Transcribe your podcast

Sometimes we don't appreciate that. Takes until they're taken away like a walk, not for any reason, not to any destination. We will take back the streets where we live, where we live in Italy, because this Labastida. And we welcome the day when we walk again. Walk together with us a bit on Idalia and Joy Perrone responsibly.


This Trinko where dot, i.e. that Pat Kenny show on Newstalk. Ireland's covid Tracker app was launched on the 7th of July as more than a million downloads, and that happened within 48 hours. But since that time, some 487 users have received a close contact alert and the Irish model has been adopted by other countries around Europe. But just what went into the development of the app and what has been learned by the agency in terms of future projects?


Gama Cryosphere is the project manager for the covid Tracker app at the agency, and he's spoken exclusively to our technology correspondent, Jess Kelly. And Jess is with me now.


Jess, good morning. Good morning, Pat.


Now, it was very successful, certainly loads of people downloading it, but it wasn't as clear cut, a guaranteed success for the team working on it. That's right.


You may remember that we had a number of expected launch dates from as early as the end of March of this year. But it was a number of months before the final version of the app was made available to the public. GAA Creosote, who you mentioned there, explained to me that work began on the app on St. Patrick's Day. Our conversation started between on Carducci, a corner, the Department of Health, the Hajazi and the Department of Public Expenditure.


And it was around this time that the contact tracing infrastructure was being put in place and these different bodies were coming together to see how, if at all, technology could help with the process.


GHA explained the initial thinking behind us that began with a more or less a scramble to try and figure out what can we use. It ultimately boils down to a problem of trying to find people who are too close for too long and and how do we identify those people? And then how do we let them know if they've been in contact with somebody who's tested positive for covid-19? And you can do that in many different ways. So at the very beginning right now, as we look back at it, it all makes perfect sense that we would end up with a Bluetooth based proximity detecting app that's privacy preserving.


But back at that point of time, we were looking at using like what data could we use that would tell us who was where? And that could be GPS data, could be telco data. It could be social media data. It could be things like identify our data is being used in different countries like that. So we were scouring the globe to try and figure out what other people were doing. And initially this started off as us using. So I think we've built like in the 100 days we spent building the app, I think we built it three times.


So the first iteration was we were building it using basically used GPS data. So we're using GPS data and time to try and figure out where people in the same place at the same time. But there's a number of issues with that. I think that version of it lasted about four days and then we moved. So and it was via conversations with colleagues in the NHS, in the UK and also in Singapore. Trace Together was released on the 21st of March and Trace Together was the first of its kind.


So basically the app we have today is derivative of that. And I think every other app globally is derivative of what they did trace together some significant differences, but it all kind of comes from that. And it was using the Bluetooth sensor or the Bluetooth radio on your phone to send a signal and receive a signal and then figure out, based on the strength of that signal, how far the phones were away from each other and then for how long would those contacts occur?


Three versions of the app within 100 days, an incredible pace of turnaround, the accuracy of the Bluetooth technology, something that had been called into question by some. So what was the thinking finally in terms of going with that model, despite the reservations?


This is an important point to dig into a bit more. You heard Gary mention there the different avenues that were explored, be that using GPS, social media or phone data, all of which would have been far more intrusive from a consumer privacy point of view. And that obviously would not have gone down well with the different privacy advocacy groups, the Data Protection Commission and the consumers, you and I, who they were hoping would download the app. This Bluetooth method meant they could gather the information anonymously.


But even with that part of the problem squared away, there were other challenges, Asghar explained.


If you look at all the options, the perfect way of doing this. So if you're to look at kind of what the gold standard for this would be would be to use ultra wideband. So in the latest greatest phone in the iPhone 11, the latest Samsung phones, you've got an ultra wideband chip and that will give you centimeter based distance. So if this was five years in the future and we were doing this, that's how we do it. And but given that we don't have that, then you have to reduce this back into what do we have and what can we do with that?


And what Bluetooth can give you is, you definitely know. So in terms of I think people become very, very obsessed with the accuracy thing to say it can't detect. Is that exactly two meters, you have to remember that the two meters is a heuristic as well. So the two meters is part it's a probability based determination of you're more likely at this level, but it won't say that if you're at three meters, you won't get infected or if you're one measure, you will get infected.


It's saying on average. So when you begin to look at it that way, then we're looking overall at the at how accurate the results are. So are we picking up close contacts? Are we picking up people that we should pick up? And that's that's one of the pieces that we're working through at the moment.


Now, you spoke there about the privacy aspect of this, and that's something many people were cautious about. You know, would the app collect information and store it on some sort of Big Brother kind of server?


Yeah, this is something that I got bombarded with on social media. People were concerned about us and I think people were right to be concerned about it. But this app has been designed from the ground up. With privacy in mind, the import of person information is not vital for it to work. Apple and Google, who created the exposure notification technology, do not have access to any form of database with personal information from the app. And you can hear the full details of the process between the U.S. and Apple and Google in the full interview, which is up on Newstalk dotcom now.


But that was very much addressed from the outset with members of the team from the Hajazi looking to ensure that the app works on what's called a consent model of data gathering, and that places the most amount of control in the hands of the user.


Now, the app had more than a million downloads within 48 hours, but they ran into, I suppose, a speed bump a few weeks ago, beginning of August, a battery drain issue. Now, what did we learn about that?


Yeah, this was a big headache for the team behind the app, and it caused around 83000 Android users to uninstall the app. I asked Garry McCarthy as to who's the product manager of the covid Tracker app if they knew what exactly went wrong.


So we do know what went wrong. So these identifiers have a fixed structure and there was a malformed identifier and it basically got bundled up on the road and distributed out to phones. And what then happened was when it got passed into the explosion notification system, it ended up getting itself into a loop. And that was it was trying to retry, marching this over and over and over again. And it just got stuck. We we found out about it Saturday evening.


By Sunday, about 7:00, we had a fix that was being pushed out. But it's a fix to Google Play services to that rolls out and that cell phones get it incrementally over time. So it's not just one that we can't just push it out. So it's not like an app update and that both removed the issue and put a fix in so it wouldn't happen again. This is it's one of those things that when you're running at speed, things do happen.


And so far as I've said often and of note, the new pay for the software and the bugs come for free and then you end up paying for the bus. I think we pay for the bugs and pains last year and last weekends. And I, I can't definitively say there will not be any more bugs. But I think in terms of things like that, we're making sure and we have to roll through to try and ensure that there's nothing in this code base that is can can create a situation that's going to cause something like that to happen again.


Have you recovered the numbers or in terms of people who uninstall the app over that weekend?


So we recovered a lot of the numbers that came back. And it's one of those things because this is of the privacy preserving nature of this. We can't tell. So we can't tell who's gone, who's come back so we can tell some things. So are our next position as of the 13th was I think we lost eighty three thousand people, something like. It's a lot of people came back once they discovered that it was safe to come back. I can only apologise to the people who suffered from this, but we're doing our best to try and make sure and recover this and trying to get it back to a point where and to build up to rebuild that trust if we've lost that trust.


Now, that was Garmo Chris, the project manager for the Kova Tracker app, speaking exclusively to Jess Kelly. The interview is in full up on Newstalk dotcom or search for tech talk, where if you get your podcast.


But, Jess, you've been updating what you are putting before our listeners because we heard about deletions and what's the news there? Yeah.


So as you heard there, the Android battery drain issue caused 83000 people to uninstall the app. Other factors like phone storage on iPhones, for example, caused people to uninstall the app. Some people just didn't want it on their phone. And so they uninstalled the app. The figure that is displayed within the app that says around one point seven million people have downloaded the app is actually an overstated figure. And the reason for that is it counts. For example, if you had downloaded the app on a Monday, you deleted it on a Tuesday and you downloaded it on a Wednesday, you'd be counted in that one point seven million figure twice because it can't actually distinguish between who was doing what and when.


That's part of the privacy preservation side of the application. So speaking Skar in the last few minutes, again, he has told me that they're going to change the way the information is displayed in the app and the way it is measured. And so instead of the figure, the number of downloads, they're going to move to a percentage of coverage because as it stands, as of today, four thousand people have downloaded the app onto an iPhone. Nine hundred 9000 people have downloaded it onto an Android device.


But because they could have been repeat downloads, people have deleted and come back to it. The figures that we have as of today are five hundred eighty seven thousand iPhones and six hundred eighteen thousand Android devices currently have the app installed as of right now. Mark Kelly, thank you very much.


Get your ticket for the win your house in Dublin draw before it's too late on the 4th of December, you could be the lucky winner of a four bedroom house in Dublin for a brand new master. Two tickets are available on when your house in Dublin Dotcom. Don't miss out on this life changing opportunity, but supporting Donegal GAA.