AI for Good Projects Need a Helping Hand

(SvetaZi_/Shutterstock)

The almighty dollar is a powerful factor in the current wave of artificial intelligence (AI) adoption. And why shouldn’t it be? For millennia, companies have relied on technological progress to grow sales, cut costs, and improve customer satisfaction. But if we take a wider view, we see there is tremendous potential for AI to benefit society as a whole. Unfortunately, these “AI for good” projects often face big obstacles to success.

Ever since big data became a “thing” over a decade ago, we’ve heard how for-profit companies have worked with non-profit organizations to find novel ways to use emerging tech for the benefit of society. Whether it’s mapping the spread of Ebola in the 2014 outbreak, analyzing patterns of student engagement in Indian schools, or tracking down human trafficking rings in the United States, we’ve seen how emerging technologies and AI can be used to help those in the greatest need.

But there’s a problem, which is that AI is difficult. At every step in the process – from assembling the data team and wrangling the data, to tuning the model and putting it into production — there are significant hurdles to overcome. If giant corporations that throw off billions in net income are struggling with their own AI initiatives, then why would a non-profit organization with a fraction of the resources find the AI going easy?

Many technology providers in the AI and big data community have devoted a portion of their revenue to charitable organizations. Often dubbed “AI for good” or “big data for good,” these initiatives provide non-profits with the ability to get software and services for free or at a steep discount.

DataRobot is hoping to take a slightly different approach with its new program, dubbed AI for Good Powered by DataRobot, which it launched in July. According to the program’s leader, Chandler McCann, organizations that are accepted into the program will have access to the same resources as paying customers.

“We are treating all of these non-profits that are participating in the program just like our normal customers,” McCann says. “That includes helping them frame their problem well, so understanding what it’s going to solve, and clarifying what they’re trying to do with machine learning. And providing our automation software to enable them to work on the problem, as well as maintain it over time.”

Much of the work of building the machine learning model is automated by DataRobot’s AutoML solution, so data science skills are not as important as they may otherwise be. “Building models is not the long pole in the tent,” McCann says. “It’s very easy to build models with automation in this space.”

Bigger challenges exist at the beginnings and the ends of the engagements. At the beginning, proper framing of the problem is critical to ensure that it can be addressed with data and algorithms. Not every problem is amenable to machine learning, McCann points out.

On the back end, maintaining the model over a long period of time, and periodically retraining it on new data, is also critical to ensure the model continues to deliver predictive value. The company will provide education and training resources to help the non-profit organization achieve success.

“The takeaway is that machine learning projects can be hard,” McCann says. “That’s why at DataRobot, we built an entire success organization to help organizations clarify what they want to do. A model is a living thing. If we’re building a model and deploying it, it has a life. It needs to be retrained over time, monitored, as well as really understanding the inputs of where is this data coming from, and who’s going to be using this prediction on the end.”

DataRobot helped predict breaks in the water delivery system in Africa for The Global Water Challenge

While McCann praised all tech vendors’ attempts to help non-profit organizations, it’s not enough to simply give them licenses to software that would ordinarily cost tens of thousands of dollars. Some non-profits have expressed a wariness of big data software companies essentially throwing software over the fence and leaving the organization to make use of it.

“From where I sit, I think it’s important to bring a holistic approach, end to end,” McCann says. “The idea is to make them sustainable.”

While it just launched the program last month, the Boston, Massachusetts company has already worked on a handful of AI for good projects. One of those was to help improve access to clean drinking water for Africans, in which it collaborated with Global Water Challenge to analyze half a million data points to identify places where the water delivery system is at risk of failure.

McCann has two full-time data scientists who are dedicated to the AI for good campaign, which DataRobot crafted with the help of GlobalGiving, a Washington D.C.-based organization that helps to coordinate and connect non-profit groups with donors and charitable companies around the world.

DataRobot is currently accepting applications for the program, and has pledged to collaborate with up to 10 organizations this year. You can read more about the program and submit an application at www.datarobot.com/ai-for-good-application. The deadline is today.

Related Items:

Big Data Yields Progress in Social Good

How Data Analytics Is Helping to Fight Human Trafficking

Hunting Down Ebola with Big Data