This week: the internet made old again, and why AI struggles with the new. Sandra Peter (Sydney Business Insights) and Kai Riemer (Digital Disruption Research Group) meet once a week to put their own spin on news that is impacting the future of business in The Future, This Week.
The stories this week
01:39 – Why does it feel like 1999 on the internet?
14:49 – AI runs up against a big data problem in COVID-19 diagnosis
Other stories we bring up
YouTube takes down David Icke 5G interview
Facebook and Twitter take down video of Brazilian president endorsing unproven antiviral drug
Online is where we go to be alone together
Has the coronvirus killed the techlash and our previous discussion of big tech in the time of corona
The Virus Changed the Way We Internet
The loneliness epidemic and our previous discussion of it on TFTW
Chinese AI beats 15 doctors in tumour diagnosis competition and our previous discussion of AI successes in diagnosis
Facial recognition is no match for face masks, but things are changing fast
Real World Masked Face Recognition data set
Listen to our new series – Corona Business Insights
You can subscribe to this podcast on iTunes, Spotify, Soundcloud, Stitcher, Libsyn, YouTube or wherever you get your podcasts. You can follow us online on Flipboard, Twitter, or sbi.sydney.edu.au.
Our theme music was composed and played by Linsey Pollak.
Send us your news ideas to firstname.lastname@example.org.
This transcript is the product of an artificial intelligence - human collaboration. Any mistakes are the human's fault. (Just saying. Accurately yours, AI)
Disclaimer We'd like to advise that the following program may contain real news, occasional philosophy and ideas that may offend some listeners.
Intro This is The Future, This Week on Sydney Business Insights. I'm Sandra Peter, and I'm Kai Riemer. Every week we get together and look at the news of the week. We discuss technology, the future of business, the weird and the wonderful and things that change the world. Okay, let's start. Let's start!
Sandra Today on this week: the Internet made old again, and why AI struggles with the new. I'm Sandra Peter, I'm the Director of Sydney Business Insights.
Kai I'm Kai Riemer, professor at the Business School and leader of the Digital Disruption Research Group. So Sandra, the holidays are coming up. What are we doing?
Sandra I think we're doing a slightly shorter episode this week. As our listeners might have heard, we're also doing episodes on the impact of corona on the future of business. So today we'll keep The Future, This Week a little bit shorter, but we'll still have time to talk about a couple of stories.
Kai So on Corona Business Insights, we're looking specifically at how COVID-19 impacts on the future of business. While here on The Future, This Week, we will continue with our more longstanding topics around the impact of technology on business, the future of business more generally. COVID-19 will, of course, come up, but we've picked some stories for today which continue topics that we've touched on previously, that show up in a new light now that COVID-19 is happening.
Sandra Okay, let's start in the past, 1999.
Kai So, Sandra, what happened on the future a long time ago this week?
Sandra So our first story comes from the MIT Tech Review., and it's asking why does it suddenly feel like 1999 on the Internet?
Kai Ooh, Ooh! I have an answer. Well, one reason is: it's slow.
Sandra It is slow. It's clogged up and slow.
Kai If people remember what it was like in 1999, dial-up modems, waiting, the thing was called the World-Wide Wait' for a reason because everything took time. And somehow, we're transported into this world because, you know, with everything happening online, work, school, university, internet infrastructure around the world has become under strain. But this is not what the article is about.
Sandra No, it's not. The article talks about the fact that the pandemic seems to have turned back the clock to a bit of a nicer, kinder time on the Web. And it lists a range of things from hundreds of thousands of people on Instagram joining virtual dance parties hosted by DJ D-Nice. And that includes people like Oprah and Mark Zuckerberg and Michelle Obama, to people trading recipes and Chatroulette making a comeback on Google Hangouts, to happy hours and book clubs and collective creativity on Google Sheets, and nice emails. So a much kinder time on the Internet.
Kai So the article points out that we're even kind with strangers, that we join meet-ups, we join online hangouts. We're joining all these things where our friends, strangers are really having a good time. And that in the time before COVID-19, where we were obsessed with portraying this perfect image of ourselves and photoshopping and filtering our selfies, and trying to be all perfect, now we're living in a more scruffy time, more authentic. We can be ourselves a bit more. We're joining yoga classes that are happening in cluttered bedrooms, we're having all these work meetings where people are dressed in sweat pants. And so it's a much more authentic, a much more normal world on the Internet, not as glossy as previously. So it feels a little bit more like in the early days of the net. So that's what the article says.
Sandra And here we paused then thought, but is it?
Kai But is it?
Sandra And there's a couple of reasons. Whilst all the things that the article points out are actually happening, and many of us have joined Happy Hour on the Internet, or even caught up with friends we haven't seen in ages, or extended a hand to colleagues we might have never talked to before to help them out with carrying out activities online.
Kai The old world still exists, right? So yes, these things are happening and there is a certain excitement about the inevitability of having been socially isolated and people are trying to make the best of it. And the article lists a few of those things that are happening.
Sandra But at the same time, there's things like the 5G conspiracy.
Kai And we're not going to give this, frankly bullshit, any airtime here. But we want to point out that this is still happening. The Internet is still a place where this dangerous misinformation is being spread.
Sandra Indeed, a good example is YouTube having to take down an interview which was watched by more than 65000 people as it was livestream peddling conspiracy theories. And let's not forget, many of these people watching it actually clicked on online screen buttons, making payments to have their reactions to the conversation be prioritised, and be put at the top of the list. There was also the news from last week when both Facebook and Twitter had to ban a video of the Brazilian president endorsing an unproven antiviral drug as a treatment for the coronavirus.
Kai And while we commend big tech to actually make a stand at banning some of these new COVID-19 conspiracies, we also want to point out that conspiracies around vaccination, flat Earth, the upcoming US elections and of course, climate change are still continuing unabated.
Sandra Which brings up the issue of big tech itself. One of the things that is very different to the Internet of 1999, and let's remember, this is before social media and before smartphones, is the fact that big tech is very much still present.
Kai Yeah, the Internet of 1999 was a place where everyone was creating their own personal web pages where we were hanging out in chat rooms or news bulletins and maybe sending email. The Internet of today is very different. It's almost like a collection of colonies. For many people, internet is almost synonymous with Facebook. So rather than having a space like in the old days, where everyone in an almost democratic way had the same access, it is now still big tech and Facebook and their algorithms which decide what people get to read, what people get to see and how what they say is being distributed across those networks.
Sandra And we've discussed in a previous episode whether coronavirus had killed techlash and, we'll include the link in the shownotes. But let's remember that the ways in which personal data is exploited on the Internet, the anti-competitive behaviours of these companies, these are very much still happening. It's just that they don't get the airtime that they used to get prior to the pandemic. And if anything, these companies are starting to play an even bigger role in our lives as people look for community. There are many people who have not been using Facebook a lot, but are now engaging much more to keep in touch with friends and family or people even who have given up Facebook who return to it. People joining Amazon Prime or other online delivery services and becoming part of an online community that might be very difficult to quit, once the pandemic is over.
Kai And we'll put a link in the show notes to an article in The New York Times, which has some interesting data about how our use of the Internet itself has changed. So one interesting insight is that we're moving to using websites more, so bigger-screen devices rather than necessarily apps. Video apps, yes. But for Facebook, for example, the use of Facebook dot com is up by 27 percent, whereas the app is holding steady, more or less. But yes, Facebook users certainly up. And so is Netflix and YouTube.
Sandra So will it last, is the question. Whether we're talking short term, is it still going to be the case another two months into this pandemic? And will this still be the case two years from now?
Kai So my view on does this outlast the pandemic as such, right, once we have a vaccine and all of this is over, and my view is no it won't. Maybe some things will stick, some apps will have come up on the back of this. Some services will have gained. Big tech might be bigger than ever. Changes that might have happened without the pandemic as well, we don't know. But will we all be in a happy clappy place online after this is over? Probably not.
Sandra And we've had Adam Kamradt-Scott in a previous episode, and we'll include the link in the shownotes, he was talking very much about this cycle of panic and neglect, that everything changes whilst we're in a crisis, such as this pandemic. But once the crisis has subsided, we very, very quickly return to our normal habits. There is also the possibility that a year and a half, two years down the line, we're still in this crisis in some way, shape or form. Say we do not find the vaccine or the economies don't recover quickly enough. And then the question is, would it still hold in that case?
Kai You make a good point, but I don't think we have to look into the future quite that far, because I think we will see, on a much shorter timeline, give it a month or two months, this novelty factor will wear off. And our colleague Mike Seymour has just written a piece for the SBI COVID-19 portal, where he makes the argument, and we'll put the link in the shownotes, that at the moment we are very forgiving. Everyone had to move hastily their office into their bedrooms. So we put up with, you know, poorly position cameras that shoot up peoples noses and show the ceiling of a cluttered bedroom. People joining business meetings in sweat pants. And that is ha-ha funny for the first few weeks, but soon we will realise that this is the only place where we're now doing business. We will still have to make money, we will still have to portray a professional image for our businesses. And so this Wild West early experimental phase will give way to a much more professional conduct online where what is being celebrated in the MIT article will very much give way to more considerate, to more measured dealings in Zoom meetings, but also in the way in which we use the Internet to communicate.
Sandra And there is another thing that might come into play if we look a month, two months down the line, which is we're still very much at the beginning of this pandemic, and once the grim reality of both the human impact and the economic impact of this pandemic come into play. People might not be as inclined to join dance parties or to exchange recipes or spend time doing DIY stands for their devices that they can use around the house.
Kai Sometimes even from Lego, I'm hearing. Is yours on a Lego...?
Sandra It on a Lego stand, yes.
Kai Is it now? Excellent. But we also want to make the point that the kinder, nicer world of the Internet, the exciting world of the corona lockdown that is being celebrated in this article, while it exists, it is a bubble that only contains certain people, but not other groups.
Sandra We spoke at length in one of our previous episodes about the loneliness epidemic, as identified by the former US Surgeon General, who estimated that loneliness was the most prevalent ailment within the community. And this is at a time where, of course, we were all connected, but we were still also all going to work and going to spend time in social environments. And in the US, more than half the CEOs were reporting a feeling of loneliness. The UK's prime minister had appointed a Minister for Loneliness. After the report had showed that more than nine million people in the UK were feeling lonely often or always, and the health effects associated with that were tremendous.
Kai And so we want to point out that many people who have been lonely at work, for example, have been excluded from activities, but were still present in the office pre this crisis, might now disappear completely from view. And so people who are not included in these online activities, be it because they are quite introverted or they find it difficult to connect with others anyway, are easily forgotten when it comes to organising all these activities online. And then there are those who just don't have the time, the capacity or the knowledge to participate. Single parents looking after their children who have lots of schoolwork, people with elderly parents or relatives who have to care for them and are worried about them contracting COVID-19. There is a lot of people who do not have the time, the energy, the knowledge or the technology to participate on what is arguably now an exciting time to be online.
Sandra Indeed. And the argument back then was that we really lacked the depth of quality of interpersonal relationships, we don't know each other well enough, or we don't get to talk about our feelings, and our interactions are often much more superficial. There is very much a danger of that getting increasingly worse with our current situation.
Kai And so what we want to say is we hope the article is right, and we can actually sustain this and make the Internet a more inclusive, a happier and kinder place. But we would also say that we all have to think about those who we might not have seen online, to try and include them and to do something about the rampant spread of conspiracy theories and misinformation and the proliferation of online scamming to pressure big tech to uphold what they have now shown in this crisis, that it is actually possible to take some of this material offline, and do this for the other spread of misinformation as well, climate change and the like.
Sandra There was another story we wanted to discuss today, and that has to do with artificial intelligence. And a really interesting article from ZDNet that talks about AI running smack up against the big data problem in the COVID-19 diagnosis.
Kai So we've often talked about AI on this podcast, and the real successes were in image recognition, in recognising cancer cells, for example, in CT images or X-rays. So that's a real stronghold for AI. So, it was somewhat surprising to see that this is exactly what scientists are now struggling with, building the kind of algorithms to do these diagnoses for COVID-19.
Sandra At first, it would seem straightforward that researchers could be using AI to diagnose COVID-19 from chest X-rays and from CT scans. But there is one big difference to what AI was able to do in the case of cancers.
Kai And that difference is data. So, in a world where we have lots of image data that shows healthy lungs, for example, and cancerous lungs, and we have doctors who can actually label those pictures and prepare it for these deep learning networks to do the image recognition, this works really well. But that data is not readily available in a pandemic with a new kind of illness that is unfolding in full force.
Sandra So there have been two problems to getting AI to successfully diagnose COVID-19. And that is on the one hand, there is very little data available to begin with. There is not a very large dataset of X-rays or of CT scans that show COVID-19 infected lungs. Nor is this data formatted in a way that makes it readily available for the machine learning algorithm. Every single hospital has data that this formatted in different ways, that is available through different channels, and with different conditions.
Kai So the article reports on initial successes in China, where researchers were able to train networks on a few hundred or a few thousand of these images. But let's not forget that in order to get the quality and the precision required, we need thousands and thousands of these images and they are not readily available from just the one source. But there's also no infrastructure in place at this point to pull together more data. And that's also not the time available in hospitals to collect data for these research efforts at a time when saving lives is the pressing issue.
Sandra And let's remember, and we've discussed this previously on the podcast, that these systems need data that is also labelled. That means that you have large numbers of doctors, of radiologists, who need to look at these images and make annotations that then can fine-tune the way the machine learning algorithm learns what the disease looks like in various patients.
Kai And so the places in which these pictures are currently produced, the doctors simply don't have the time to annotate the pictures to train these algorithms. And while some people might say, wait a minute, there's also so-called 'unsupervised learning techniques' where AI might learn from pictures that are not labelled. That is correct, but those algorithms need even larger quantities of data, which, again, are not available.
Sandra So in time, these algorithms, of course, will be able to diagnose COVID-19. But this will take, first of all, some investment in a global infrastructure for sharing these images, and secondly, a lot of time. So paradoxically, they'll be available probably when we no longer need them as urgently as we do now.
Kai And so, again, this shows that AI, while really well able to predict outcomes on past data, when we're confronted with novel phenomena and data is simply not available, this technology is not readily useful.
Sandra And interestingly, we see similar issues crop up in other places where AI doesn't have a past dataset to draw on, and we're seeing a similar problem with facial recognition. It was never designed to recognise people with face masks on. So even systems like the ones on Google's Pixel 4, which were built less than a year ago, really struggled to recognise a person that has a face mask on. Same thing of course, holds for Apple's Face ID. And whilst for face recognition, there are easier work arounds, and indeed, in China, we've seen companies that have asked their employees to share photos of themselves wearing face mask, and based on that photo, companies have created thousands of simulated images of fake people in masks to be able to then train machine learning algorithms to recognise people with masks on. Researchers from Wuhan released a real-world mass face recognition data set to help people around the world do this. They've got 50000 images of 10000 fake people, and about 90000 images of 500 real people.
Kai So once again, this shows the limitations of AI when we do not have data readily available. However, we do want to point out that we can do something to prepare for the future, in order to break the cycle of panic and neglect. And that would be to bring in place a global infrastructure for data-sharing that protects privacy, but allows researchers in a pandemic, in a situation like now, to pull together data much more quickly and therefore train algorithms to be used at a time when it's actually needed. And we also want to point out that there are some real uses for AI when it comes to fighting COVID-19.
Sandra Let's remember, AI is helping in this pandemic, it's helping with discovery of drugs and understanding how the virus works. Since we have previous data on similar viruses that can model the ways in which this virus interacts with drugs, or interacts with people's bodies. We also have good modelling around how viruses spread in communities. So AI has helped understand how such viruses would spread through communities, and help contain that spread.
Kai So it once again shows that AI is all about the data, and where we have previous data that we can bring to bear on the problem, AI is immensely useful. On the more novel aspect of a problem such as COVID-19, it takes quite a while for these algorithms to be trained and data to be available.
Sandra But that's all we have time for the day, we wish you happy holidays.
Kai Tune into Corona Business Insights and of course, The Future, This Week.
Sandra And thanks for listening.
Kai Thanks for listening. See you soon.
Sandra On The Future....
Kai Next week.
Sandra This week?
Kai Yes. But next week.
Sandra On The Future, This Week. Next week. Thanks for listening.
Kai Thanks for listening.
Outro This was The Future, This Week made possible by the Sydney Business Insights team and members of the Digital Disruption Research Group. And every week right here with us, our sound editor Megan Wedge, who makes us sound good and keeps us honest. Our theme music was composed and played live on a set of garden hoses by Linsey Pollak. You can subscribe to this podcast on iTunes, Stitcher, Spotify, YouTube, SoundCloud or wherever you get your podcasts. You can follow us online on Flipboard, Twitter or sbi.sydney.edu.au. If you have any news that you want us to discuss, please send them to email@example.com