Daphne Koller is a pioneer in several aspects of artificial intelligence (AI) and machine learning. She also helped advance online education in its early years by co-founding worldwide learning platform Coursera in 2012. We talked to her about all of these things and how they will impact the post-COVID-19 world.
Daphne Koller
Founder and CEO, Insitro and Co-Founder, Coursera
In light of COVID-19, what do you foresee happening to traditional education as we know it?
What we’re seeing right now is a tremendous acceleration of the move to online and hybrid instruction, which is going to be critical in the fall and probably well into even the remainder of the next academic year. When we have some way of treating or preventing COVID-19, there will be more of a return to face-to-face instruction.
I think that because of the need to develop, for the intervening year or more, a better format for the online education space, many will realize that face-to-face instruction was certainly not the optimal or only way of teaching. And as a consequence, as part of the development of new and better teaching methodologies that utilize this hybrid format, much of that will remain even after face-to-face teaching is more possible.
So I think we’ll see at the beginning, as we’re seeing right now, a very significant switch to online that will remain throughout most of the academic year next year. And then it might revert back to a certain extent but not all the way back to where we are today.
Much of your research is focused on the potential of artificial intelligence, and how we can use that to better in biomedical engineering and sciences. What potential Do you think AI has to improve biomedical sciences and find cures for diseases with which we’re still struggling?
Much of what we see around those has been transformed by engineers and scientists applying machine learning and AI techniques to ever increasing data sets. And that transformation has not yet happened in the biomedical sciences, partly because they’re much more complex than almost anything else we do. They deal with natural systems that are much more subtle, and had a poor understanding of the engineered systems we typically end up dealing with, where we kind of get to write the rules.
Here, we didn’t get to write the rules; evolution wrote the rules and we know they’re very, very complex and partly because, up until recently, there have not really been the kind of tools that allow us to measure or create biological data at the right scale to drive machine learning.
That has changed in the past few years with the development of an increasing range of incredibly valuable tools that can help us engineer large numbers of biological experiments, and measure their output in a scalable way so we can get the kinds of datasets that will drive machine learning methods, and help elucidate biology and identify interventions that can make a meaningful difference.
I think that over the coming years we will see an increasing number of therapies that are very much inspired by and use the tools of data science to identify the right patient populations, or even to identify the target or design the drug. And over time, we’ll create even more of those tools that help create value throughout the farm or in the value chain.
Given the realizations we’ve made as a result of COVID-19, how far off would you say major advancements in these fields are?
The use of machine learning to help improve patient diagnostics in relatively circumscribed tasks like radiology or pathology, that trend already began before COVID-19 and will presumably be accelerated by that.
Things that are much more challenging like designing a new drug for Alzheimer’s disease, who knows? That’s a much further-out prospect, and hopefully we eventually get there.
There are multiple ways in which machine learning data science and AI can contribute to human healthcare, and some of those will happen much earlier than others.
We know AI has plenty of applications for good, but we also know hackers and other malicious actors can use it as well. How do you foresee AI impacting societies?
So first of all, let me just say that any powerful technology can be deployed for both good and evil: nuclear reactors or nuclear bombs; CRISPR is a therapeutic or research tool, but can also be used to design virulent pathogens. I mean, almost anything, you can say, has a good side and a bad side.
I think AI has already demonstrated multiple ways in which it can be deployed for just bettering everyday life. I mean, every time that we go on to the web and search for something on Google, that uses AI underneath the hood. The kind of speech recognition that can be helpful to all of us, but also to people with disabilities, it’s incredibly valuable.
Automated driving — even if we don’t get to full Level-5 automated driving — but just the way in which cars can help protect us from inadvertent accidents. All of these use AI techniques, so I think the list is very, very long.