Seminar by Balachander Krishnamurthy, lead inventive scientist at AT&T Labs Research will present the seminar
Data: 14 APRILE 2023 dalle 11:00 alle 13:00
Luogo: Aula Kelsen, Via Galliera 3, Bologna
By now almost everyone has heard of ML disasters. They occur often and unfailingly receive press coverage every few months. This is not new in computer science: it happened with security incidents 2 decades ago (and continues to this day) and then with privacy violations (over a decade ago and continuing today).
But ML related problems, especially bias, is more insidious, harder to detect, and much harder to fix. Computer scientists and statisticians take full advantage of the large amount of data available today, fast processing capabilities, and apply a variety of clever algorithms to generate predictions in a number of arenas. When they are related to beating chess and go champions, there are few negative consequences (except to the hitherto human champions). However, when ML is used in self-driving cars, replacing Human Resources, facial recognition etc., the risks are higher.
In this talk I will cover some well-known ML bias incidents and enumerate the reasons behind them. I suggest ways by which AI governance with the pillars of technology, harnessing non-technical governance talent, and training can help reduce (not eliminate) the risks. No background is expected for the material presented.
---
Balachander Krishnamurthy is a lead inventive scientist at AT&T Labs Research. He is one of 66 AT&T Technical Fellows in the company's 146-year history – the highest honor given by the company to their technical employees. His focus of research is in the areas of fairness in Machine Learning, AI Governance, Internet privacy, and Internet measurements. He has authored and edited ten books, published over one hundred and thirty technical papers, holds eighty-five patents, and has given invited talks in thirty-five countries. He co-founded the successful ACM Internet Measurement Conference in 2000 and the Conference on Online Social Networks. He has been on the thesis committee of several PhD students, collaborated with over eighty researchers worldwide, and given tutorials at several industrial sites and conferences.