Giving AI Some Common Sense

Ron Brachman, visiting professor at Cornell University, presented his Distinguished Lecturer seminar "Giving AI Some Common Sense" on Wednesday, April 24, 2024 at Northeastern University's Curry Student Center (Room 433).

ABSTRACT

With AI’s recent leaps forward, there is immense excitement about its potential and its burgeoning application. But despite demonstrating some amazing capabilities, AI systems continue to make unexpected, unhumanlike blunders, making their behavior worrisomely unpredictable and undermining our trust in them in real-world situations. Surprising gaffes by chatbots and bizarre actions by “self-driving” cars reveal that they just don’t have what we would call common sense.

So how can we fix this problem? In this talk, Ron Brachman shows how we might take some first steps. He begins by examining what it means to have common sense, exploring the role it plays in cognition. Brachman illustrates when and how common sense comes into play and will talk about how to start building a common sense capability, based in part of work that has been underway in AI for a number of years, and on a novel bigger-picture approach to the architecture of an artificial cognitive system.

The talk concludes with some discussion about the criticality of common sense in two key challenges: the intelligibility of intentions and actions in AI systems, and the crucial ability – noticeably lacking in current systems – to take advice and act on it. Brachman makes the case that AI systems without common sense should never be allowed to act autonomously in the world, and that AI still has a rich set of under-addressed research problems that it needs to tackle. The talk is not technical and uses numerous easily-understood examples from the everyday world.

Flip through Brachman's slides here:

‍

BIO

Ron Brachman is an internationally known expert in Artificial Intelligence. He recently retired as a Professor and Director of the Jacobs Technion-Cornell Institute at Cornell Tech and is now a Visiting Professor at Cornell University. He is currently working with Schmidt Sciences on new AI programs and serves on advisory boards for Nokia Bell Labs and Laer AI, among others. Previously, he was Chief Scientist of Yahoo and Head of Yahoo Labs, and served as the Director of the Information Processing Technology Office at DARPA, where he created the program that gave birth to Siri. Earlier, he held research leadership positions at AT&T Labs, AT&T Bell Laboratories, Fairchild/Schlumberger, and BBN.  

Brachman holds a B.S.E.E. degree from Princeton University and S.M. and Ph.D. degrees from Harvard University. He has served as President of the Association for the Advancement of Artificial Intelligence (AAAI), as Secretary-Treasurer of the International Joint Conferences on Artificial Intelligence (IJCAI), and as Board member and Treasurer of the Computing Research Association (CRA). He has won multiple awards and is a Fellow of the ACM, IEEE, AAAI, and AAAS. Brachman has published numerous technical papers and is a co-author of an important AI textbook on knowledge representation and reasoning. His most recent book (with Hector Levesque), Machines like Us: Toward AI with Common Sense, was published by MIT Press.