More Microsoft News at https://dWeb.News Tech
The COVID-19 pandemic has upended every aspect of society and, at the same time, the urgent need for a response has spurred collaboration, innovation and creativity. Microsoft has created a Resilience series to highlight some of the research projects, technologies and ideas that emerged from the pandemic and the people who made them possible. To understand how this pandemic has affected the research community, we talked with Peter Lee, head of Microsoft Research and Incubations. Here is an excerpt of that conversation. Learn more about societal resilience as an emerging research area on the MSR blog.
I’ll start with a big question. What have you seen over the past year-plus, both within Microsoft Research and in the scientific community as a whole?
Well, one thing is that when the pandemic started, researchers dropped what they were doing and worked on problems of urgent priority. We had this sudden realization that we needed to put things aside and marshal a response for the world’s science and technology needs. Here at Microsoft, we immediately started a hackathon — a special hackathon just for COVID-19 response — and we had over 1,100 people from across the company put aside what they were doing and respond. I personally found it incredibly moving just to see the outpouring of concern and direct response by my colleagues here.
Maybe we thought they would be summer projects last summer and then August passed, and we are still working on these things today. And I think every major organization is thinking about this idea of societal resilience and how our research fits into that.
And something that has emerged from Microsoft and for researchers generally is that a lot of the ideas and technologies and science that we possessed prior to the pandemic got seriously pressure tested. For example, the whole idea of using messenger RNA as a foundational platform for vaccines — that was a theoretical concept for a couple of decades. And scientists at a lot of great research institutions thought in principle it should work and the normal course of science and technology development was proceeding. But then during the pandemic there was a need to pressure test: Are these things for real or not? And sometimes they worked and sometimes they didn’t. And learning what things that didn’t work or didn’t work as well can be just as important.
What are some things that didn’t work as well?
Well, there is this theme about data and the knowledge from the data, and yet there were some things that we just couldn’t answer. Really simple things. Like, what is the capacity of each hospital and clinic around the world to treat COVID-19 patients and what percentage of that capacity will be used up over the coming weeks. For about the first four or five months of the pandemic, we couldn’t answer that really basic question, despite all the digital data that we had and all the analytics.
So, one of the big things that happened, with the work at Microsoft and other places, is building new types of models and dashboards that give you some predictive ability to fill those gaps. In computer science we are all about predicting the future. And we want also to predict the impact of various interventions. If we impose an 8 p.m. curfew on all bars, what will that do? If we ban large sporting events, what would that do?
Then there is the test of the idea of real-world evidence. When you’re doing research that is contributing to our knowledge in medicine, new ways to treat people or diagnose people’s illnesses, the standard way we do that is through a randomized control trial. You do a scientific experiment where you try to control for what computer scientists would call confounders. You’re trying to derive cause and effect as opposed to just statistical correlations. It’s the gold standard for scientific discovery in medicine.
But there’s this idea of real-world evidence, where you observe every doctor treating every patient and just record it all. Now, you use all that data, this real-world evidence, and you do data mining to answer any medical question you want. This is a concept that is incredibly alluring in an era of big data, when we have all of these magnificent data analysis problems and capabilities.
As a computer scientist, before the pandemic, I thought real-world evidence was straightforward. If we just had all the data. What I learned this year is, wow, the problem is a lot harder.
During the pandemic the largest early access protocol in U.S. medical history was authorized for antibody plasma. Over 100,000 people enrolled. You would think observing what happened to those 100,000 people we would be able to tell if plasma worked or not. But even though the use of antibody plasma has a 100-year-old history in medicine, determining its efficacy for COVID-19 turned out to be much more difficult than anticipated. We learned many things, including that we lacked some important machine learning and data science tools. And another serious complication is that all this work was happening at light speed and under a very public and sometimes political spotlight.
So researchers were not only trying to work quickly but also in the spotlight.
Right. This is a different kind of science we’re doing. The normal paradigm of scientific discovery is we work on problems, publish papers, meet at conferences and debate. Then scientists around the world work on these things, replicate the results in clinical trials and debate. This happens over the course of five to 10 years. For the pandemic, we tried to do this all in one year this year. It’s a new paradigm. I use the term “crisis response science.”
Newton and Einstein weren’t brought in to save the world. There is something different going on. And what we’re finding is it wasn’t just the pandemic. For climate calamities, like the wildfires in California and Australia, researchers and scientists were brought in. I mean, heck, if an asteroid were about to hit Earth, who would you call? Yeah, you’re going to call scientists now, which is such a different paradigm for scientists.
And there’s this question of why are scientists and researchers of all stripes so important in crises? And it’s because of their natural inclination for dedication to the truth. Scientists have had to develop a culture and a mindset where your career might depend on this experiment turning out a certain way. But if it doesn’t turn out that way, you are still absolutely duty bound to publish the results that maybe show that your hypothesis is just not true. That authenticity ends up being very important when you are trying to respond to a crisis that can have huge impact and disruption on people and on structures like political parties or governments or power structures. For example, we learned in the antibody plasma effort that it is a big deal to ask people who have just gone through the difficult experience of recovering from COVID-19 to now turn around and do another difficult thing, which is to donate plasma.
Microsoft is launching a series about resilience to highlight a few of the people and efforts that came out of the past year. You mentioned the predictive dashboards. Can you talk about a few other things that went well?
Sure. One other thing in regard to data is that the amount of scientific research and papers produced during the pandemic was just overwhelming. It swamps any ability for human beings to absorb the information. So early on, Microsoft helped create CORD-19 (the COVID-19 Open Research Dataset). And then just a couple of months ago, Microsoft built on that to launch Microsoft Biomedical Search, which allows searchers to use natural language queries. And the use of AI to smartly synthesize the vast amounts of data was just so important, and it’s just amazing how much got accomplished there.
I was part of The Fight Is in Us, and it involved big tech companies, health-care institutions, nonprofit blood centers, for-profit pharmaceutical companies and media companies. And they all came together, meeting every week, and they’re still meeting, in order to come to a common understanding of the hard problems to work on and find places to collaborate.
You talked in the beginning of our conversation about this new research area of societal resilience and how jobs are being created to address what will certainly be a permanent part of the field. Can you say more about Microsoft’s research role going forward?
Well, one thing that’s really been important for Microsoft specifically is helping people do their work, whether as a student in school or information worker or on a factory floor. And right now, it’s really important to Microsoft to understand what is work going to be like. We don’t really know what the future of work will be like.
Prior to the pandemic, I felt like we had that pretty nailed. We could have a lot of science to design buildings and conference rooms that worked a certain way. And we understood about trying to accommodate all sorts of things for people to be well supported in their work. All of that was really based on years of experience and some of it on a lot of research to really understand. Well, all of that’s up in the air now. So what’s happening is there’s been a reenergizing of all that research in these areas. There was research that was really big 10 or 25 years ago and it’s all coming back now. Within Microsoft we just have this renewed emphasis on future of work.
Why is it so important for Microsoft to invest in research?
Microsoft is in a very special place right now because we’re out here to help everybody be better and more successful. Our business model depends on less division in the world, more equitable access to technology and to good jobs, nations getting along. The more all that is true, the more our business goes up. I don’t think that was planned, but that’s where we are. That has put all these charitable aspirations in complete alignment with our business model.
And so we go back to what we thought would be those summer projects last year and, as we recognize the permanence of these things, we have decided to create new job descriptions, new roles, new organizational structures around this idea of societal resilience — long-term careers where people at Microsoft can earn a living and build a career focused on these things.
Read more about societal resilience as an emerging research area and learn what you can do on the MSR blog.