behavior, and the challenges of evaluating large-scale AI models. Her work has been instrumental in highlighting the importance of understanding not just what AI models do, but why they do it, a concern that has become central to the field of responsible AI.
Early Career and Education
Hooker completed her undergraduate studies in computer science at the University of Toronto, where she first became interested in artificial intelligence. She later pursued graduate studies at the same institution, earning a master's degree and a PhD in computer science. During her doctoral research, she focused on machine learning and neural networks, laying the groundwork for her later work on interpretability.
After completing her PhD, Hooker worked as a research scientist at Google DeepMind, where she contributed to projects involving large-scale machine learning models. Her time at DeepMind allowed her to engage with some of the most advanced AI systems in the world and deepened her interest in understanding their inner workings.
Research Contributions
Hooker's research has addressed several fundamental questions in AI. One of her notable contributions is the concept of "forgetting" in neural networks, where she investigated how models lose information about certain data points during training. Her work on this topic provided insights into the dynamics of learning and the potential biases that can arise from data selection.
She has also been a vocal advocate for the importance of evaluation in AI research. In a widely cited paper, she argued that the AI community often over-relies on a few benchmark datasets, which can lead to misleading conclusions about model performance. She proposed more rigorous evaluation practices to ensure that AI systems are robust and reliable.
In addition to her technical work, Hooker has written and spoken extensively about the social implications of AI. She has emphasized the need for diversity in AI research and the importance of considering the perspectives of marginalized communities when developing AI technologies.
Leadership at Cohere for AI
As head of Cohere for AI, Hooker has overseen the launch of several open-source models, including the Command series of large language models. She has championed the idea that AI research should be collaborative and transparent, and she has worked to make Cohere's models and research publicly accessible. Under her leadership, Cohere for AI has also focused on multilingual AI, developing models that perform well across a wide range of languages, which is a departure from the English-centric focus of many AI systems.
Hooker has also been involved in efforts to improve the interpretability of large language models, which are notoriously difficult to analyze due to their size and complexity. Her team has developed tools and techniques to probe these models, aiming to uncover the mechanisms by which they generate responses.
Recognition and Impact
Hooker's contributions have earned her recognition within the AI community. She has been invited to speak at major conferences, including NeurIPS and ICML, and her papers have been widely cited. She has also been featured in media outlets as a leading voice on responsible AI.
Her work has influenced both academic research and industry practice. Many AI labs now prioritize interpretability and fairness in their development processes, in part due to the advocacy of researchers like Hooker. She has also mentored numerous students and early-career researchers, helping to shape the next generation of AI scientists.
Personal Life and Advocacy
Outside of her professional work, Hooker is known for her commitment to increasing diversity in technology. She has spoken openly about the challenges faced by women and underrepresented groups in AI and has worked to create more inclusive environments in research settings. She has also been involved in initiatives to bring AI education to underserved communities.
Hooker continues to be an active researcher and a prominent figure in the AI landscape. Her work exemplifies the growing recognition that building powerful AI systems requires not only technical expertise but also a deep understanding of their societal impact.