Resources

  • Articles
  • Authors
  • Quotations
  • Search
  • Tags
  • Topical Guide
  • Videos
HomeResourcesAuthors

Irina Rish

Portrait of Irina Rish

Irina Rish is a prominent researcher in the field of Artificial Intelligence, with a particular focus on achieving Artificial General Intelligence (AGI). She leads the Canada Excellence Research Chair in Autonomous AI, overseeing a large team of students, postdocs, interns, and collaborators. Her work explores the challenges and possibilities of creating AI systems that can generalize to a wide range of tasks and problems, mirroring human-level adaptability and learning capabilities.

Rish’s research delves into the complexities of out-of-distribution generalization, aiming to develop AI agents capable of learning and performing tasks significantly different from their training data. She draws upon principles from statistics, machine learning, and classical AI to create systems that are not only capable of mastering specific skills but also demonstrate the capacity for continuous learning and adaptation, akin to human cognitive flexibility. Her work resonates with transhumanist themes by exploring the potential for AI to augment human intelligence and solve complex global challenges.

Her presentation at the MTAConf 2024 focused on the technical aspects of AGI, emphasizing the importance of creating autonomous, multi-tasking systems capable of performing economically valuable work. Her engagement with the MTA highlights the intersection of AI research with philosophical and theological considerations regarding the future of humanity and the potential for technology to shape human evolution.

Videos by Irina Rish

Keynote - The AI Scaling Revolution and the Future of Intelligence
55:10

Irina Rish

Keynote - The AI Scaling Revolution and the Future of Intelligence

2024.04.13

Irina Rish explores the revolutionary impact of scaling in artificial intelligence and its implications for achieving artificial general intelligence. She discusses how foundation models—large-scale systems trained on massive, diverse datasets—have dramatically improved AI performance across domains, often surpassing specialized approaches without architectural innovation. Rish examines the mathematical "scaling laws" that predict model performance based on compute, data, and model size, while also addressing emergent behaviors, phase transitions, and the balance between caution and progress in developing increasingly capable AI systems.