From 3M Health Information Systems
AI Talk: AI for social good
This week’s AI Talk…
I heard a talk at the 2020 International Conference on Machine Learning a few weeks back on the topic of “Doing Some Good with Machine Learning.” This talk was given by Dr. Lester Mackey, who got his PhD in Computer Science from UC Berkeley and did post graduate work at Stanford. He is currently a researcher at Microsoft. AI for social good has always been an area of interest to me. I decided to dig in this week and do a themed blog on this topic! I’ll start with the presentation from Dr. Mackey.
Doing some good with machine learning
Dr. Mackey’s talk was mostly autobiographical. He sketched out his decade-long effort to focus statistical and machine learning (ML) techniques on social causes. He outlined his journey from his formative years at UC Berkeley to his tenure at Stanford where he picked up statistics, machine learning and beyond. In addition to his individual commitment to apply machine learning to good causes, I was impressed and enlightened to hear about numerous organizations that have similar goals. Below are some of the organizations and resources mentioned in the talk:
- DataKind —Their tag line: “Harnessing the power of data science in the service of humanity.” They have a DataCorps that brings together pro bono data scientists in focused teams to solve specific social problems using data science and AI.
- Data Science for Social Good — a summer internship program with the University of Chicago that focuses on teaching machine learning skills to aspiring students interested in data science projects with social impact. Similar programs are now available at CMU and other institutions.
- Statistics for Social Good — a working group established by Dr. Mackey at Stanford to help organize data science projects and solutions there.
- Data Science for Social Good — a Google group which brings together anyone interested in promoting social good using data science.
- Statistics for Social Change — a github data repository for a variety of public data sources related to social issues along with a list of partner organizations.
- Harvard Dataverse — a repository of research data, the depth and breadth of which blew my mind! It has over 99,000 datasets for all kinds of problems! In response to the current pandemic, this repository has all sorts of Covid-19 related datasets.
- COVIDcast — a site that predicts COVID-19 prevalence, developed by CMU’s Delphi group in which Dr. Mackey participated. The Delphi group is an epidemiological forecasting institute that has been involved in forecasting the spread of influenza.
- Statistics Without Borders — a group within the American Statistical Association that is bringing together pro bono help in applying statistics and data science to social problems.
- Solve for Good — a platform where problems that need to be solved can be posted and addressed.
- Data science contests — there are a variety of platforms for posting and addressing data science challenges. Innocentive, Drivendata, Dream Challenges and Kaggle.
Dr. Mackey ended the talk with a call to action that covered four precise themes:
- Teach more ML for good (academic schools)
- Publish more ML for good (all researchers)
- Incentivize more ML for good (call to corporations to promote pro bono work)
- Prioritize more ML for good (everyone)
Wadhvani Institute for Artificial Intelligence
I would be remiss if I don’t point out the work done by this institute in India. Its CEO, Dr. Anandan, is my former classmate and close friend. And I worked for Romesh Wadhvani, one of the institute’s founders, three decades ago in Pittsburgh! Wadhvani Institute’s explicit goal: AI for social good. They are funded in part by Google and the Bill Gates Foundation. The roster of projects they have undertaken shows their commitment to the cause of social good. They have developed deep learning models that detect pest infestations using pictures taken from cotton crop. Timely detection leads to corrective action. Another application they are in the process of fielding is screening for low birth weight babies. India is one of the hotspots for low birth weight babies. They use anthropometry technology to do this screening. Anthropometry? I had to look it up! It is the science of measuring the dimensions of the human body, mostly used in anthropology. Using a smart phone, they take the dimensions of the baby and predict the birth weight (using AI), geo-tagged for location and transmitted to a server. Of course, given the current pandemic, the institute is developing solutions surrounding Covid-19 as well. All in all, it’s an impressive effort to do good.
I am heartened to see so many organizations focused on what is good for society. My faith in humanity is restored. Perhaps it’s time for me to do my small part.
I am always looking for feedback and if you would like me to cover a story, please let me know. “See something, say something!” Leave me a comment below or ask a question on my blogger profile page.
V. “Juggy” Jagannathan, PhD, is Director of Research for 3M M*Modal and is an AI Evangelist with four decades of experience in AI and Computer Science research.