Perspectives on AI and Cloud in the Fight Against COVID-19

Dr. David Shaywitz, Harvard-trained MD, shares his knowledge surrounding the use of AI in the spread of the virus and vaccine development in this Tech Barometer podcast segment.

By Jason Lopez

By Jason Lopez March 24, 2021

In the days after COVID-19 officially became a pandemic, curious minds wanted to know how new technologies could join the fight against the virus.

In this Tech Barometer podcast interview recorded in February 2020, Dr. David Shaywitz shared his insights about how AI and cloud computing can help humans tackle the complexities arising from the pandemic. 

Shaywitz studied medicine at Harvard and earned a Ph.D. in biology at MIT, and was recently a visiting scientist in the department of biomedical informatics at Harvard Medical School. He co-hosts the podcast on digital health called TechTonics and is the founder of Astounding HealthTech, which advises senior biopharma research and development leaders about digital and data.

The advancement and proliferation of big data and AI technologies are making it possible even for small organizations to sift through vast amounts of data and suggest options for solving problems across different industries. Big data and AI technologies helped the small software-as-a-service company BlueDot see the first warning signs more than a week before COVID-19 was officially identified by the World Health Organization.

He said BlueDot is an example of data technologies and human curation working together, similar to the approach taken by cancer data company Flatiron, acquired by Roche for $2.1B in 2018.

“Their algorithms distilled an extensive amount of news and other information into a couple of daily highlights, which then human curators sorted out to see what might be interesting,” he said.

RELATED

AI in the Cloud: Questions Many Aren’t Asking

He said telecare, telecommuting and tele-education will grow increasingly important in daily life around the world.

“Technologies are immensely powerful,” he said. “We need to come up with pragmatic use cases that deliver palpable benefits people can appreciate.”

He sees cloud computing making it possible for almost anyone to have unlimited, secure, safe compute and storage. 

“A few years ago, when I was chief medical officer at DNAnexus, it was clear the cloud was enormously empowering,” Shaywitz said. “It's taken a little while, but at least the healthcare ecosystem seems to have got the message – even five years ago, many medical centers were visibly anxious about moving the cloud.”

He said now almost everyone recognizes cloud computing is essential, easier to manage than ever before and less of a roadblock to research and analysis.

“It seems so clearly advantageous now that the discussion has moved onto the next series of challenges.”

RELATED

 Coronavirus Gives AI and Big Data Chance to Shine

Transcript (unedited):

Jason Lopez: A world health organization report in February about the steps being taken against the Coronavirus mentions where technologies like big data, AI and cloud are being used to understand and fight the Coronavirus and its spread from December through February 104 strains of covert 19 were isolated from people in China who had contracted it and the genome of the virus was sequenced, showing it was not SARS nor the flu, but a new disease and the Chinese government deployed big data and artificial intelligence to track the spread of the virus within the country and to help identify risks for various populations.

Jason Lopez: Researchers outside of China have been using machine learning to track the spread of the virus globally, sifting through huge amounts of data from news reports, social media and flight patterns just to name a few of the sources. Facebook is sharing data with Harvard researchers to help forecast the spread of the virus. And nd that's just an example of a few things that are going on behind the scenes to talk about technology and combating this pandemic. We turned to David Shaywitz cohost of tectonics, a podcast on digital health. He got his MD at Harvard, Ph.D. in biology from MIT and most recently was a visiting scientist in the department of biomedical informatics at Harvard medical school. And he joins us now from his home in Silicon Valley. Uh, so David just wanted to get some context on how technologies like AI and the cloud are being deployed, you know, from behind the scenes to deal with the Coronavirus.

David Shaywitz: Sure. Jason, I mean, one of the initial responses you could say, it's all the fact that we're so deep into this means that if the tech was so effective, we wouldn't be in this situation. So in some level that's cheating and kind of getting to the bottom line. I think what, uh, a pandemic or what this situation we're really kind of any stress has, is a real challenge to a lot of the hype around technology and, and demands us to see, know what's real and what's a promise and also highlights both the opportunities for technology to make a difference as well as some of the, the limitations. So, um, to sort of step through some of it. I think one of the first areas where technology, let's say AI has played potentially a role or at least had been highlighted, is playing a role is in the area of surveillance.

For example, there's a really interesting actually Toronto company. It's a, it's a B Corp company, Cornell called blue dot. That is a sort of infectious disease surveillance company. And there are multiple organizations and groups and companies that try to do something like this where it sort of monitors all of the world's news, um, and then tries to incorporate using a combination of manual curation and AI to, um, to try to highlight things that seem unusual and it should be flagged for attention. So for example, they were among the first in very late December to flag some unusual cases of pneumonia associated with a live market in a Wu Han China. Now, what I think is interesting about a blue.is there was actually founded by a physician entrepreneur, the ID doctor, who had sort of gotten through the SARS epidemic in Toronto just under two decades before. And really wanted to try to see if they could use technology who wasn't intrinsically, he said an entrepreneur.

He just wanted to really try to figure out how to apply technology to this. He wound up working in a fashion that reminds me very much of the data company. I'll flat iron, where thought iron is a, is sort of the ecology data company that provides almost regulatory grade data based on clinical information and what their whole thing was integrating both digital and data and all of that stuff with manual curation and in a very similar way. Blue dot highlights really the importance of manual curation and what they do with, they have the AI, no go do some of the translation obviously for Oreo, all the world's languages and they and really integrate a lot of information that it would take human being so long to go through, but ultimately it's sermons of making the judgments. They essentially tried to distill all the information into a kind of a couple of daily highlights which then human curators sort of go through and try to see what might be interesting.

In addition to that. Jason is even though they sort of, you know, reported or had this finding the impact of what they actually found on the world is at present. Unclear. So I think the surveillance is one area. Um, another area where it's, you know, people sort of highlight the potential is in, uh, drug development and, or vaccine development. And again, the idea is that it really, really difficult to make a drug from scratch. And almost everything that goes through the drug development pipeline as an idea does not come out the other end as a successful product. There's a number of reasons why generally very good ideas still don't have traction. You know, chemistry is incredibly complicated. Biology is really complicated. The human body's unimaginably complicated. Um, and so it really turns out they're just a series of problems to solve. But there are some thoughts that technology, you know, in AI and figuring out how protein folds, kind of understand some toxicology in some instances can potentially be modeled and certain steps may be able to be accelerated.

David Shaywitz: And so there are different companies that are sort of trying to work on that. But even if they sort of crack that in general, there are often so many other steps after that that even if, um, you know, the thing was the new novel coronavirus they just really figured it out what it was. It was amazing in the sense that they were able to get the genome sequence relatively soon out, you know, within a month or so of the thing that identifies and it takes a while to go from that. Even that knowledge. Okay, we have the genome that's great about this virus. How do you turn that into actionable information? Now, one company that may be doing this is, um, a company called Madrona, which is trying to make vaccines and they have sort of a proprietary AI platform that only under the DNREC don't really share a lot about that apparently tries to take with aspects of the, of, of a gene where that, uh, the protein encoded by a gene is most likely to be most useful for something like a vaccine. But again, we're not really going to know. It's going to take a while to figure out if this works, even though they're really moving almost unimaginably quickly. I mean, based on some of their work, which in turn was based on previous work because the new virus is similar to the African virus, they are actually expected to start clinical trials in a year. And that's sort of, I mean an imaginably fast for the accident.

Jason Lopez: Because of AI that they're able to move that quickly.

David Shaywitz: You know, it's really hard to tell. I imagine that they would say that they were able to pick out the specific app, which to know what part of the virus to pick out and use for a vaccine was as a consequence of, you know, algorithms and

Jason Lopez: I see, let's actually not call it AI, let's call it just big data, unstructured data, you know, that kind of thing. Right. How long have researchers been using big data and using analysis of unstructured data with computers to do research into genomes and viruses and bacteria? How long have they been actually using it? Or is it still sort of an experimental thing?

David Shaywitz: Right. So if it's sort of two parts of the question, the genomics community has been really having been on this really, really, really early and they've been fairly early adopters of cloud or at least some people have. But what confuses a lot of people about when you think about, uh, you know, large organizations like pharmaceutical companies, in particular, is, on the one hand, they haven't seen or data-driven organizations and they have a lot of data, but big data is sort of almost, it says I'm a little bit of a different meeting. Big data has as much more sort of both comprehensive and it's all sort of um, you know, generally for interlocking, whereas what you sort of wind up having it almost like the Minnesota model of data where it's like the land of a thousand lakes. So there's a little bit of a Lake here of data all over the Lake there and each of these lakes to push the metaphor or zealously guarded by whoever is sort of owned or responsible by that data.

I mean their data that's collected for a very specific purpose and it has intrinsic value in itself. But then the question is, can you sort of pulling it, is there value and how do you pull it together more effectively? And that's really where the industry is. Now. There's a general sense that my gosh, how we're handling data and all of these little silos doesn't make a lot of sense. And when it'd be great if we could, um, kind of pull it, pull this together in some useful way. And I think a lot of companies are struggling, you know, how to do that. And on the one hand, as you know, the idea of what's really exciting for me to see is how things like the cloud, which used to be are something that just made help people in particular really nervous. It was, it was viewed as, Oh no, we want to keep our data, you know, on our local server versus in the cloud somewhere. People finally recognize that in many cases, you know, for, for the kind of the major providers, in particular, the cloud is far more secure than um, uh, having the data in sort of, you know, some personal data room or something. So,

Jason Lopez: Well I imagine all the data that you were saying. You know, it's kind of a people thing here in terms of the 10,000 lakes idea. But imagine if they could actually even get deeper into like flight manifests and actually know where people are. Of course, now you're getting into the privacy issue. Boy, it's very, very complex.

David Shaywitz: It's complicated, right? And I think, I mean I think they do have actually de-identified flight information there and you know, they do try to uh, to understand this. If you look at it for all of this information, I was reading some article recently about how they're trying to track the so-called patient zero. The basic assumption is that there's a new coronavirus that originated from a back somewhere cause I think that's where all these viruses eventually come from. Someone basically eats a bat apparently or a bad infect something that somebody eats and it turns out there aren't any bats near that market. And Wu Han, the Nerf bat is like a relevant bat somehow, like a thousand miles away. So you know, people are, you know, going back and it seems, you know, the market was I guess in December, but they will keep her going back in November and just trying to show you that it's really hard to track the very beginnings of something, you know, in theory where you might've been able to, to stop something.

So it is really a, you know, what you're pointing out is so interesting because in the one that we have an, a population level is an incredibly rich information, but on an individual level, you know, tracking down this very specific type of thing can be really, really hard. And then how do you turn this into something actionable? Like so much of what I read about tech and AI is this could do this or it may do this or might do that. And it's, it's like a combination of this, an urgent need with some promise again of what's going to happen. And I think the technologies are immensely powerful. So I think trying to kind of tone down the rhetoric and really try to come up with some very pragmatic use cases that deliver a palpable benefit that people really can appreciate.

Jason Lopez: I guess one final line of questioning here is one final topic and that is touching on something that you brought up a little earlier, which is what can be learned. I mean the pandemic is on us. What do you think might happen as a result of all of the scrambling people are doing right now to do research and to implement a lot of technologies?

David Shaywitz: Absolutely. Well I think it's a great question. I think it is an opportunity, you know, exposure to stress both exposes the fragility and has a chance to, for the organizations, the people, the processes that can learn from it really helps robust defy them and make some sort of more and more able to deal with uh, adversity. Actually think one area where we're really going to see this, how we're going to be coping with the, you know, all the corn teams that are coming up. The two areas I'm thinking about involving technology, our Telecare and then all of the telecommuting stuff. So, and tele-education.

Jason Lopez: Yeah. You don't really know if something can really work unless you're forced to have to do it.

David Shaywitz: Let me make two points about that. Um, I mean one specifically or the data, but the first point is, you know, it's the famous quote various, they attributed to everyone from Caesar to a Mike Tyson that everybody has a strategy until they get punched in the face. Um, and I think at a certain level that that's true. But the flip of that, Caesar said that, well, something along those lines. Um, but, but then the other thing would people say is that's also so important about data. So what happens is we were talking before about these large data lakes that then might turn out not to be that useful. And so what happens is if you collect all your data with the idea that theoretically curious people could use it to answer anything, it's like theoretical, theoretical, theoretical. And then when people try to use it for practical reasons like, Oh, it doesn't really work that well and that's why what seems to be much more effective is when you try to figure out what is your actual question you're trying to ask and then really ask it and iteratively ask it. Cause as you start to figure out what data you need, the goal is to try to do it as soon as you can as the whole. I mean that's fundamentally how startups work where they try to figure out here this is our working hypothesis, but let's actually try to talk to a customer.

Jason Lopez: Yeah. That reminds me of why the cloud has been so powerful lately for startups because you don't have to put together a bunch of boxes for your data collection and then hire an it team to oversee all that

David Shaywitz: 100% you have to have, you've totally taken that off the table. Essentially everybody can have unlimited, secure, safe computing everywhere, compute and storage everywhere. It's immensely liberating and empowering for startups and that's how come all the startups use it.

Jason Lopez: Yeah. And so if, if you have an idea to do some vaccine research, you don't have to start a data center to go do it. You can just get AWS or wherever you want to go and, and you're up and running.

David Shaywitz: Right? I mean, this is like, you know, when I used to be chief medical officer at DNA nexus, I know it was just so conspicuously obvious there that to the cloud is enormously empowering. And it was funny because even five years ago there were a lot of medical centers are like, Oh, I don't know about the cloud. We're a little bit concerned and now it's actually obvious that that's the safer place to go. There are different approaches to cloud. Obviously, if, you know, sort of failing it, it seems, um, I mean almost sort of, um, it seems so clearly advantageous now that I think that the, you know, the, the discussion has moved onto the next series of challenges.

Jason Lopez: David Shaywitz is a Harvard-trained medical doctor and scientist who co-hosts the podcast on digital health called tectonics. He's also the founder of astounding health tech, which advises biopharma companies engaged in research and development. There are all sorts of resources on the web regarding the Coronavirus. If you want to see a real-time dashboard of the global spread of the pandemic, the center for systems science and engineering of Johns Hopkins university publishes one. It's in the notes accompanying the post of this podcast. This is the tech barometer podcast. I'm Jason Lopez. Check out our other podcasts at www.theforecastbynutanix.com.

Jason Lopez is executive producer of Tech Barometer, the podcast outlet for The Forecast. He’s the founder of Connected Social Media. Previously, he was executive producer at PodTech and a reporter at NPR.

© 2021 Nutanix, Inc. All rights reserved. For additional legal information, please go here.