Aleksandra (Saška) Mojsilović believes AI can be a positive force for the world.
The IBM Fellow is Head of AI Foundations at IBM Research and Co-Director of IBM Science for Social Good, so she knows a few things about applying AI to humanitarian problems. We caught up with her to talk about machine learning, COVID-19 and collaboration…
Q: AI has incredible potential to help address the many challenges on our planet. Can you share some recent, real-world examples that inspire you?
A: The idea of our program is to work with organizations on the forefront of challenges like poverty, hunger and other injustices, to figure out where AI and technological data can make a difference. They really inspire me to think about what’s possible.
One of the projects we’ve been working on is how AI can help the opioid crisis, one of the largest health issues here [in the US]. When people get addicted to opioids, it usually happens in a benign way: [patients] have a pain incident or surgery, get a bottle of pills and are largely unguided… so take more than what’s needed and become addicted. So, we are studying healthcare data, prescription data, and use machine learning to tease out stories of addiction and understand which prescribing behaviours are risky and what type of populations are more vulnerable in order to come up with better prescription guidelines, create health care policies and identify vulnerable individuals.
Another example is that we are working with CityLinks Center in Cincinnati, OH, an organization administering social services to low-income individuals. They take a holistic approach, so we are applying machine learning to the success of their interventions by analyzing years of interventional data in order to understand which services or combination of services are most effective in uplifting people to help them get through their programs.
See how Saška and IBM give back – with science.
Q: That’s incredible. What AI applications have you seen fighting COVID-19?
A: Everyone knows how AI models can generate things like images, music or fake news, but it is possible to teach AI to create things that are more useful. So right now, we are also looking at molecules and compounds… Can we train AI to create antibiotics or generate a framework that creates new drugs? When COVID-19 happened, we asked: Can we apply AI to COVID-19 and create molecules that could be candidates for a cure?
We applied AI to create new molecules that will be candidates for a covid drug, which might significantly accelerate the process [to finding a vaccine] since the drug discovery pipelines are very long and costly. If we can create a really good candidate that has a high probability not to fail in a chemical trial, you can compress that pipeline enormously… So, we recently created 3,000 completely new molecules and put them in a visualization tool so researchers can touch them, understand their properties and figure out which are most likely to succeed.
Q: What are future applications of AI in global health?
A: We’ve been deeply looking into fighting and managing pandemics for quite some time and although we didn’t predict COVID-19, we had a range of AI projects that examined how we can accelerate ways to predict or pinpoint disease before it happens, how we can better fight it and how we can accelerate finding cures. We’re also looking at drug repurposing. It turns out that there are many generic drugs that can be repurposed, so can we train AI to read scientific publications and find drugs that have potential to be reused for something else?
Another project that I find really cool is that COVID-19, like most other modern diseases, essentially originated in animals and jumps into humans. We have a project that was started in the time of the Zika virus to understand the transmission mechanism… We’re looking to train AI to learn about all of the viruses in animals that we know of, in order to predict other animals that are possible carriers or have the potential to create disease… Ideally, we want to use AI to fight epidemics before they occur. These are the kinds of wonderful applications of AI that lends itself to be helpful, as opposed to writing fake news…
Q: It seems like this work requires a huge amount of cross-industry collaboration. What have you learned about fostering that atmosphere?
A: [Collaboration is] really the best way to learn things, especially to learn something complex. With our NGO partners, there are so many things that we don’t know that they know, and things that they might not know about technology that we do. So when we work with each other, listen to each other, we learn a lot and, in doing so, come up with something that each party on its own would never have been able to come up with.
Being able to foster diverse points of view and come together in a completely new way is where the best innovation happens. If I look at the work that we did in our program, it is not so innovative because we are great scientists, but because we sat down and listened and then asked questions about what would help our NGOs — we cannot come up with recommendations unless we really understand the problems. AI is part of this complex discovery system, so it’s important to understand the bigger picture, have diverse thoughts and multiple points of views. When we cultivate research areas and it’s not just white, male scientists, we have a better chance of creating solutions, not just systems. This way we can be more innovative, out-of-the-box and inclusive. When we talk about this, it’s not some sort of philanthropy: we actually created assets and tech that our company can leverage in so many ways. I think it’s an argument for all of us to think about: that when we focus on doing good, everyone benefits and the return is spectacular.
This interview has been edited and condensed for length.
Hear Saška speak about Science for Social Good at the at the IBM Think Digital Event Experience May 5-6. #Think2020
Questions or comments? Drop us a line at firstname.lastname@example.org