

AlmaLinks Tel Aviv hosted Prof. Zachary Lipton to summarize where we really are in AI research today. Zachary Chase Lipton is an assistant professor at Carnegie Mellon University. His research spans both core machine learning methods and their social impact, with a concentration on deep learning for time series data and sequential decision-making. His work addresses diverse applications, including medical diagnoses, dialogue systems, and product recommendation. Below are the main highlights from the event:
- Deep learning is a powerful tool that has unlocked new capabilities in pattern recognition. We can use statistical regularities in datasets to build tools for language translation, speech recognition, computer vision, demand forecasting, and assisting in medical diagnostics.
- While these capabilities are real and exciting, they are not magic: the picture painted by many companies and the media typically contains fundamental misconceptions about the limitations of these systems.
- Big challenges include (i) guaranteeing the performance of a system even when the environment changes in small ways, e.g. that an X-Ray reader should still work even if the equipment is upgraded to a new machine that produces qualitatively similar but superficially different images; and (ii) making the leap from learning associations to causal interventions. Truly intelligent systems should reason about what would happen if we made different decisions.