 We're back. I'm Jay Fidel here on the 2 o'clock block on Friday on Think Tech Hawaii, and this show is likable science. Science is very important to us. It's our middle name, if you will. And one of the things we do is we talk about likable science with Ethan Allen, the host of likable science every Friday one way or another. Welcome to your show, Ethan. Good to be here, Jay. Nice to be talking to you again. You know, a lot of this is inspired by the MIT newsletter that goes a bullet and goes around. It's really worth subscribing to that if you want to know the exciting things happening in science and technologies. It's really a community service that they perform. And one of the threads that we see repeated over and over in the MIT letter is AI, artificial intelligence. And you and I have talked about that before. And one of the interesting things popped out what yesterday or today that you noticed, what was that? So they're pointing out that the Apple Watch, which is able to monitor your pulse rate, can actually now be used to predict if you're in danger of going into atrial fibrillation, which is a very bad heart condition where your heart stops pumping effectively and basically just sits there and quivers. And if that happens for very long, you basically drop dead. And this is, I mean, it's a very sophisticated prediction to be able to do that from a simple monitoring of your pulse, basically. Yeah. Well, but you know, this is interesting because this is an example of what the scientists do. They come on our show on the research in Manoa segment on Monday afternoon. And they talk about all the data they have. And a lot of these PhD candidates or research scientists, you know, they are using data they got from somewhere else, maybe NASA, about things happening in space and geophysics. And they, the project is not so much getting the data, the data they can get free most of the time, I think. Right. It's the interpretation of the data. Exactly. It's managing the data, putting it in the right fields and then, you know, sort of connecting the fields with the right interpretational algorithms and ultimately coming up with conclusions that means something. And so the value proposition, the added value is not in the data, but it's in the knee analysis of the data and coming to a conclusion that means something. Well, yes, but you can't do that without access now to these huge multiple sets of data that even two or three years ago we didn't really have. But the more and more data sets are being made more and more available and more and more software is available to look at, link and compare these multiple data sets in meaningful ways. Yeah. And so it is, right, it's allowing people to look at stuff that's sort of been there for anyone to see before, but to look at it in new ways and to be able now to extract new information out of it. Yeah. And one of the elements of that, I mean, this sounds like really basic stuff, but one of the elements of that is that we have a mature technology on sucking data out of a spreadsheet, which anyone can enter, right? You don't have to be a programmer or even an input operator to enter data on a spreadsheet, sucking that up into a sophisticated database program that will analyze it in any way you want. And it's that connection that makes it so easy. So A, we've got free data, tons and tons and tons of increasing all the time, getting more sophisticated. And then B, you can use it. You can suck it up into your analytical program in no time at all. And now you have some, you know, you have major possibilities. So graduate research and for that matter, faculty research is greatly aided by the ability to do that. Sure. You may recall about a year or so ago on my show I had Alon Benere, who is an anesthesiologist at the Vet's Hospital in Seattle, Washington. And he got curious about the issue of suicide among veterans, veterans of Iran and Afghanistan wars, because it's a huge trouble. More people, more of those veterans have now died from suicide than actually died in the fighting. And he then, and this guy, he's an anesthesiologist, right? He's not really a computer guru kind of guy. And he taught himself enough for this Python software, which is one of these packages that helps put together a search open source. Yeah. And it searches multiple databases. And he put together a system that looked at these people who had committed suicide or attempted suicide at their records and all the records and their interactions with social workers or interactions with police, their, you know, economic aid, their food stamps, their jail, anything that might suggest any sort of social disruption, a social fabric of a life as it were. And he was then able to come up with these people who committed suicide or attempted suicide, had these certain common patterns in these sets of data, which really were gathered for very different kinds of purposes. And what was very intriguing, of course, you could find there are other veterans who had not yet committed suicide who had these same patterns. Make predictions with that. And yeah, you would think these people now should be sort of targeted by the military for sort of special intervention to come up to them and say, hey, let me help you here. What's interesting is the data you're talking about in that context is subjective data. It's in text. And it's not like numbers. It's not like a real simple data table with numbers and certain phrases. It's text and it has to be interpreted to become data. So this goes back to our conversation with it last week, you know, about taking a large volume of text, hundreds of pages, what have you, and AI, artificial intelligence will interpret that and then summarize it for you in a paragraph. It's really quite remarkable. Well you can take all these records, these, you know, these records that are subjective about what a person did in his life and how he engaged with, you know, with police and other government offices, and you can make that from a pretty amorphous body of text into actual data about what happened. Exactly. Just as now, you can take and look, for instance, at social factors of an infant and see their socioeconomic status of their parents, educational status of their parents, the zip code they live in, issues like this. And you can make predictions, is this kid going to be ready when he enters, he or she enters kindergarten? Are they going to be up to par? Are they going to be lagging behind their peers? Are they going to be ahead of their peers? That's pretty scary, actually. Well, it's tremendously useful. It is useful. It's a question of putting on resources, you know. But, you know, when I say scary, I mean, you find out there are people who are not going to do well in kindergarten, and then the burden becomes that you know this is a matter of science and, you know, analyzing the data, and now you know you have to do something. Well, the question is whether government is going to be able to do that, or you've just, you know, blown in the wind here. Now you know there's going to be a problem, so have at it. There's no way to solve the problems. Yeah, but, I mean, almost all those problems are very classically well-established now that early intervention is far cheaper. Every dollar you spend on that child at age two, you're going to save $10 or $100 down the road, giving that adult social services later on, you know. So it makes for a persuasive pitch. You go and say we have X, you know, people in this population who we know as a matter of our analytics are going to run into a problem, so it's up to you to fund a program that will allow us to intervene. If you don't do that, here's the result. If you do do that, here's the result. Right. I mean, sometimes they're more obvious, the thing that came out a couple of years ago, that you can now download an app on your iPhone that will, you can put this app on and go to bed, and this app will basically monitor you as you sleep, and will tell you if you have sleep apnea. And that, again, there's a- At home. At home. At home. Or anything. No. So, yeah, before to get monitored for sleep apnea, you had to check yourself into a hospital where a whole skull cap of electrodes probably spends several thousand dollars to find out if this was happening to you, and now there's a downloadable app that will- Wow. They can do it, actually, even if you're sharing your bed with a partner, it can tell it with two people. It can tell if one of them has sleep apnea. Wow. I mean, it's very sophisticated. You, your partner, and that. Yeah. Right. That's reasonable. Right. Very scary. It's really important to health. In that case, it's easy to see the benefit, because both apnea and stress in general are critical to health. And if you don't do anything about it, your life expectancy will be affected. Likewise, if you have a stress situation where things in your life are stressing you out, you really have to be aware of that and deal with that, otherwise it will have an effect on your health and your mental state, which is part of your health. And I think if you had, you know, data from sleeping or data from some other source, we can talk about one of the sensors and sources might be available. And the app could analyze that and say, Ethan, you know, you're not in a good state of mind here. You have a lot of stress in your life. You have to do something, and it might even suggest- it's like one of those diet programs, you know? You have a lot of stress in your life. Here's what you have to do, and this is our recommendation for today. It's the kind of thing you have to avoid today. It's the kind of thing you should do. Because I think, you know, we live in a world that's very complex and filled for all of us. We're locked into this world. It's this highly civilized, mobilized and technological world, and it's very stressful for us. You can say you have a great life, but you probably have a lot of stress anyway if you live in this world. And so the problem is that's not good for you. That creates heart disease, that creates cancer, that creates stroke. So you have got to do at least something to avoid that kind of stress. It creates long-term health effects, not just in you, but for instance women who are under chronic stress or in pregnancy, it affects their physiological state of their fetuses and ultimately the health of their children. So I take your Apple Watch, and maybe we have to add other sensors to it, maybe not. And I find patterns in your pulse, maybe in your temperature, which the watch can probably read under the wristband, and I compare these against patterns that led to a bad place. And then it makes a little beep, and it says, you know, you're stressed, or you're having fibrillation, or whatever it is. Your insurance rates are going to go up. This is out of Star Trek. I tell you. It's out of Star Trek. It is. I mean, this is a real watch. Your watch is going to tell you, you know, what's the word. It's coming soon. Right. Ten minutes. Go call them that. Zone out, guy. But it goes to the fact that, A, we have, you know, an increasing ability to analyze data. B, we have tons of data. We know how to collect it and keep it, and open it up for access. And the next point is, what was it? C? Well, after upon it, I mean, you know, is that we are inventing new sensors all the time. Right. So if I say to you, well, I could probably tell temperature in the inside of my wristband. I don't think it's any question you could tell the temperature inside of your wristband. I'm sure if it isn't already the case that there will be detectors for diabetics on the Apple Watch that will be able from their sweat. Not invasive. Right from their sweat, will be able to analyze insulin levels and automatically feed back, you know, feed in insulin or whatever they need, or tell them to, you know, start showing down on the watch, you know. You feel a little poking. I mean, calling out of watches or what, calling your iPhone a phone, right? I mean, it's only one of its myriad of functions and probably the least important. Yeah. Yeah. Well, you know, we can sit here and figure out other sensors that we have. We have three of them now. We have pulse. We have temperature. We have sweat, analysis of sweat, which I don't think is any question, you know, given what's in the watch now these days with the apps and all that. In fact, given the fact that you could, even if you didn't have room in the watch to do all you want to do, you could report that to your phone. Right. And then your phone would do it. Right. You know, by Bluetooth, what have you. Yeah. So bottom line is, you know, health is really a great subject to have AI work on these few sensors, but maybe additional sensors too. Oh yeah. Yeah. It's exactly. It's a huge area where you've got a lot of this very messy data, right? You've got data that's analog. You've got data that's sort of emotional, verbal data. You've got something that's very unique quantitative data, and you somehow want to smush all this stuff together and, as you say, pull meaning out of that and mine it for everything that's hiding in all that big mess. Yeah. Let me go some more. So what time do you get up in the morning, right? I get about 5.30. Well, the watch knows. Right. You know, assuming you wear it. Right. What time do you go to sleep? How long are you in bed? Mm-hmm. When are you eating? Maybe the watch can figure that out too, from one thing or another. These few little sensors, these few little data points, can lend so much the noise level in the room. Is it really? And your voice. It can hear your voice. Can it? Right. It can tell the stress in your voice or whatever it is it's looking for. As I say, my wife was recently driving a rental car that began urging her to stop and get a cup of coffee. Apparently it was detecting some changes in the way she was driving and suggested that she was getting tired. Yeah. And her car. It's a car that doesn't even know her. Yeah. It's a rental car. The valuable point there is that this is growing and it's all because we can interpret this data in a way to be very useful for us. Right. And so you start out with, what would be very useful, then you do the interpretation. Yeah. What's useful to me is it's not just to watch. It's everything. Oh, it's everything. Right. Yeah. It goes back to our conversation about having government work and little blockboxes about to. But square. I mean, you know, even, you know, a few years ago they began for elderly folks having, essentially, home monitors that would detect when they're up and moving. It would detect when they would get up from bed or go lie down. So that, again, a healthcare monitor at a distance can say, oh, Coenso is now up and it's one o'clock in the morning. They don't usually get up at one o'clock in the morning. I wonder what's wrong. Exactly. And call them up. Yeah. Well, there was a loud sound and it was on the stairs a minute ago. Maybe he fell down the stairs. You know, in the old days, not too long ago, a lot of this technology for elders was pretty clumsy. You weren't around your neck. It was a big thing around your neck or in your, I don't know, around your waist. It was silly. But now, with miniaturization, you don't need all that. We moved on. Sure. I mean, if you recall, probably when the cell phones first count, they were big, chunky things that did one thing and one thing only. You could make phone calls on them sometimes. Yeah. Yeah. Well, you know, we've been talking about, you know, miniaturization, we're talking about watches and clocks and time and interpreting time and it just makes me feel we need to take a break. Ethan Allen, the host of Likeable Science, will be right back for this further discussion about AI. Ethan Allen and me, Jay Fidel, here we are talking about AI, artificial intelligence and how it comes together with newfangled sensors of all kinds and very small sizes that do amazing things. And scientists and research fellows and scientists who come up with ways to interpret that data and make it useful for us. And that seems to be happening increasingly. You were talking about geology a minute ago during the break. Can we talk about that for a minute? Sure. So, I mean, it used to be to get genomic genetic data. You needed real tissue off somebody. You needed, initially, it needed to be fresh, good tissue off of somebody, you know, off of body. Then it got to be where you could pull us out of old bones, ancient bones, very ancient bones, dinosaur bones. Now, I was just reading, the paleontologists can dig in the floor of a cave and out of the dirt there, essentially start doing genomic analyses on the dirt and pull out human genes that will allow them to say what kinds of people lived when in this cave, whether they were Neanderthals or early Homo sapiens or one of our Australopithecine ancestors or whatever. Just amazing stuff. We had somebody on the show two weeks ago, see Emilio Herrera was his name. And from UH, from so west at HIGP there, he deals in magnates, paleomagnetism. And using, you know, the phenomenon of magnetism, which is complex and very powerful and ubiquitous for sure, you can do dating like that. You can do a whole historical dating over millions of years and find out exactly what happened, get a history, sort of a paleontological history of something. And that's just, you know, of a magnet. I mean, essentially keeping the data, working the data, getting other data. You know, it's sort of like a cocktail idea. Let me show this at you. If I have one drug, I know pretty much what the side effects are. The FDA made them tell me. And if I have two drugs, well, there's something could fall between them, and maybe I know the one and know the other. But I don't know what happens when you put them together. Right. And so three or four or five or ten, okay, ten drugs. Well, who knows what the side effects of the cocktail are. Right. Ten. So the same thing here. If you had just one data point, one data table, looking at some kind of, you know, sort of simple data, one dimensional data point, you know what you were doing with that. But if you're really ambitious, you want two or three or five or ten. And when you get to ten, you have a cocktail of so much data that it bottles the mind of what you can study and learn from that data. And that's where we're going, isn't it? Right. And this is all, actually, if you consider, this is all sort of what they call systems thinking, right? Systems, one of the properties of systems is you put a bunch of simple little bits together and get them interacting a certain way and sort of magical things happen. They call them emergent properties, right? For instance, after the Big Bang, we have time and space as properties that emerge that did not exist before the Big Bang, you know. And now they're here for much of part of our lives. You couldn't have predicted it from what went before. And similarly, I mean, as you, once you guess, as you put the ten compounds together in your cocktail or whatever, sort of magical, almost amazing emergent properties start coming out, the richer and richer your data input is and the clever and cleverer algorithms are to extracting those. Yeah. And extracting and predicting, right? Predictive technology is the most exciting. And there's a whole field of mathematics about how you predict. It's not so easy, but there is a field of mathematics that specializes in predictions. And that, you know, sort of, I think is changing or will change our world. Oh, yeah. So I was just reading before I came over here in Los Angeles. They now have this thing called Predpaul predictive policing. And they tested this in one of the 31 districts in Los Angeles during 2013 and 2014 using this. And this system looks all kinds of not just where crime occurred yesterday and two weeks ago and all, but it literally it's looking at traffic flow, it's looking at temperature, it's looking at rainfall patterns. Pulls together huge sets of different kinds of data and then tells us where crime is likely to occur in the next 12 hours. Stay away from there. And well, then you send the police there because they send a small number of officers to a very limited area. And guess what? You know, the bad guys show up and indeed these, this one district was able to reduce their crime rate more than any other district in Los Angeles over that time period by using this thing. Isn't that interesting? But I want to seize on the last few minutes here together with you, Ethan. I'd like to seize on something you said about traffic. I'd like to create a case study. OK. And let's make Honolulu. The city and county of Honolulu are a case study. I want to give you some data points, OK? So we have weather, we have rain, you know, and everybody knows that in Honolulu when it rains, the traffic slows down. People get, you know, unexplainably timid or something when the traffic slows down. We have road conditions, which, you know, you can determine that from a drone or a satellite. You can see where we know it's all lousy, right? You have the existing infrastructure, you have the traffic lights that work, you know, the way they work. You have certain things happening in the community. You have certain events, certain institutional openings and closings. You have the schools, the businesses, the government, all these factors. They could all be digitized. They could all be made into data points, not only direct numeric data points, but those kind of subjective to objective transitions that we learned about in the last MIT newsletter, you know, summarizing the long document and all that. So you can have data about pretty much how our society here in this city works, OK, and how our traffic works and how we work. On a full moon, I feel differently on a full moon, that kind of thing. And I put it all in. It's not 10 data points. It's way more than that. And then I put it into, you know, some kind of some kind of processing and I have this huge algorithm that tells me not only where I can go to avoid the traffic, but it tells the city what they can do or it tells the traffic lights automatically what they should do in order to expedite the traffic. It tells the planners what they have to do to avoid the congestion. It tells the buses where they should go and not go. It tells everybody individually what they should do in order to minimize the congestion. And yet sometimes, you know, you don't even need that sophisticated an algorithm that is, and I think I was mentioning this before, that they recently put 20 drivers into a course and had them all driving around. And as human drivers will want to do, they begin sort of clumping up and getting these little traffic bubbles and traffic jams and all. And they dropped into the mix one car that was a self driving car and that immediately reduced the problem. Having simply one car that behaved truly rationally and well, basically suddenly diffused, broke apart the clots, made the traffic flow much more smoothly, made everyone run better. And this car was on its own. Right. It wasn't getting data points from anywhere. Right. Making its own. Other than its basic neighborhood. It's just its own sensors looking around. And there's nobody doing the elaborate sort of watching how everyone was going and, you know. So I'll give you, if that were the case, so then multiply that one car times a million. Now all the cars are that smart. And on top of that, you know, take HPD, take the traffic management systems of the city and make them smart too. We don't have any congestion at all. Don't you think? Yeah. No, I foresee that we could and should have transportation systems that would move people around quickly, smoothly, efficiently with a relatively low carbon footprint. And yes, a lot of us to get places much more neatly. Yeah. I mean, everything from the new Bikki, you know, bike share system is going to be put in place. Included that. Light rail, you know, whatever you want. There's lots of different options that are getting part of things. It's SimCity on steroids. There we go. It's everything you ever wanted your city to do. And it's all automated and it talks to all of us. We all engage with it. But, you know, one thing that comes to mind from this discussion is that it's not just the traffic. It's everything, right? It's every system that we need, every system that affects our lives together. And as this data mining is sort of going to finer and finer grains pretty soon, the inevitable result is that somewhere in that data, they can say, where was Jay Fidel at 4.15 p.m. last Tuesday, and they'll be able to tell you exactly where you were, where you were driving, what you were doing when you made that left turn illegally, you know, whatever it was. And some people don't like that. No, I mean, you remember the vancams. I like the vancams. That's what the vancams are very good. But the statute that adopted the vancams and gave everybody a ticket based on a photograph of them speeding, I guess it was, went away, it was repealed in the same session. It was amazing. And so there was a certain resistance to this. And so I give you this case study where, you know, we're living in a kind of nirvana. Everything works well. Our biggest bugaboo traffic congestion is gone. And so many other things in the city are rationalized. Our community is efficient in every way you can imagine. People would resist that, wouldn't they? Oh, yeah, right. No, people, you sort of look at a place like Singapore where, you know, chewing gum can get you in big trouble. I mean, once you begin to get that level of data analysis, they'll be able to tell them you're chewing gum. And if the powers that be decide that we should not chew gum, well, you're going to be in big trouble. Yeah, and that's the tension. Right. The technology that avoids the congestion and gives you all kinds of other benefits is all tech. Also technology that knows where you are and what you're doing all the time, arguably an invasion of your privacy to the nth degree. I mean, in New York, when they put GPS in the taxi cabs so that the owners of the taxi cab could see where the driver was driving or sleeping or having lunch, the drivers sued. They all got sued the owner group. OK, and guess what? They lost because inevitably, don't you think this is so tantalizing? The possibility of using AI to manage our lives together, our city together is so attractive that privacy, I'm sorry to say, will have to fall second. Yeah, I mean, the fight between the taxi cabs and groups like Uber and Lyft these days, to me, there's a group who's desperately clinging to the past and this old model. And I'm sorry, but, you know, the buggy whip industry is gone. You know, countless people were very upset to the buggy industry disappear, you know, the guys who made the buggy whips, you know, I'm not a lot now. Nobody wants my product anymore now, but it happens. You know, they say there's a message in every show, Ethan. And this course is likable science, so we have an orientation towards science and progressive and progress and whatnot. And Ethan and I talk about these things. So what we learned today is that buggy whips are dead. You heard it here on Think Tech. Breaking news. But it's been a great time with you, Ethan. We'll do it again soon. I so enjoy these discussions. I look forward to it. Aloha. Aloha.