Earlier in November, we sat down with representatives from the UNICEF Innovation Fund and Thinking Machines to discuss how they work together to implement “tech for good” in the Philippines.
For Stephanie Sy, founder and CEO of Thinking Machines—a UNICEF Innovation Fund portfolio company—her company is building AI and data systems to help organizations in the Philippines make smarter decisions. In recent years, they have collaborated with UNICEF and the Asian Development Bank to formulate a new understanding about poverty in the slums of the Philippines using geospatial data analysis.
The following interview has been edited and consolidated for brevity and clarity.
Oasis (OS): Tell us a bit about yourself and your journey within tech. How did you end up founding Thinking Machines?
Stephanie Sy (SS): I started my career as a technologist in San Francisco, working for a few startups and Google. After that, I decided to come back to the Philippines, where I grew up, to start a data science company. In traditional industries, you need a lot of capital to found a business. But technology is the one space where you can build companies faster and change a lot of lives if you have smart people, computers, and internet access.
The Philippines is a country rich with human capital. Back then, there wasn’t much of a data science industry and I figured I could give it a go. In the last six years, we have built a lot of tools and helped many organizations in the government and private sectors to embrace decision-making processes that utilize AI and big data.
OS: Nowadays, many organizations are trying to implement artificial intelligence and machine learning solutions. How has the Philippines embraced the two technologies?
SS: AI and machine learning have been buzzwords in the last seven years. But the problem is that a lot of executives don’t realize that data solutions are like an iceberg: AI and machine learning is the shiniest piece of it, while 90% of the work, like data mining, cleaning, and processing, is underwater. It involves matters like where the data comes from, where the data should be stored, and how to process the data.
What we want to do is empower organizations to make better decisions through the application of AI and big data. For example, we give the staff of the Philippine Department of Education and Health tools and training to process a lot of data quickly.
OS: What are some unique insights that you’ve gleaned on tech for development in Southeast Asia?
SS: We still have a huge talent gap. We don’t have many places where people can learn how to be great data scientists. There are two reasons for this. R&D in academia is still pretty weak here. And looking at the industry side, where is our Google? Where do smart fresh graduates go to learn how to be a great data professional? These are still nascent in Southeast Asia.
What does exist here is data preprocessing. A lot of American companies outsource their data labeling work, which involves manually tagging data and training a machine to recognize, say, a fire hydrant. In a lot of cases, these tasks are outsourced to the Philippines, India, Vietnam, or Malaysia. This doesn’t help the countries develop their own AI talents or an AI industry, because the work of identifying raw data is limited and it does not help you to become a data scientist.
OS: What are some projects that you’re working on in the Philippines?
SS: With UNICEF and the Asian Development Bank, we do a lot of work to develop open source data sets, models, and geospatial analytics tools for the public to use and improve. We get to see people take these resources and build their own tools on top of it.
Whenever Thinking Machines works with a government institution, we strongly encourage them to make the data that they collected available to the public. For instance, the Department of Health in the Philippines open-sourced their national COVID-19 testing data and updates it on a daily basis, which is amazing. Government departments like the National Economic Development Authority of the Philippines and the Department of Budget and Management are also more open to the idea of opening data to the wider world than they were five years ago.
OS: What’s an important point that very few people agree with but you believe in?
SS: I strongly feel that machine learning models are not going to be able to deliver the kinds of fully automated experiences that people hope for. People expect to pass off hard decisions to machine learning models. The problem is that machine learning models only mimic the patterns and learn whatever you show them. If you present a biased set of judgment calls, they will give you predictions that are biased in the same way.
So, a data scientist needs to be ethical when they utilize the final data.
This article was produced in collaboration with the UNICEF Innovation Fund, which provides early-stage funding and support to frontier technology solutions that benefit children and everyone else in the world.