Michael
Toker

Michael Toker is a computer scientist and PhD candidate, focusing on analyzing the internal computations of large multi-modal AI models—systems that can process and combine information from more than one type of input. His research aims to understand how information flows within these complex neural networks, how it is stored, and how outputs are generated, such as in text to-image models, which generate images from written descriptions. By understanding the roles of different components within the model, Michael seeks to enhance model capabilities and mitigate unsafe behaviours in its outputs, such as biased, misleading, or harmful responses.

Michael seeks to enhance multi-modal AI model capabilities and mitigate unsafe behaviours in its outputs, such as biased, misleading, or harmful responses.

Michael holds BSc and MSc degrees from the Technion – Israel Institute of Technology and recently interned in Nvidia’s AI research team. He founded the Google Developer Student Club at the Technion and was part of the organization team of the “Doing Good” CS Hackathon for three years. Michael lives in Ramat Gan with his fiancée and their cat. He enjoys running, biking, playing chess, and listening to podcasts and audiobooks in his leisure time.