 Hello there, and thank you for joining us. My name is Jeremy Howard. I am the co-founder of Fast.AI, and this is the other co-founder of Fast.AI, Rachel Thomas. Hi, I'm Rachel Thomas. My background is I have a PhD in math and I previously worked as a software engineer and data scientist. More recently, I'm the founding director of the Center for Applied Data Ethics at the University of San Francisco. And as Jeremy said, I'm also the co-founder of Fast.AI. So Rachel is a mathematician slash data scientist slash software engineer turned data ethicist. I am the opposite. My major was actually in ethics and nowadays I'm a professional data scientist. So we both together are super interested in combining data and ethics and policy and society. And we found ourselves in a funny position last week where we were starting recording our big annual course, which hundreds of thousands of people a year watch right at the time that the COVID-19 thing was exploding. And so we did some work looking at COVID-19 in terms of from what the data looks like. Much to our surprise, the report we wrote has been read by well over half a million people. The video we created as part of this course has been watched by thousands of people and there's a request for us to do more of that. So here we are doing more of that and today we're going to tackle a topic which I think is likely to be the most important topic we've ever tackled. So let's jump in. So this is I think the hardest situation I've ever been in trying to teach and the reason why is because I am in the United States of America. And here in the US over the last 24 hours, there has been this extraordinary new movement, which is basically to say, let's just let people die on the assumption that that's going to help the stock market. And so if I seem a little less than jovial today, it's because I am less than jovial today. So I'm not going to be talking about that in particular. Hopefully it's actually pretty obvious that. Well, economists are already saying it would cost about $5 trillion before we even start taking account of the lost productivity due to the fact that people would not be going to work anymore when all of their friends and family are dying around them. So economically not a great idea and you know, societally maybe one of the most evil things I've ever heard in history. But we're going to not focus on that. We're going to focus on a potentially interesting part of a really important possible solution to the longer term. The first thing I wanted to mention is we've been talking a lot about the importance of stopping the spread. And a key part of that has been locking things down as soon as possible. And interestingly, we're already seeing in the US the impact of that. In the US, Kentucky and Tennessee right next to each other. They are testing at similar rates. And so you would expect them to have a similar spread. Kentucky got in ahead of Tennessee when it came to doing a lockdown in terms of the, you know, they actually just declared a state of emergency back on March the 6th. And you can see already the difference between the number of cases in Tennessee, positive cases, positive tests versus the number of positive tests in Kentucky is an extraordinary big difference. So the stuff we are now seeing starting to happen around the Western world with these lockdowns, again and again, we're seeing the earlier, the better. The orange bar here is the number of positive cases in Tennessee. And the blue bar here is the number of positive cases in Kentucky, whereas the lighter blue and lighter orange is the number of tests that have been done. There's still people saying that maybe not that many people are actually dying and finding all kinds of obscure ways to murder the data to suggest that even Italy is actually not that bad. But the fact is, if you look for example in the region around Bergamo in Italy, the number of people who died last week was four times higher than the number of people that died a year earlier. So like regardless of how you think about what kind of testing is being done or, you know, what kind of post-mortems are being done or whatever, there's four times more people dying of all causes in this region. This thing is killing people at an astonishing rate and it's not even close to peaking. So obviously we need these lockdowns, but the question is what happens next. And we saw last week this tweet from the Surgeon General saying, don't buy masks, they're not effective, except for healthcare providers. And this has been kind of a pretty common refrain. It's certainly not the only the US Surgeon General that's saying that. And there certainly is a big issue. Doctors are saying there is not enough protective equipment including enough masks for hospitals. So PPE is referring to this protective equipment, masks, gowns, eye shields. There is a huge shortage. And so perhaps the reason that a lot of policy folks are saying masks are not useful is to stop people buying them. I mean, this is what he's saying here, don't buy masks in the hope that that will cause there to be more masks available for folks in hospitals. Interestingly, when you actually look at what hospitals are doing, they're saying, please donate us protective equipment. These are things we need, not just protective equipment, but even testing equipment. But interestingly, they're even saying handmade masks would be useful. So it's interesting to see what the actual hospitals are saying they want. And so when hospitals are saying we need masks and even handmade masks would be useful, it starts to question this idea that like they're not useful for the public. So like, why do hospitals want them if they're not useful to the public? Like, how does this work? There's also some interesting empirical data, very rough empirical data. But if you look at all the countries which have flat curves, they are all countries actually where mask wearing is extremely normal. And not wearing a mask in places like public transport would be considered very antisocial. So there's this kind of country level data that suggests that, well, something's going on in these countries and is masks part of that. Taiwan does not appear on this at all because they basically didn't have enough cases to appear on this graph at the time that this tweet came out. They only have 153 cases and two deaths in the whole of Taiwan, even though they're right next to China where this thing blasted off. So what happened in Taiwan? They are making 10 million masks per day. Now, I thought that sounded crazy unlikely, but actually this is the leader of Taiwan. And here she is saying, yes, Taiwan is now making 10 million masks per day. And they are a small country. I think they're like 30 million people or if you like a small region, depending on your politics around this. So that kind of got me interested, like thinking, well, if masks are important and a small country like this can create so many. You know, what does that mean for a country like the US? Is that something we should be doing? And so I started doing some research and I saw, for example, University of Michigan study on pandemic flu, not on COVID-19, but there's similarities. And they found that combining hand washing and using masks makes a huge difference to the spread of pandemic flu. 75% reduction. Now, 75% reduction would be the difference between a pandemic and something that would just die out and fizzle away. So this is again indicative. It's not proof. So I started digging in some more and I asked the wider community to help. And we now have a list of 30 papers, or nearly 30, maybe it's 28 papers that have actually studied the question of the efficacy of masks as a way to stop spread and transmission from many different angles and many different types of masks. And all of the research I've seen points to any kind of mask as being helpful to significantly stopping transmission. So this has been quite a learning journey for me. And so I've really focused on what does the data say and just looking at primary research papers, not at government recommendations, not at blogs, but at actual science. And we'll provide a link to these papers with the video. One of the most interesting things I discovered is in the Czech Republic, which has a really fantastic healthcare system and is doing a lot of testing. They actually made mask compulsory for the whole country since March 18th. If you want to go anywhere, you have to wear a mask outside the house. And they have these little illustrations. The check here means your mask protects me, my mask protects you. And they're really focusing on the social side that if I have a mask on and I might be sick and I don't know about it. I am actually doing a pro-social thing by helping you to not get sick. In the whole country, there's only been two deaths. And look at how we're not seeing the exponential growth in new cases. It's actually flattened out. So this is a super interesting situation. So I started wondering, like, whoa, how did that happen? How did this country, like very unusual in Europe to see mask wearing? So how did this happen? And there is this fantastic story about what happened. You can find it at tiny.cc slash masks rule. I strongly recommend reading it. But basically what it says, in fact, let me jump to it now. What it says is this guy, Peter Ludwig, who is a kind of a social media influencer who focuses in science communication, made a video about the importance of wearing masks. And that kind of kicked off this huge community reaction. It's an amazing story. And this is the video. This is Peter. And after really digging into the data and the virology and the science of this, I am pretty convinced that everybody wearing masks could be a key or probably is a key part of the solution to stopping the lockdown, eventually getting the economy back on track, restoring normalcy to society and reducing the spread. I am not as good a science communicator as Peter is. So what I'm going to do is I'm just going to tell you what he said with his pictures. Okay, so I get no credit for this. I'm just repeating what he said because maybe you don't speak Czech. Okay, so Peter started out his fantastic video by pointing out Mongolia. Now Mongolia is right on the border of China. So lots of people moving backwards and forwards, right? And somehow at the time that he did this, the 14th of March, they had only one case. Today it's gone up to four cases. It actually turned out that one case checked themselves out of hospital and against the law went and hung out with 200 people over the next few days, which is a terrible, terrible idea. But how come Mongolia had only one case? And so Peter said he noticed that something striking him out in Mongolia was they had a policy that you have to wear masks since January, which is pretty extraordinary. So Peter went on to describe, and I'm using some of my own slides now. His slides are much more, his presentation is much nicer than mine. But I wanted to show you some additional information here. So he described how there are different kinds of masks. There are what are called respirators, the most common type being the N95. It's just a rating. And they're full seal hospital grade things, right? Versus surgical masks, which are these just loose fitting pieces of cloth. The key thing here is that these different pieces of protective equipment protect you from different things. But what's interesting is, as Peter mentioned, there is research that shows that even the basic loose fitting pieces of cloth actually works super well against a wide variety of viruses. So this is actually the research that he's pointing out. And this comes from smartairfilters.com, but the actual data comes from this report, Davis Adele. They've just made it into a nice picture. And what they found is that against 0.02 micron particles, surgical masks are super effective, vacuum cleaner bags are super effective, dish towels are super effective, cotton t-shirts are super effective. And actually in this study, they had two different sized particles. They also looked at 0.1 micron particles. I think it was 0.1. And they found the same things were even more effective, because 0.02 is particularly small. They also looked at kind of doubling them up. And they also looked at how well can you breathe through these different materials. And they actually came up with some suggestions around things that are easy to breathe through. So in this case, two t-shirts, easy to breathe through and end up being pretty effective. So anyway, you can check out this fantastic article or the original paper for yourself. But basically the story they found in this research and a lot of other research I've looked at is for stopping the spread of viruses. You don't need N95 respirator masks at all. I mean, they're great, but lots of other things work great as well. So what's going on here? How does this work? COVID-19 is a 120-nanometer-sized virus. So Peter points out that how can these simple pieces of cloth protect from that, because they have bigger holes in 120 nanometers. The point Peter made, though, and described again and again in all the research I've looked at, is that COVID-19 is spread through droplets. In other words, you cough and it goes into your face, right? And specifically it goes into your eyes or it goes into your nose or it goes into your mouth. Also talking. Sometimes I've noticed when I talk little bits of spit fly out. I'm sure we're all like that. So these are droplets. And the droplets is what the mask catches. So there are three categories of transmission described by virologists. There's contact transmission, droplet transmission, airborne transmission. So contact us when you actually touch someone or something with the virus. And this is a big issue if you touch. We talked last week about sanitation. If you touch a surface which somebody who is infected has touched also, and then you touch your face that can get you sick. What we're talking about for masks though is the coughing and talking, which seems to be such a common vector for spread because that's actually how on the whole this particular virus is kind of expelled. Airborne is a nightmare. Airborne is when things, they aerosolize. They turn into an aerosol. They float in the air in tiny, tiny, tiny like sub, I can't remember what it was, 0.5 micron droplets for ages. That can actually happen in a hospital. There are certain procedures in a hospital which cause the droplets to aerosolize. And the kinds of masks we're talking about, the basic masks don't really help there. So in a hospital setting where there's potential aerosolization, you need the good masks. So actually Oxford University just published today a new report about this, where they described how basic masks are fine, except for if you're in a hospital around an aerosol generating procedure, or AGP. And this is a list of AGPs. So you're probably not doing these things at home, right? So you're fine to use whatever, couple of layers of T-shirt, right? In the hospital even. And this is what the University of Oxford is saying from their research. This is with COVID-19 research. This is not just any old research. So like basic masks, homemade masks, this is why the university we looked at earlier is asking for homemade masks even, because they do really help. So there's been a lot of research already, empirical research showing that these basic masks actually help in practice. So here's an example for SARS. So this is from SlateStarCodex.com, a summary of some research. They have a link to a paper about SARS, which looked at how folks that were wearing even surgical masks got no infections at all. This is the Oxford study that I mentioned. And so their conclusion, again, standard surgical masks can be used in most community-level clinical interactions, let alone at home, right? So they weren't explicitly looking at home, but we've seen from the other research that basic materials are just as good as surgical masks or very close to. The research, empirical research has also found that basic masks do help in practice, empirically in practice. They do help in non-healthcare settings. Again, not COVID-19, but swine flu. When they looked at the result of a flight in which there were folks with swine flu on that flight, those people that wore a basic mask were 0% that wore masks got sick, competitive 35% of non-wearers got sick. So there are huge differences when you use even basic masks. This is totally in line with what virologists say. So this virologist says, COVID exposure dose matters. This is really important, kind of when you're thinking about the science here. Masks can help anyone because they reduce the amount of virus released or taken in. And the immune system wants a low dose. So if you keep the dose low, then the virus has to divide many times to get to high levels, and then the immune system's got lots of time to kick in. So this is kind of counter-intuitive, but this is actually what the data is now showing us, and the science is now showing us, is that the transmission of this virus seems to occur primarily through sustained exposure. So in Hong Kong and Singapore, they define that as less than 6 feet away from somebody for at least 15 minutes in Hong Kong, or 30 minutes in Singapore. So this is super important for understanding how this virus works. It's not about just passing somebody in the street, unless they cough on you, but actually being around somebody for an extended period of time and close to them. So a key point which we absolutely know now, we've got plenty of data to show this now, is that many people are infectious before they have symptoms. So it's not okay just to wait until you're coughing before you start putting a mask on. Something like 50% of people with the virus are asymptomatic, and they are spreading it even then. So this is a key reason why you have to get the mask on early. So these little pictures here, as Peter is describing, how when you wear a mask, you stop other people from infecting you, but you might be infected with it. No idea. You might never get into symptoms, right? But you could be spreading it to your friends, your family, your community, your coworkers without knowing it. So you need to wear a mask for that reason as well. So he describes this as a selfless motivation. And this has been a key part of the messaging in the Czech Republic. It's this idea that this is such a pro-social behavior to wear a mask, because you are protecting the people in your community, the people in your workplace. So if we look at it, we saw that graph earlier that showed there's quite a few countries now that are controlling this without a lockdown, like Korea. And in every country that has controlled it without a lockdown, or isn't in a lockdown anymore, I should say, they're doing four things. They're doing rigorous testing. They're doing rigorous contact tracing. So if you're tested positive, who's everybody you've been near? They quarantine anybody who's found positive. And everybody's wearing masks on, you know, on public transport at work, you know, anywhere that they're around other people outside of their family. So like from a data science point of view, this is, you know, we don't need proof, right? None of these are economy destroying levels of cost, right? Masks. I mean, we can make our own masks for like 10 cents. So, you know, these are things that allow us to get the economy going again and avoid people dying. So we save lives and potentially trillions of dollars. And it's not, this is not like some crazy idea that might just work. This is something that lots of countries have done and it's already worked. So it's like, why isn't everybody doing this? We need the lockdown still to get to the point because so many countries like the U.S. and the UK are way too late. But, you know, in a few weeks or a few months, once we're past the apex, this is what we need to be doing. It's going to be super hard to do the masks bit because we're, you know, our governments have been stigmatizing this. You know, they're saying you don't need them. If you wear a mask, you're taking it away from hospitals. So there are now all these Asian countries that acted quickly to contain the virus. And when you actually look at like just the traffic on the streets, they're kind of back to about normal. So there's lots of data you can look at. In this case, this is Tom Tom traffic congestion, right? And it's just fascinating to look at these countries where the last seven days versus historical average. This is from the Financial Times, I believe. Sorry for accidentally cutting off the citation, which I tried to keep. Yeah, you can see how there are countries that are nearly back to normal. People are moving around again. So Peter used, you know, used all that to try to get rid of the first idea, which is we don't need masks. We do need masks. So that was his cross that off. So the second argument he dealt with is there's no point saying you need masks because there are no masks. So his point was, well, make them go and make masks. And so here are pictures of people in the Czech Republic after this video came out making masks and hanging them up on mask trees for anybody to come and pick up that needs one. And one of the things you notice here is the masks don't look professional. And I think this is really important, right? Because if you have a polka dot mask, nobody's going to come up to you and say, you're stealing professional hospital grade masks. You know, they're cute. And so part of the Czech community reaction was to have, like, you know, humor and, you know, just make it as fun as possible, get everybody involved. And so even kids have their koala. I don't know why there's koalas in Czech Republic. I guess it's just there on a visit. It has its own mask. There are all kinds of interesting things you could use for these masks. So one chap pointed out that 3M has a household filter which costs $22 and gives you enough for 60 masks and it is actually officially virus rated. So, you know, if you want to go fancy, you can spend $0.30 and cut those up. As we've already noted, things like cotton should be just fine too. But yeah, string, staples and hot glue, $22 worth of this stuff, maybe a bit of cloth on each side to make it more comfortable. So the call to action that Peter made was basically, he said, I want everybody in the Czech Republic to wear a mask. Everybody, 100% of the country. Which, I mean, that sounds crazy to me that that could be possible because at the time this was something that had a social stigma around it. So he pointed out something really interesting, which is that this would not be, if this was successful it would not be the first time that something with a social stigma turned around. Just like in most Western countries, where I remember when I was growing up, dogs would poop on the sidewalk and you would leave it there. And he pointed out that we managed to change the social expectation that if a dog poops, you pick it up. And before that changed, no question, people would have been like, oh my gosh, you're picking up poop? That's gross. Whereas now, it's like, excuse me, you seem to have left your dog's speakable matter on the sidewalk. Please pick it up. Thank you very much. So, you know, these things can change. Social expectations can change. And so his hope was that if you see somebody without a mask, you know, people will stare at them and think, who is this antisocial person? So he made three particular asks. He said, share this video. Take a selfie wearing a mask and use the hashtag masks for all. Well, actually, the Czech version. But let's pick the English version. And influencers should set a good example for their followers. And so I think it would be super cool if, like, you know, celebrities started rocking their own homemade, colorful masks, maybe pictures of them making them themselves, you know? I mean, I think that would be super important. And, you know, the point is that this is a little word cloud of, like, the most important values to folks in the Czech Republic and the biggest one was kindness. And so he is saying, like, this is the most kind thing you can do is to have a little bit of inconvenience yourself in order to protect those around you. Oh, my God, it was successful. So today, when politicians appear in public, they're all rocking masks. People are putting mask pictures on the public transport, on the statues. It's extraordinary. So there's a lot of recommendations out there for face masks. Michael Lin, who is a biochem researcher, has lots of great information that we can share similar kind of stuff with what we've described about, but some more links. Actually, a article that just came out today, we might need to update this, is that, no, actually, this looks correct. So he's actually already got the new thing. If you put a mask in the oven at 160 Fahrenheit, 70 centigrade for 30 minutes, it sterilizes it, and you can use it again as much as you like. No alcohol. So there's actually a new Harvard study I just saw today, which actually has shown this. Interestingly, you can see this is coming from Wei Xin QQ, Chinese sites. A lot of the best information we have is coming from Chinese sites, because they're the ones with months more experience of this. So a really interesting thing that appears all the time in, like, the training that doctors talk about for mask fitting is if you want to know if your mask's working pretty well, go somewhere, like, get something smelly, like some sugary water or some smoke or something, and see if you can still smell it with your mask on. And if you can smell it a lot less, then you're blocking the particles. So that's kind of the rule of thumb that you keep seeing people talking about. Okay, so that is what I wanted to say about masks. I guess, you know, I I'm not a big YouTube influencer, thought leader, whatever. I'm just a data scientist. But I guess my job is to look at data and try to figure out what it says. And the data is super, super clear on this. It's not like there's no 100% perfect proof randomized controlled trial for this exact illness disease. Such a thing would be impossible because it hasn't really been around long enough. You can't really do a randomized controlled trial where you say hey, you're not allowed to use a mask and you are. But from a kind of a Bayesian prior perspective, the evidence is so incredibly strong that this is a good idea. The cost is so low. The upside is so huge in terms of life saved and the economy being able to get back up and running again. It seems like madness to me not to do this. So I will repeat Peter's cold arms here and say let's make this happen and particularly, you know, let's see influences come out wearing masks. Start start making masks. We're going to have links with this video to how to make them at home. It's very cheap. It's very easy. And let's hope that we can make happen in the US and the UK and other countries what the Czech folks have done so incredibly well.