PassW0rd – 8th November 2017

Speaker A: This program is brought to you by Resonance 104.4 FM. If you like what you hear and want to support our work, please make a donation at fundraiser.resonance.fm.

Speaker B: Hello, how are you today? What are your metrics telling you about your state of health? I’m Peter Warren, feeling a bit tired and run down actually. Luckily, I don’t have any monitoring devices tracking my sleep, or pedometers counting my steps. If I ever need to do more walking, a small white dog will let me know. Quiet, Sid! So this is Password, and in this show we’re examining some new apps and devices and data pools. They promise to make us healthier, saving lives and money. It’s our money, since we all pay for our embattled National Health Service. And assistive technology promises to deliver swifter, more accurate diagnosis, as well as improving the quality of life for older people and those with chronic conditions. To do that, of course, These gadgets have to collect our data. Some of it is really quite intimate. The frequency of our bowel and bladder habits, for example, and women’s menstrual cycles, and what we all do in bed. One of the UK’s top scientists is leading a high-powered team to investigate the ethics of collecting, storing, sharing, and using all these facts and figures from sensors in our underwear. Implants under our skin, or wearable tagging devices. Sir Nigel Shadbolt is investigating Respectful Things in Private Spaces: Ethical Data Handling for Very Personal Devices— RETIPS for short.

Speaker C: When we present this, we talk about this spectrum of types of data from open, non-personal, through to closed, highly personal. And, and as you shift through that set of interests, you need to understand how Open data can have huge benefits, so the anonymized health data that we know of can make real benefits to a health delivery system and to researchers and to actually our own understanding of what’s going on around us. But at the point at which I am now starting to emit and generate and collect my own, not just health data, but fitness data, well-being data, what’s the deal between my generation of that and ownership of it and various third parties using it? And we have open in the title, but that doesn’t mean everything is open for everybody to exploit and plunder. The whole notion should be here that we are putting individuals in control where it’s appropriate. If data is a new asset class, then people need to understand their rights and responsibilities to that asset class. And again, health is a very good example of that. And we can imagine situations where you as an individual make an informed and conscious decision that you will share rather sensitive data about your current medication, your current medical context, because you’re in need of acute and ongoing analysis, and you may well be prepared to share that with a cohort of patients like you. And so the PatientsLikeMe movement, the people who try and consent people in for a wider good that also benefits the individual, that level of control, people kind of do understand that when they engage in medical trials, when they engage in volunteering their data, but simply assuming that this, because it’s come off device X, is property of company Y, will not do.

Speaker B: I think that’s really one of the issues, isn’t it? The biggest issue about all of this is knowing how your data is being used.

Speaker C: We interviewed somebody from a pan-European project, and what they’re asking people to do is to put an app on their phone which gives data about them. It sits there and just gives data, data about them. The reason that they’re doing that is so that they can collect people information from normal people, but also from Parkinson’s sufferers, so that they can actually find what people are likely to develop Parkinson’s and whether there is any commonality. In the same way that in the National Health Service we’re familiar with giving blood, this seems to suggest that we can give data, we can donate data. Yes, and I think people would understand that, but the— and this is a very clear aspect of the incoming General Data Protection Regulation, which is is to have explained to the data subject what is going to happen to the data. The right of that individual to request that data not be processed in that particular way. So there is some quite useful regulation that’s going to start to structure this landscape, and rather than seeing it as a great big threat, what we should be doing is regarding this as starting to rebalance the data landscape and saying, yep, okay, now how do we enroll people into the kind of study you’re talking about. How do we persuade people that actually for some of this we need their, if you like, anonymized normal data? So what does the base level performance of those people who don’t have a condition look like? So what are the natural statistics of you carrying around these devices when you are reasonably normal in your health before you can begin to then compare and contrast the pathological cases? But the conversations needs to be, I think, rebalanced towards that end. And people say, oh, consenting is all too difficult. We see the consequences of uninformed individuals who, when they realize what has happened, withdraw their consent en masse. And that can’t be where we want to get in many of our public service contexts.

Speaker B: Sir Nigel is, with Sir Tim Berners-Lee, a co-founder of the London-based Open Data Institute. And one of the inventors of the World Wide Web. He’s in charge of a center of excellence called the Petrus Internet of Things Hub, which combines the UK’s top researchers from 9 leading universities. By an odd coincidence, Nigel is also the name of an artificial intelligence— not an AI, but an AGI, artificial general intelligence— created by Chimera Systems with no specific purpose, Chimera claims. As an AGI technology, Nigel determines required functions through unsupervised learning by observing sensor data. Nigel will fuel an explosion in peer-to-peer transactions and bring about a greater transformation in global business than even the internet.

Speaker A: Moon is shite!

Speaker B: The founder of Chimera Systems told me that economics, physics, and probability, not biological process, are the secret of his company’s success.

Speaker D: Neuroscience obviously is a part of biology. What we are arguing, what, you know, when we started doing this research and started looking into quantum mechanics and started realizing that we believe that intelligence is, for the lack of a better word, and I want to stress that, that for lack of a better word, it’s some sort of a force that decides how these concepts of particles, whatever you want to call them, collapses into an experience that we go through. Which version of these futures do we actually experience? We are now arguing, and at some point we will release a research paper The only problem is we hate writing and we don’t have time for it. But we are arguing that intelligence is a part of physics and space-time, not biology. And that’s a different approach than I would say the most of AI researchers out there. And so we are more focusing, you know, people always ask me, well, what kind of tasks does this technology do? What we concluded with is that actually doing a labor task, which is what most people actually think of when they ask what kind of task this AI does, doing a task is not important. It’s the result of the task that increases or decreases the probability of getting to certain versions of the future. And we know that. We know that for a fact because people today are not dumber than people 100 years ago. Because today we are outsourcing tasks to machines and to stuff we build. That doesn’t make us less intelligent, but we are doing that. We are building machines to take on some tasks, and the only thing we really care about is the results of each task, not the task itself. So we are arguing that in our minds, when we, you know, when we live our everyday life, the process of thinking is— and of course, this is just a theory, we can’t actually prove it— the process of thinking is us humans doing what we call quantum modeling of the future. We are trying to model every version we can think of in our heads, and we’re thinking about the actions we can take, the reactions that might come, what other actions we can do after that, and we’re basically simulating the future. And once we find a path to whatever goal we want to achieve, that’s when we do the action. Only after we simulate. Our own future in our head. And I think that’s what artificial general intelligence should be all about. It’s about machines observing the world, comprehending the world, being able to do quantum modeling, figure out which versions— or I’m sorry, which actions to take.

Speaker B: That’s the theory. In practice, businesses and government are banking on AI to improve the efficiency for example, of the National Health Service, by using it to find patterns in the great big pools of data they can collect from patients’ records. We know, because the International Red Cross has told us, that the current state of the NHS represents a humanitarian crisis. So how can AI, which is just a set of mathematical equations, save us? Password reporter Charlotte Roundtree put that question to Jacqueline de Rojas, the chair of the Digital Leaders Board and president of TechUK, the government’s think tank.

Speaker E: It’s a really great question, and I’m not sure I’m into saying what’s wrong with it, but what I am saying is that with all of the pressure on the health service, we’re living longer, We certainly are using more services from the NHS, and with austerity on top of that— so it’s being used more, there is less money— I would say that the only state we can afford is a smarter state. And what that means is that we need more automation in services so that we can improve and extend our reach in terms of services into the citizens of this country. Obviously we are in times of austerity. How do you think we could get the message across that in times of austerity we would need to change, digitise the NHS? Surely it would be quite an expensive thing to do. It’s interesting that you say that. Is that the case? I mean, I think that’s the case if we do what we do. If we do more of the same, I would tend to agree with you, but If we’re going to disrupt our thinking and change the way we consider how we could deliver public services, you know, we think about even in our consumer world, 85% of all consumer interaction will be with machines by the year 2020. And if you overlay that onto the NHS, then maybe we can do things differently. Rather than do more of the same and think we have to grow it that way. I think technology gives us the opportunity to disrupt and rethink how things are delivered. It’s quite a broad term, automating and digitising the services we’re talking about. What more specifically do you think we could use technology for? Well, I’ll give you a great example. My mother has Alzheimer’s Alzheimer’s dementia, and you could argue that she needs reasonably high-intervention care. On the other hand, with machine learning and AI and robotics, the fact that she asks Alexa 24 times a day what the time is— or 24 times an hour, actually, she would ask— you know, the great thing with Alexa is Alexa doesn’t get cross that she’s asked that many times. A human carer from the NHS might think that that’s not an important thing to do, and we probably don’t have actually the funding to deliver that kind of attention to someone like that. So in a way, we can see that machine intervention could help in situations like that. But in general, accelerating and optimising routine admin processes is clear. That’s clear where technology could help. There’s a lot of citizen transactional relationships with the NHS, with government, spanning everything from paying taxes, claiming benefits, prescriptions, all those kinds of things where we could use technology to automate, and we could then use AI and machine learning to automate increasingly sophisticated tasks. So I think it could remove quite a lot of pain points for people who find the current service slow and inefficient. So, um, a lot of people do think that technology will take away a lot of those human interaction and care that we’re looking for, but you think it may be a bit different where it might allow of a helping hand to people, say, with Alzheimer’s? There are very clear cases where humanity needs to be writ large in communications and interventions, of course, but there is the case where surgery is being, being affected remotely. We have access to the best, best, best surgeons without actually having to travel. So I think we just have to think in an open-minded way and not worry about the way we’ve always done things. And consider that a large percentage of what services we’re offering could certainly be delivered with an augmented machine-based backbone, and then 20%, let’s say, could be very high-intervention humanity, which is very important in lots of cases in the NHS. I think there is a big opportunity for machine-based backbone in the NHS where we can deliver services to patients who really could apply for repeat prescriptions, remote surgery, all of that. There’s a big rainbow of intervention that we could use, but maybe it’s human intervention by exception rather than by default, and that would enable a much higher increase in public sector productivity and maybe remove pain points for people who do find the public service from government and the NHS slow and inefficient. But I also offer an example in Norway where they are tagging patients electronically, and what that does is the consultants, I think, are saving something like 2 hours a day because they’re not going to the bedside, the patient isn’t there, and then they’re wondering, oh my gosh, where they are, and they’ve got to go and find them. No, they know exactly where the patients are because they’re electronically tagged. And I think that’s fascinating in terms of the sorts of things we can do to help both the patient and the consultant side in NHS services. I hadn’t heard of that, but that’s definitely an interesting take on things. So healthcare services aren’t quite as digitised as maybe they should be. Why do you think the UK is such a good place to start talking about this kind of change and innovation? Well, I do think necessity is the mother of invention, and because we are going to run out of money to fund the NHS, A&E, Accident and Emergency, is under enormous pressure. And we can use technology like we can use drones to deliver blood services rather than roads. It means much, much more efficient, and it starts to create the kind of optimism I think that we need back inside people who are reshaping it, because we can really make a difference here. But if you ask why, why the UK, I think because we have to. We’ve got no option. As I said earlier, the only state we can afford is a smarter state. And a smarter state with a capital S, and that means highly automated, I believe. We have huge amounts of technology these days which can monitor so many metrics in your body that as long as you are wearing the device, then I’m told even heart attacks can be predicted. So we can raise standards through the use of technology. But we do absolutely have to create more educational investment and input into ensuring that nobody’s left behind from an inclusion perspective.

Speaker B: Jacqueline de Ros and her experience with an elderly mother sliding into dementia is all too common. It inspired author Paul Kitcat to imagine a future in which friendly robots would be able to care for our elderly relatives. Perhaps even better than we can.

Speaker A: I’ve written a novel in which I dramatize the arrival of robot carers in a care home, and of course while I was writing it, during the course of writing it, robot carers did indeed begin to appear in care homes and various other robotic developments took place. So I felt while I was writing that I was very much on the zeitgeist, which was an interesting sensation. But it was originally prompted by the fact that my mother, who died 2 years ago, was herself in a care home at the end of her life. And really the whole subject started in my mind because I was there one day feeding her— she had lost the ability to feed herself— and I observed the level of care that she was getting in the care home, and for no No fault of the carers who were employed there, they simply weren’t able to give her the kind of attention that would have really helped her to combat her declining powers. So that was the start of a thought process.

Speaker B: One thing that’s interesting in this though is, do we therefore then think that AI is better than people at doing things?

Speaker A: I don’t think it’s a question of it being necessarily better. Well, I think we will be before very long seeing quite human-like robots appearing among us. But if we go back to where we are right now, it’s very obvious that old people in care homes spend quite a lot of time effectively fending for themselves, and that’s something they can’t really do. And I think it’s also pretty clear that one of the reasons for cognitive decline is the lack of social interaction. I think that all the kind of puzzles and games that people like to think of as keeping them sharp, they’re all very well, but the thing that really keeps the brain on its toes is contact with other human beings, and particularly with intelligent, socially minded human beings. And the carers in care homes may be very intelligent, but they’re not socially minded because they simply don’t have time to be. And so the residents really are left to themselves. Anything that came along that would provide them with some kind of stimulation is a good thing. And if AI can replicate some of the, uh, things that you’d expect from a human being in a social situation, then that would be very good for the residents of the home. The robot carers that are currently appearing aren’t that sophisticated. They don’t probably have conversations with old people that are particularly like a human conversation, but it’s a step in, in an interesting direction.

Speaker B: Do you not think that there’s an issue here? Aren’t there sort of dangers of anthropomorphism? Aren’t there dangers of dependence that are going to occur as a result of this?

Speaker A: The point you make about anthropomorphism, I think, is very interesting because in the long run, it’s pretty clear that scientists are working, or some scientists in robotics are working towards creating robots that are very like humans. And at that point, it’s not so much that we will anthropomorphize them as they will become very anthropomorphic to the point where they may be indistinguishable from human beings. And that’s—

Speaker B: there’s already that professor in Japan, isn’t there, who’s created a, uh, a model of himself that does look frighteningly like him?

Speaker A: Yes, yes. And actually, one of the things that I read while I was writing my novel was that there are a couple of things that really make this work, and one of them is that we tend to form emotional bonds with machines especially if they’re slightly unpredictable. So if they’re inclined to not be perfect, we tend to bond with them better. And I was— when I read that, I was reminded of my first car, which was a very erratic Morris Minor, and it used to go wrong quite unpredictably, and I loved it. And I’ve never really loved another car as much as I loved that one, even though every other car I’ve had since has been far more reliable. So that was one thing. And then the second thing that they said was if you have a a robot that has synthetic skin, the most important thing is that the skin should have, it should have a temperature that is close to human, and also that the skin shouldn’t be a uniform temperature. If you think about human skin for a minute, if you touch the top part of your arm, it’s warmer than the underside, for example. And they were working on the idea that if you have synthetic skin that has that in it, then a robot would feel very like a human being. So if you have variable temperature of the skin plus slightly erratic behavior, you end up with a robot that people are going to get emotionally attached to and think of as quite human.

Speaker B: But this is interesting, isn’t it? I interviewed a Cambridge academic, uh, several years ago, uh, when we were talking about this, and one of the points that he was making is that what makes humanity isn’t the, uh, the, the accuracy of the robot or the computer. It’s its frailty. It’s its ability to create errors. It’s that fallibility that gives us humanity. And for some reason, we’re trying to eradicate that. And as you say with your very touching story about the Morris Minor, it— that’s because it had some character. And perhaps your latter cars had less character, that they’d had that ironed out of them because of their reliability. Do you think though that the robots then, as companions, what you’re suggesting is that we do need to put some sort of fallibility, some quirkiness. We may actually need to have a robot carer that’s a bit like Nigel Farage, so that it irritates.

Speaker A: I think that would be a very extreme case. But I think, yes, some fallibility would be required. And in fact, in my novel, part of the novel is narrated by the robot, by the synthetic human. And one of the things that she tells the son of the elderly woman who she’s about to care for, if he gives her permission, one of the things she says to him is, I do make mistakes, but not major ones, because I’m connected via a kind of version of Wi-Fi to everyone else. So if anything catastrophic goes wrong, it’ll be stopped immediately. But small mistakes and small errors and slight fallibility makes me more endearing and more human-like.

Speaker B: That’s Paul Kitcat, author of We Care 4 U. It’s fiction, but not fanciful. In Japan, robot carers already exist. There’s one called a Robear with a cute face like a giant teddy. Residents of some Japanese care homes also enjoy relationships with robotic pets like this cat purring away in the arms of Password’s Jane Wyatt at London’s Science Museum.

Speaker E: Who’s a pretty pussy then?

Speaker F: Who’s a pretty pussy?

Speaker E: Who’s a pretty pussycat robot? Just Unbelievable robot pet.

Speaker B: Hmm, a cuddly catbot that purrs when you tickle its tummy. Might sound like a poor substitute for a real cat, or better still, a real human being. But when you are lonely, these interactive machines can really help. Dutch entrepreneur Michael Rouse goes a lot further than that. He claims that his new psychology app can help save lives by talking people round when they’re suicidally depressed. It’s not a catbot, but a chatbot called Tess. You can send it a text when you’re feeling low.

Speaker F: In general, that’s one of the main things that helps people. So it’s certainly a part of what this system does as well. Beyond that, what the system also does is it uses these things we call coping mechanisms. So This is, for example, when someone goes to a psychologist and they’re in a session and they’re talking about loneliness and isolation and that they feel anxiety to break out of it and to try to reach out to other people and to try to find a way to make themselves feel better. And in such a situation, the coach or psychologist would simply think of a way to talk with this individual and to see if they can think of how they could perhaps make them feel better. And if there’s a certain way that they can maybe think of convincing them to, for example, go out and go to a meetup where they can meet other people or do some volunteering or That’s really more of the sort of practical-based advice. It’s often part of cognitive behavioral therapy. A totally different way of making them feel better would be trying to make them look at the situation from a different angle. So if someone’s feeling really sad about something, you could also talk, talk through the situation and try to go some steps deeper and try to figure out, like, is that really what’s making you sad? Or is it really that, for example, in this case of loneliness, is it really perhaps that you’re still grieving about the loved one you lost? And it might be far deeper. So those are all examples of different coping mechanisms and different ways of figuring out what’s really going on and in which way you can make someone feel better. And those coping mechanisms, those small conversations are written up by psychologists and then uploaded into the system. And when the system finds out that someone is feeling very sad about a particular topic, then there will be numerous different conversations available that can be used to talk through that situation and make that person feel better.

Speaker B: So, I mean, one of the things that everybody listening to this will be saying is, how much does this cost? You know, America is a legendary place for psychologists and psychiatrists, and it’s also legendarily expensive. How much does this service cost?

Speaker F: In general, we work with— in a business-to-business context, So we work with large hospitals, we work with large groups of community mental health centers and also employers, and they pick up the bill to make it available for free for the end users. So for the end users, it’s for free and their providers or employers pay for it.

Speaker B: Okay. So somebody goes along to a psychiatrist— sorry, a psychologist in America, and they say, okay, ‘Yeah, but as part of my service to you, I can offer you this service as well, and that’s going to cut your bill, but it also means that I can get a lot of information about you and help you in a much more targeted way.’ Yeah, it’s really indeed meant to improve the quality of care and also very much to improve the access to care.

Speaker F: ‘Cause the system is also implemented, for example, for a large public health department and their hospitals. And there are a lot of patients who go to these hospitals and they are diagnosed with a chronic illness or potentially even multiple. And then their disease becomes more of a burden because they’re also feeling, having some feelings of depression and anxiety. Which makes it harder to stick to the exact regimen of taking medication, which makes the disease worse, which then in turn makes their mental health, their mental state worse as well. And that’s also a context where TESS is offered and people can simply start using it like that, not even with seeing a psychologist at the same time. But prescribed by their doctor and just getting access to mental health care for the first time. Because like you said, it is often very costly, and one of the main reasons people don’t seek care is treatment cost. And in those contexts, often, especially here in the, in the U.S. with Medicaid and Medicare, these are groups that often do not have enough money to have access to Behavioral health services. And in this way, they very much for the first time get access to behavioral health services. And a large part of helping them and making them feel better is done by our system. But at any time when they need more help than talking with our system, then some psychologist from that organization takes over the conversation. And can help them or can schedule an appointment with them. And in this way, their resources are very much used in the best way for really helping people that really need in-person therapy. And all the people that can already get help and feel a lot better by using our system will be helped by our system, which allows the hospital system to suddenly help so many more patients at the same time.

Speaker B: So why did you decide to come up with this system? Why was it that you suddenly said, hey, I know what I want to do, I want to come up with an AI system that helps people who’ve got mental health issues?

Speaker F: Yeah, so it very much started with myself. I, between when I was 10 and 20 years old, I was diagnosed with 3 different chronic illnesses. And that was a bit much to take in at the time. And luckily one of my doctors told me, hey, I know this is hard and I know this is tough, especially because you don’t know how these diseases will develop. But if you manage, if you can manage your emotions better and if you can make sure you don’t experience too much stress, then you might be able to influence the course of how they will develop. And that’s then what I did. I went and see a psychologist and for helping me learn to manage these emotions and also for some feelings of depression. And that really helped me a lot. And then in the years following, I realized that I was often talking to friends and colleagues and then especially when they were feeling bad, I was able to make them feel better. And I was really just repeating the things that I had learned. I was repeating these coping mechanisms that my psychologist had taught me, that my psychologist had been talking through with me. And I was just helping them in a similar way. And this is also a concept in literature known as being an expert by experience.. And that’s very much where that idea came from.

Speaker B: So when— I mean, can I inquire what those chronic diseases were?

Speaker F: Yeah. Epilepsy, inflammatory bowel disease, and psoriasis.

Speaker B: Okay. And one of the things that we’re told is that mental health problems are going to grow within our society. Nobody seems to really know why, but we’ve been told to expect an epidemic of things like depression. Does that mean, do you think that systems like yours will become part of our daily lives?

Speaker F: Yes, because that’s really the issue I saw in Europe, but even more when I moved to the United States, is that there’s the demand is just so much bigger than the supply and the supply is also not scalable. So from the beginning of ages, people have been doing one-on-one therapy face-to-face, and later they found out that group therapy also works very well. So one therapist and maybe 5 patients, and that’s really sort of where it stopped with regards to scalability. There, there were of course self-help books, but they require a lot of motivation, and those turned into online Self-help courses also often require quite some motivation. And that’s why I really believe, because I saw how I was able as an expert by experience to help people, that I believed, hey, nowadays these conversational systems are getting much and much better. And if they’re only somewhat as good as I am in having these conversations and Basically talking about the right topics and talking through a situation in the right way at the right time, that’s the most important part. And that’s really not too hard of a challenge. The harder challenge is to exactly understand every single nuance of what someone’s saying. Yeah, that’s really when we started building and started testing and realized, hey, this is actually doable. Getting diverse results and, hey, people are appreciating this and this is helping people. We’re on to something here.

Speaker B: And another health tech startup business that believes we’re on to something is Amir Butt’s Tumor Trace. Using a small handheld device, the technician can spot microscopic changes in human tissue that give an early warning of pre-cancers. And diagnose actual cancers more swiftly and accurately than a conventional pathology lab at a fraction of the cost. Amir Butt believes in countries like India where there is no National Health Service but only private paid-for diagnostics, this new invention will save many lives.

Speaker G: Tumor Trace is a company with a technology which allows you to characterise the presence of pre-cancer and cancer in tissue that’s presented to the device that we have. So in less than 2 seconds you can get the characterisation. There is a change, there’s a magnetic change that occurs in cancerous tissue as it becomes more and more cancerous. You can actually pick up this change using things like atomic force microscopes and magnetic force microscopes, and we have a patented method which allows you to detect the same kind of profile change, magnetic profile change, using the interactions of a photon of light and valence electrons— electrons that are on the outer shell of elements. So, you know, biological molecules, as they begin to become more cancerous, the properties of the valence electrons change because different compounds are created There are changes in the proteins as the cell becomes cancerous, and those are reflected in changes in the structure and the layout of valence electrons. And valence electrons are just electrons on the outer shell of molecules and atoms.

Speaker B: Right, okay, so they’re electrons that are on the outside and you can see those, and because you can see that change, you can pick it up.

Speaker G: Yeah, it is quite a genius invention that we have because As I said, you can detect the same thing using expensive and bulky equipment that you normally find in laboratories, nanolaboratories. You know, there can be anything from 300,000 to 500,000+ machines, and you can detect the same kind of change in tissue, but obviously it’s not very practical to have those machines in clinics and to be carried around into sort of villages, etc.

Speaker B: Oh right, so this is a very, very portable technology that you’ve come up with?

Speaker G: Yeah, so you can have it in hand luggage.

Speaker B: Right, so what size are we talking about?

Speaker G: Think back to your school days and when you had a microscope, it’s about that size.

Speaker B: It’s incredible. And AI is very important to this. What does AI do?

Speaker G: I’ve said that there is a change in magnetic profile that we can pick up, and we use sort of quite advanced physics and mathematics to determine this magnetic profile. That magnetic profile, you can graph it, and actually you can see very, very clearly what that— the, the graph of normal tissue as opposed to that tissue as it progresses through the stages of cancer. You can, you can visibly determine, because of the different shape of the graph that you get back, that there is cancer present in the tissue. Now, so that’s, that’s a visual interpretation. If you want to use the data and classify that data using automated software. There are various ways that you can do that. What we do is we use artificial intelligence and deep learning neural nets to characterise the underlying patterns that exist in the data. And actually, because we have a very, very strong underlying pattern depending upon what stage the cancer is at, we can train our learning algorithms with a very small dataset. Normally you need thousands and thousands of datasets to actually train your algorithms, but we get quite good accuracy with a very small number of training set.

Speaker F: Okay.

Speaker B: So what this basically means then is that you’ve got the records from previous analyses of tissue that you’ve done, and you’re able to very, very quickly map those across the analysis that you’re currently doing, and that says, yes, this is something we should look at.

Speaker G: Yeah, so like all artificial intelligence algorithms, you need to train them with existing datasets that you know the outcome for. For example, if you’re training your algorithms to recognize letters of the alphabet, you present it with different images of the letter A, and then you tell it that this is a letter A. And then eventually you have— you’ve trained your algorithms, and the next step is you test the algorithm to see how accurate they are. And then you present something and you say, what is it? And if it correctly characterizes it, then that’s a good result. If it’s not a good result and if it does it wrong, then you feed that back into the algorithms for the algorithms to be trained. And so once you’ve trained them, then you your algorithms, all the different parameters that you have in your sort of network of variables are set, and the idea being that they’ve learnt through feedback loops to determine the letter A very, very accurately. And it works in the same way as we use the algorithms. And over the next 3 or 4 months, we’re training the algorithms on a much larger dataset than we have of 4,000 from Graz University in Austria, and we fully expect our current accuracy is around about 90%. We fully expect our accuracy to be over 95%.

Speaker B: It’s a new technology that’s still at an early stage of development, but when TumorTrace is rolled out, as they say, it will become even more valuable. Incredible as a way of collecting data on how cancers develop. Where, when, and in what sections of the population. This is all information that could help to save many more lives beyond the actual patients who are involved by finding patterns pointing to causes and informing research. One patient in Switzerland, Pierre-Michel Legris, has become a strong advocate of this mass data collection. So much so that he has invented an app to enable him to collect his own vital measurements instead of returning to hospital for tests while he was recovering from leukemia and a bone marrow transplant. When I spoke to him, I observed that he was becoming his own doctor by taking charge of the these test results.

Speaker C: Exactly.

Speaker H: So as soon as you need to do computational analytics such as AI, and as soon as it is about personal data, then you have to comply to some, I would say, obvious questions. Typically, to make sure that the subject of the data has the same rights than you might have as the owner of the data. So if the data describes your life, clearly you have the right typically to know how the data is used, how it is processed. Typically an AI machine will process this data. You have other rights, so the right to erase this data in some cases, or to deal with the consent of sharing it with some other persons. It’s really I would say obvious, right? Exactly that, the same that you, you might ask for your bank when you put your cash into a bank. They shouldn’t spend it without your explicit consent, and it’s the same with data.

Speaker B: Okay, because I mean, this is a big issue, isn’t it? I’ve just been doing an interview with a guy who was talking to me about an AI system that they’ve put together that does cancer analysis. And one of the things that he’s been saying is that there are issues to do with data in Europe, that you’re not allowed to put data into the cloud, medical data, um, and that this has a— has big ramifications for doing AI analysis because you can’t get hold of information very, very fast. And He says that that’s one of the things that AIs need.

Speaker H: And the other thing is that anonymization of data for me is a kind of crime. Because it’s better to ask the consent, explicit consent of someone to use the data for a purpose. And if you anonymize the data, then you cannot give any outcomes back to this person because you cannot go back to the person because it’s anonymized.. And if you keep the link, then you can analyze the data and give outcomes to this person. Our task with PRIVE as a company is to provide all the tools, software tools, so you can collect this data according to different regulations. There is GDPR in Europe, HIPAA in the US, and other medical compliance structures.

Speaker B: Absolutely. Healthcare statistics should not be used for any kind of marketing purposes, nor for alerting medical insurance companies to the prevalence of any disease or life-threatening condition. When our personal data is being collected, we need to be sure that it will be used with respect to keep us healthy or make us better. And of course, the data pools also help to make products and therapies better. I’m Peter Warren with Password here on Resonance FM, exploring the ramifications of artificial intelligence in healthcare. After this, you can hear A World in London with DJ Ritu. We’ve been finding out about how artificial intelligence can prevent suicide, diagnose cancers, and help cancer survivors to take control of their testing and recording of results. In an even more ambitious scheme, the EU Commission has funded a team of researchers to spot the early signs of Parkinson’s using big data. Parkinson’s is undetectable in its early stages, but by compiling data datasets from thousands of people, some of whom may go on to develop the crippling disease, they hope to find remedies, if not a cure. Professor Leontius Hadjileontiades is the i-PROGNOSIS project coordinator.

Speaker I: Yes, actually the tests that are taking place in the hospital when you go for the medical evaluation are, let’s say, related with the movements and the use that you use of smart devices. For example, they have a test about tapping. So when you write an SMS, we measure the way that you write SMS in your mobile, not what you are writing, but actually what is the time that you press the key, what is the flying time, that is called when you go from one key to another, how much is the pressure that you press there. So these are indicators that relate with the fine motor skills. So this is one of the things that changes. Also the voice changes. So when you talk on the phone, we record the first 75 seconds and we analyze locally on the phone. So we get features from your voice and we send this, we take these features and actually everything goes to the cloud. So there are information that comes from the change of your voice according to the everyday use of your phone. And also when you walk around, how the walkability is exploratory or not. So this goes to the, let’s say, behavior of the depressive mode. Suddenly you go outside, not only nearby places. And also we analyze the selfies, so we— the frequency of the selfies, but also the probability of smiling in the selfie. So also the behavior goes to the depressive mode. So we analyze also some ideas when you go to the next stage of the project, of specific people we monitor with some smart belt where we analyze the bowel sounds, because the bowel sounds relate with the mobility of the bowel and the constipation, which is one of the very early, early symptoms of Parkinson’s disease. So there are information that comes from the everyday use, or in another stage of the program, specific patients or specific people we monitor at home and. And then we try to see alternative means where we use this IoT and smart devices in everyday, but with totally unobtrusive way to identify these indexes.

Speaker B: This would not be possible without the Internet of Things. This is a completely new treatment.

Speaker I: Yeah, it’s something that gives us a very dense, let’s say, sampling of information, because otherwise you couldn’t have this information information. And since we’re creating all this big data, we need definitely to have machine learning and artificial intelligence behind in order to construct models that are semi-supervised or almost unsupervised in order to identify if your behavioral vector is closing up with the prodromic symptoms of the Parkinson’s disease.

Speaker B: With all this big data promising a more personalized type healthcare. National Health Service planners may well be banking on lower costs and better outcomes from artificial intelligence. But of course, as with any new technology, cybersecurity is an issue. Hospitals are still recovering from the WannaCry ransomware attacks earlier this year, which saw thousands of patients’ confidential records locked away by cybercriminals demanding large amounts of money for their safe release. Tom Gaffney of the Finnish cybersecurity company F-Secure told me that the hackers were able to strike because most of the hospitals affected were running XP, an operating system that is no longer supported by Microsoft. We also subsequently found out in the National Audit Office inquiry that many of the systems had gone unpatched because other systems dependent upon them would have failed if the systems were patched.

Speaker C: Patch update that Microsoft, you know, they periodically put out their updates once a month, as you know, to protect against things that have been discovered that might cause some weakness to various platforms. But this one allows elevated privileges, which lets the software deploy its encryption onto device and therefore lock the content for the user. So the fact that these hospitals have been vulnerable means that they haven’t patched their software? Exactly so, unfortunately. Now, hard to be somewhat overly critical off the bat because it’s the sort of security dilemma permanently these days is the only way to keep things completely safe, of course, is to disconnect them from the internet, but the world just doesn’t work that way anymore. Although that does beg the raises the question, doesn’t it, that perhaps a network like the NHS network should be on a private network. Why is it internet-facing? It doesn’t necessarily need to be, does it? Well, again, the nature of the world is it makes sense that doctors can access information on the hoof, so to speak, or using systems floating around within hospitals. And that information, you know, by definition then needs to be available online somewhere.

Speaker B: Without enhanced security, putting more new technology and AIs into the health service seems foolhardy. Even before this year’s ransomware attacks, we know that more than 300 medical devices such as heart pacemakers, X-ray machines, and incubators for premature babies are capable of being hacked quite easily. We even reported the this on password 4 years ago after the United States Department of Homeland Security issued a warning. Add to that insecurity the unpredictability inherent in machine learning by robots and AIs that feed on patients’ personal data, and the future of healthcare begins to show its dark side. Still, as the writer Paul KitKat points out, there is nothing new about humans using dolls, effigies, or automatons. And just as we evolve, so too do our robot companions.

Speaker A: If you imagine that the development of robots would be to, to make them increasingly like intelligent human beings, then I think what you would expect to see would be more sophisticated interaction, let’s put it that way. And if they— if a robot with a degree of sophistication encountered my mother on one of her best days, they’d certainly have to be ready for an argument because she loved it. And so really, she wouldn’t have been very well cared for by a robot that just agreed with her. She wouldn’t have liked that at all. So it would have been quite an interesting combat. I think it’s a few years off yet, but not, not as far as some.

Speaker B: I don’t think it’s— I mean, it’s interesting, isn’t it? The Internet of Things, which is going to be gathering the data for the great AI revolution. Yes, the Internet of Things. A lot of of people have been saying, actually, we don’t want Internet of Things devices in our houses because I’d rather do the tasks that people are saying that they could do myself. Although there is one group that significantly says that they do want Internet of Things devices, and that’s old people. And these, the 4 reasons were given. The chief one was it meant that they could stay in their homes longer. Yes. The second one was that it allowed them to have more independence. Another reason was that they felt safe from being monitored and surveilled. And the final reason that they gave was it meant that they wouldn’t be lonely. It’s something of a paradox, isn’t it? Because if you think that you’re not lonely when you’re with a machine, then that is a bit odd.

Speaker A: It is odd. That’s a very interesting piece of research that you’ve quoted there. I can see easily how each of those things might work, but it’s the last one that’s the most dumbfounding, isn’t it? Because you can certainly see, yep, having someone keeping an eye on me, yes, that makes sense. I don’t mind, you know, having menial tasks taken care of, yes, I can see all of that. But what you’re telling me is that older people People are saying, yes, I’m pretty happy to chat to my fridge if there’s no one better to talk to. Well, maybe that’s how we all end up, you know, you’ll settle for whatever you’ve got.

Speaker B: And if it’s a talkative fridge, then okay. And then there’s another very interesting point again in this world of loneliness, which the government is predicting that we’re all going to be moving towards as more and more people get depressed and there are more and more old people. We picked up a bit of research where a woman had forbidden her child from talking to the Alexa, uh, house control system, and she said that she was— that the child was not to treat the, the system as though it was a, an entity. And the child had turned around and said, but why? You tell me to treat my teddy bear as though it’s an entity.

Speaker A: Yes, well, that’s a pretty good answer, isn’t it? Really what that’s reminding us of is that right from the start of our life, we invest material objects with personalities and emotions and everything else, and it’s our imagination that’s at work there. So it’s a pretty reasonable comment for her to make, isn’t it? Because Alexa does something that the teddy bear never does, which is actually answer back. Or say something interesting. So yeah, I think we just don’t— people think that we might find it kind of repellent to deal with human-like robots, but it seems to me that we, you know, we’ve grown up with dolls and facsimiles of humanity through our whole life, and why not, you know? If it’s convenient and interesting, I think we will accept it.

Speaker B: And if Tess Nigel, Robear, and co. can save lives as well as cutting costs without putting cybersecurity at risk. Then perhaps we should welcome them into our hospitals and homes. You can find out more about the ramifications of new technologies like these at our website, Future Intelligence, and we’ll explore them further in next month’s edition of Password. Thanks for listening.

Speaker A: And goodbye. This program has been brought to you by Resonance 104.4 FM. If you liked what you heard and want to support our work, please make a donation at fundraiser.resonance.fm.