Date & Time:
January 13, 2022 11:00 am – 12:00 pm
Location:
Zoom
01/13/2022 11:00 AM 01/13/2022 12:00 PM America/Chicago Kush Bhatia (UC Berkeley) – Learning When Objectives Are Hard to Specify TTIC Talks at TTIC Zoom

Learning When Objectives Are Hard to Specify

Register for Zoom details

Real world deployment of learning systems which interact with humans requires aligning what these systems optimize for with underlying human objectives and values. A major hurdle towards accomplishing this has been that it is hard for humans to precisely specify what it means to do the desired task well. 

In the first part, we will take a multi-criteria viewpoint towards these underlying objectives and develop a framework for selection of such value-aligned models when the data comprises pairwise comparisons across multiple different criteria. In the second part, I will present our work on understanding the consequence of over optimizing a misspecified objective. We find evidence about the existence of phase transitions which could pose challenges to safe deployment of such learning systems.

Host: Nati Srebro

Speakers

Card Image d44d702

Kush Bhatia

Ph.D. Candidate, University of California, Berkeley

I am a graduate student in the Computer Science department at UC Berkeley where I am fortunate to be advised by Peter Bartlett and Anca Dragan. I am interested in problems at the intersection of statistics, optimization and machine learning, in particular on aspects of aligning the objectives of the ML models with those of humans.

Before coming to Berkeley, I spent 2 wonderful years working at Microsoft Research Bangalore with Prateek Jain and Manik Varma. I completed my undergraduate studies at IIT Delhi with a major in Computer Science where I was advised by Parag Singla for my undergraduate thesis.

Related News & Events

Card Image 9fed1fd
Card Image
UChicago CS News

Five UChicago CS students named to Siebel Scholars Class of 2024

Card Image 31e102b
Card Image
UChicago CS News

UChicago Computer Scientists Bring in Generative Neural Networks to Stop Real-Time Video From Lagging

Card Image bbcfe7e
Card Image
UChicago CS News

UChicago Team Wins The NIH Long COVID Computational Challenge

Card Image 320bc3e
Card Image
UChicago CS News

UChicago Assistant Professor Raul Castro Fernandez Receives 2023 ACM SIGMOD Test-of-Time Award

Card Image 096db1e
Card Image
UChicago CS News

Computer Science Displays Catch Attention at MSI’s Annual Robot Block Party

Card Image 0c81085
Card Image
UChicago CS News

Professor Heather Zheng Named ACM Fellow

Card Image 92588f2
Card Image
UChicago CS News

Q&A: Ian Foster on Receiving the 2023 IEEE Internet Award

Card Image 7a306ff
Card Image
Video

Ian Foster – Better Information Faster: Programming the Continuum

Card Image 6b6d1a2
Card Image
UChicago CS News

Professor Fred Chong Named IEEE Fellow

Card Image 6e4eb5c
Card Image
UChicago CS News

Associate Professor Diana Franklin Named ACM Distinguished Member

Card Image fe821b3
Card Image
UChicago CS News

UChicago’s Parsl Project Pivots to Sustainability and Community with New Grants

Card Image 3afe375
Card Image
UChicago CS News

Trending Now: How Netflix Chills Our Free Will

arrow-down-largearrow-left-largearrow-right-large-greyarrow-right-large-yellowarrow-right-largearrow-right-smallbutton-arrowclosedocumentfacebookfacet-arrow-down-whitefacet-arrow-downPage 1CheckedCheckedicon-apple-t5backgroundLayer 1icon-google-t5icon-office365-t5icon-outlook-t5backgroundLayer 1icon-outlookcom-t5backgroundLayer 1icon-yahoo-t5backgroundLayer 1internal-yellowinternalintranetlinkedinlinkoutpauseplaypresentationsearch-bluesearchshareslider-arrow-nextslider-arrow-prevtwittervideoyoutube