 Medicine is a perfect, fertile field for AI. There's so many tasks and so many things that it can do. I didn't realize I was using AI until I started thinking about it. I use Dragon, and I can use my cell phone now to dictate into Dragon, I mean it's an amazing thing the newest versions of the Dragon medical 99% pick it up right out of the box. That’s Kurzweil’s stuff right?

: As I recall there’s a different researcher that actually developed on that, though Kurzweil is obviously a brilliant man. The discussion around that technology, about voice recognition technology, also is a great way to talk about the difference with AI, what's changed. How is AI different than the last three four or five generations of digital technologies? Voice is one of those, so we've been working on voice recognition for as long as computers have been around but it's not really helpful when it's at 70% efficacy, 80% efficacy. Even 90% efficacy can be a little troublesome. We might think 90% accuracy in lot of things is good, in voice recognition not so much. But when it works it scales infinitely, it's suddenly available for everything and at 99% accuracy suddenly we just realize “wow we can use voice.”

That's why it's it appears as though those 24 by 7 microphones, what we call speakers in our home, like Alexa and Siri and so forth, those microphone speakers, that's why they occur as having suddenly appeared. It's because the technology suddenly reached a threshold, it's useful, and boom everybody can use it in seemingly infinite number of applications.

 Voice recognition really improves life, I know in radiology they get their work done much quicker. That's one application of AI in medicine. What are some of the other applications that that you can name off the top of your head?

I really put them in four buckets: drug discovery, I have patient records that's in the nomenclature all over this, I have logistics, you know just the moving of supplies around a hospital, and then you know we have medical robotics about which I've written.

And so in those four I'd say that the one that's exciting for me is drug discovery. We have a kind of increasing sensitivity we might say about human trials on drugs. Yet we have this treasure trove of past drug trials. It can't be tapped right now or it's very difficult to tap it because the data is dirty, it's clustered in a particular way where we don't often know what data was collected and then it was collected in different ways. So we might look at an old test where they ask the patient did you smoke yes no, another one might say how much did you smoke, and another might say how long ago did you smoke. They're all trying to get at the same thing, three different studies, over three different times maybe over three different sponsors and three different companies.

How do you merge them, if those were three different drugs, into something as a collective? You can't normalize them or average them, you want to keep that fidelity, so the innovation that's coming in data integration, this is very boring low-level IT technology, is going to allow for that to come together for analysis. This is hugely exciting.

 What other applications do you see for robots and for AI?

 You know logistics is an obvious one. So I was in Pittsburgh where we saw this r2d2 looking thing shuttling supplies around a hallway. This is a great way for anybody concerned about jobs in AI to think about the jobs that are going to get taken away. Jobs will get taken away. Here's one shuttling: supplies around a hospital. But you know, no job gets automated that humans want to do right? The way we've thought about this traditionally is dirty dangerous jobs those are the ones that get automated but even the ones that are just boring we find benefit automating these sort of jobs. Nobody trains themselves to shuttle supplies around a hospital, it's not a human task in some sense, it doesn't use our skills. So logistics, the supply chain of hospitals also. It's dreadfully inefficient. I was talking to a large Hospital in Manhattan who doesn't really even pay attention to the inefficiencies because they get measured on standard of care not the efficiency of their business, but I think the increase of some of these tools will help bring efficiency and lower the cost of delivery.

 There's a lot of ways, they don't even keep track. A nurse opens this sterile gown it gets wasted, throw it away, I'm not even sure it gets tracked. I think they just built into the cost so high to account for it. Which is not very efficient and we can't keep doing that, we have to track these things better, track supplies track who’s using what.

 I can say another example is patient records. So AI is going to make a difference in patient records. It has to some degree. I mean I look at the fundamental IT infrastructure of it, the guts of the technology, and the guts of the technology say that we need to integrate heterogeneous data, data that's not alike, and heterogeneous data doesn't like to be integrated.

This is the reason that healthcare.gov had such problems, it really was brought to its knees for a time, it's because heterogenous data doesn't like to be merged data. The British National Health Service is trying to do a similar sort of data integration of patient records, Italy has a 30 billion dollar effort trying to integrate health records, and I hate to say that they're likely to have unhappy outcomes because the technology is really just emerging for a new way of integrating heterogeneous data.

 They made the same mistakes we did by allowing 90 vendors to come in and do 90 different EMRs?

 They may be making different mistakes but similarly at the IT infrastructure level the same mistakes. I virtually guarantee will be made.

 That was a tragic mistake actually. In my office now I can get an EKG or an x-ray report sent to me through the EMR it's perfect it comes right in, but I have to print it and fax it to somebody else. I don't have the ability to send it directly from my computer to your computer. And faxing is the least secure of all. It's so inefficient, we're still printing things out and faxing them, I mean it shouldn't happen.

 Well this touches on other parts of our discussion where we really need to look at incentives. You know what are we incentivized to do and if we're incentivized just to continue meeting a practice of standard of care then faxes versus email really have nothing to do with it and we'll just keep doing what we're doing.

You know another way to think about it is how we will introduce new information thinking that's going to solve some sort of problem. It often doesn't it. And we experience this in our day to day lives, when we will in the United States go to find a mortgage application and the resulting mortgage paperwork has become thicker and thicker, no one's incentivized to make that smaller but that deluge of information on a home mortgage is ultimately giving you all the information you need but it doesn't help in your understanding.

 No it certainly doesn't, and when a person goes home from the hospital with a stack of papers that thick and there's instructions it's overwhelming actually to them. Every prescription has a four-page blurb with it and no one's reading them anyway.

 This is actually the perfect place, that physicians, clinicians, researchers, administrators, everybody in the healthcare industry already experiences many of the issues that that we are confronting in AI because you already in the healthcare profession you already are interacting with patients on issues of explain-ability and understanding even when somebody is sitting right across from you they're alert they’re nodding their head they still may not understand what you're talking about and you have a very difficult time, but you're completely used to the issue of your context being different than theirs and the level of understanding being different and this is very similar to what we confront with learning algorithms or any number of other parts of the totality of the AI system.

The Current Applications of and Opportunities for AI in Healthcare

 Part of that is regulatory too, I think the FDA doesn’t allow drug companies to re-use old studies, you have to redo the wheel every time.

: We, as a society, need to have a conversation about what we want the FDA to be. It’s not for me to say, it’s not for any one person to say. But if what we determine as a society is that any failures are intolerable, than that falls out a whole bunch of other regulatory processes. But on the other hand if we can accept a certain amount of error in a certain context, then that allows for a different level of experimentation.

One of the difficulties about all of these issues is the degree of innovation that takes place in general in healthcare. Fresh out of medical school you’re taught the latest procedures. You might think of yourself as the most innovative, but you’re not the most experienced. Maybe ten years in, you’re experienced, you’re appropriately seasoned, but you’re no longer the most innovative. It’s a tension between those two.

The regulatory regime we have set up isn’t rewarding innovation. It rewards practicing to the standard of care. So we have, reasonably, come to these conclusions, and set up everything including our health insurance to accommodate these societal conclusions. But we don’t have the larger conversations that society needs to have about the totality of the system, what that means for the outcome.

 What do you think about the FDA’s recommendations about robotic surgery, particularly with invasive, as it relates to cancer surgery?

 I think this goes to the point about what the FDA should be doing, about how much they should be regulating. I think in these issues I am deeply suspicious of the possibility for misinterpretation on what we even mean by surgical robotics, and how that may be misapplied.

I think when we go down these paths people often think of robots as being fully automated. They don’t think of robots as being augmentations, little assistance robots.

 Yeah, co-bots, collaborative robots. When I had the good fortune of working in the Obama administration we were looking for the future of robotic research for the federal government. I was privileged to be among the best minds at the table, we came up with the idea that these collaborative robots would be ubiquitous. That’s the best thinking we have. Rather recently. These co-robots would be all around us. And they’re going to be augmenting us. So I was working with one surgical robot that would allow my shaky hand to be completely smooth.

That’s a robot. And it’s augmenting me. And I not being an expert as a regulator, would say that’s pretty easy to argue that that’s going to help my efficacy.

 Yet it didn’t. Many of the studies showed that the robotic surgeries were no more effective than conventional surgeries, yet they were more expensive. Which surprised me. Am I correct in saying that?

 I think in the totality of these technologies and the totality of this environment, what we might see is perhaps the outcomes are no different but the surgeon is less fatigued, maybe the outcomes are no different but there’s a lessened likelihood of catastrophic outcomes for example. Then we might say, they’re worse because the surgeons ignored them, didn’t follow the directions. So I don’t know. I would say the jury’s out on that and the robot I saw a year ago could be very different than it is today.

Balancing FDA Regulation With Technology Innovation

 In terms of applications in medicine, what do you think is holding us back, what are some of the problems that you see that are keeping us from moving forward? You mentioned earlier a lack of collaboration sometimes between the providers and the software engineers, but what other things are holding us back?

: We talked a little bit about the things that are holding us back from the adoption of modern technologies, that's how I had phrased this, modern technologies, in the healthcare system. You know it's worth repeating that the issue of data integration has historically been a very big problem to do proper analysis. I was on participating with this team with a non-governmental organization they gave a lot of money to address this issue of child mortality. I had not been aware that in the United States there was this community that had child mortality rates that would rival a developing country. It's really tragic and then so this NGO was spending millions of dollars to try to address this problem, they had some first-rate physicians from UPenn from UPMC and from Vanderbilt working to address this issue. And they start with data, they spend some amount of money trying to collect data to then do analysis, they spent about 18 months, 2 years, a couple of million dollars of the grant money, on this database integration and I asked the physicians why did you do that? And they said well that's just what the IT people said. I asked the IT people why did you do that and they say well we used the best technology available at the time. But they spent a couple of million dollars you know 18 to 24 months and they integrated three databases. The new world you're going to see is being able to integrate 30 databases in a fraction of the time at a fraction of the money. That literally will save lives because the outcomes from those sort of research projects will come to the public sooner that's it that's a huge innovation, but it's not terribly telegenic to talk about data integration.

 It’s an amazing innovation, how's it going to happen though? How are we going to go from where we are then in that hospital system to where you're talking about?

 That particular innovation will come from innovations in mathematics. So in the same way that we have had innovations in physics that allow for new semiconductor approaches to emerge in that industry, this is an innovation in mathematics that will augment maybe even fork off from calculus, and enable a new digital way of thinking about these sort of problems.

 Do you think we need more processing power to do that? Are our processors powerful enough to do that?

 It's a great question. So in the totality of the planning part of “sense plan act” you have the storage, computation, and networking. And in the computation part of that that has improved sufficient to allow that mathematical innovation to take root and be applied into computer science. That will continue to improve but that's not a limiting factor for this particular application, regardless of the size of the database.

 I agree that the databases that we have, that we have to do better, integrating them. That's really a real slow down in drug technology and patient care and patient records.

What do you think about privacy issues when it comes to medicine and electronic records and data gathering, with our watches that measure our heart rates and report to Apple, you know and stuff like that, do you have any concerns about that?

 I do, I have many concerns about that and I would say many of our reactions are misapplied so we currently have this restriction about applying Social Security numbers to patients the problem or phone numbers, now those laws should be repealed. Because engineers have to somehow connect a patient to data and if they can't do Social Security numbers they'll do a workaround and that workaround very well may be worse. This gets to the sort of systems intelligence that I would ask everybody to participate in which thinks about the data acquisition, the data collection ,and storage, and then the resulting actions. The acquisition of data is certainly a place around which we need to be very concerned and the metadata problem the description about the data remains a bigger problem than many people appreciate but if people can participate in the holistic systems intelligence systems view of AI, they can participate in the conversation to see whether this is appropriate.

Because I don't have an answer and no one really has an answer and our understanding of this problem is very different than today, 2019 2020, than it was in 2016 2015, and therefore it's probably going to be different in 2020 to 2023, so we need to be continuing experimenting about how our reactions feel relative to the technology. And the data available it's very concerning if the data that's collected is say stored for all time and somehow carelessly distributed. You know there was an example recently where some 50,000 MRI images were on a public hospital computer right? This is tragic, it's because often people have this myopic view of what their job is in relation to data security. “Oh that's ITs problem” you know this particular problem, IT thought it was the hospital, the hospital thought that was the IT department, they're kind of pointing back and forth to each other. Each has their own little point solution instead of people constantly reflecting on what's the total solution here, double-checking “hey do we have the security protocols in place?” at multiple steps.

You know we do these sort of guardrails in other parts of our lives, in transportation we talked about this a little bit you know, we've put up sidewalks to separate from cars, we've put up bollards on sidewalks increasingly to make sure we keep cars separate, and that allows for new experiments, allows for autonomous vehicles to go down the street with a little more assurance that they're not going to be crossing over to the sidewalk. Because these cars will fail but they fail in different ways that we don't really appreciate. And so guardrails guidelines, that's far more effective right now than regulation in helping shape the conversation, because regulations necessarily are overreach just by the nature of it. So us having a conversation, physicians are very well placed to be part of this conversation, can help put in the structure that allows for new tools, new innovations, could come and just be placed in and applied with patient.

The Implementation Hurdles Associated With Adoption of AI and Robotics

 Some people think that MRI scans can be used like genetics, to learn about people. I’m not so sure, because MRIs change, genes don’t change all that much. But that brings me to the issue of genomics, and the huge impact of AI on genomics. And how that’s going to affect medicine.

: I use that as yet another opportunity to invite people into the systems view. We today are making decisions today about how we use or restrict those innovations. It’s important that we’re part of that conversation. When we as a society started introducing cars into our civilization, and you certainly think the original inventors of the car thought about what that could do for transportation, they would have no idea that we would have strip malls, suburbs, and urban sprawl.

 That’s the canonical example. About what it was going to solve. But it introduces new problems. I think people need to be engaged in this conversation today to say “what if?” so we can put guidelines and guardrails around these technologies so we can introduce them in ways we consider to be safe. The scary part of some of these propositions is that we’re going to have companies that know more about us than we do. We have companies that know more about my physiology than I do, so they know I crave sugar and fat and they market to me Doritos. They’re almost addictive, you know, “I can’t resist Doritos.” We will have those sort of behavior triggers around which companies will have more information than we do around the data and the interpretation of the data. More than we should even be expected to know.

 We had no idea ten years ago about Google, about Facebook, about data mining. They gave all this cool stuff for free. And we thought this was really neat. But they were gathering data, perfectly legally. You remember free 411? When you first had cell phones you had to pay for 411. Then there was a service called free 411. You could make 411 calls for free. That was Google testing out their voice activation. They did it 2 or 3 years and stopped.

 These sort of issues permeate our world. I don’t care so much if I am manipulated into buying dark chocolate. And Google doesn’t care so much if their ad placement to me is flawed. But there are many more high consequence contexts for the application of AI. Drug databases are 1, airplanes are another. The context under which we think of AI is really important. I care about my privacy at Google, my safety less so. At Boeing, the opposite. I also care about the context. I care about my password to a news site not a lot. The password to my bank, quite a bit.

And then we have the timing of our understanding. One reason regulation doesn’t work so well is our understanding of internet data in 2016, very different than today. And we can expect it will be different a few years from now. So regulation isn’t really the mechanism by which we have a dynamic conversation around the guardrails for these technologies.

 The way I see it, AI is going to be a bit like the arms race. There’s going to be a few countries that will race to develop AI, not just medical applications, there are military applications, transportation, there’s engineering and manufacturing. It’s going to come down to China, the United States, maybe Russia, using this technology all along. It’s going to be this nuclear arms race. Do you think so?

 This idea was popularized by Kai-Fu Lee, we shared a doctoral advisor and I think he’s fantastic, his writing is fantastic, as an executive and an investor he’s demonstrably effective. On that issue I can agree in a narrow domain. If you’re particularly talking about learning algorithms, that’s fair enough. Data matters, datas really the whole thing. As we were talking about earlier with Google, their experiments acquiring data and all the many ways we can acquire data without us knowing is a big deal.

I think there are more stumbling blocks for a society than we may appreciate to embrace AI. We don’t want resistance to AI. I’ll give you an example. The end user agreements that we see pop up on our devices, multiple times a day, I doubt you’ve read one. I’ve never, I’ve glanced at them and I’ve never read one end to end. That’s tough for non-lawyers or people who get bored. Here’s the scary proposition: those agreements can be violated without you knowing. They’re particularly pernicious with regard to children. It’s actually illegal to be advertising to children on their mobile devices. How would we know if they’re violating it? You have no idea for your kid’s safety or for you. This is a societal conversation. If a society gets sufficiently burned, or sufficiently upset about that possibility of abusing the privacy of children, there may be a resistance to the adoption of AI. And then who wins in that circumstance is not a black and white scenario. We need to think of this as a total systems, we need a degree of social systems intelligence. Where did the data come from? How are we using it? What’s the output? This manifests itself in many many ways.

The Acquisition of Data and the Issue of Privacy

 The application of AI to facial recognition is kind of an interesting topic because I hope it doesn't set us back, that issue of people resisting the facial recognition. Yes there’s flaws in it and there's bias in it, but it doesn't mean we should throw the whole thing out. It means we have to get more data. We have to analyze that data and understand where it comes from. You know because there's certain cities in this country that I think banned it altogether, your city.

: San Francisco. They’ve banned in certain applications. Both of us may have traveled to Europe a couple of years ago and we saw facial recognition being used for immigration control. You know I was really recently at the Detroit Airport I think maybe even much as six, nine months ago, and Delta Airlines was using facial recognition to confirm the photograph, matched your image, your face matched to that on the passport. It’s really strange because you'd think a person could have done that with roughly the same speed yet Delta Airlines had this and they're rolling it out to more airports now. Facial recognition right before you board and Delta makes no issue about this being somehow experimental, it looks like it's now just part of the background, part of the infrastructure.

Facial recognition has a great place to be in some applications. I don't love it on my phone but it's there and it's working increasingly effectively and now is a little faster. I also think of access control being actually a pretty nice application. So one story that was told to me by this facial recognition company Alcatraz who friends of mine funded in Palo Alto. Alcatraz had a story about deploying this in hospitals where the nurses were exchanging key cards, and that being to get in the supply cabinet or a drug cabinet, even worse, you're exchanging cards maybe for really good reasons but it'd be so much better just to have facial recognition and not to have any potential abuse of the key cards.

 What do you think the future is for medicine? In the near future and far future what are the kind of nice things that it can do for us?

 The way I invite people to think about this the way I think we started off this conversation getting at which is these evolutionary approaches, which is really how I continue to think of AI as a evolutionist, I guess you know instead of AI as a revolution, this is how often people think of AI. It’s just incremental evolutions of technology, incremental changes in our jobs, incremental improvements in safety ,in in a lot of different domains.

The combination of these can be drastic you know if you have a reservoir behind a dam and it raises just an inch you may not even be able to notice that but the volume of water was immense. So these little changes in a lot of different technologies in our healthcare system could have drastic impacts, but I think little by little is how we're going to see AI begin to influence our lives. We talked a little bit about drug discovery, we're gonna improve data integration, and that's gonna have a really profound effect. We're gonna begin to address some of the issues of privacy, hopefully around health care records, and that will have a profound effect. We will have an increase in in robots that are useful in logistics, useful in the everyday movement of goods in a hospital, that will have profound effects. These literally will save lives going back to our example about the infant mortality if we're able to decrease that by even a little bit lives are saved and this is just information technology at work.

Facial Recognition and the Future of AI in Medicine

: You know we were probably going to get into this question of bias at some point, and this is a good place to touch on that, you know the issues of data and bias and data and collection and where this is going manifests itself in every part of this system. It's again why I ask people to have a sort of systems level intelligence, if we look at just a collection of data the sensing of data you may think that the radar on top of your autonomous car, it may be objective, you know certainly the thermometer or an air temperature, air quality, those are objective, but it was within the last hundred years, well within the lifetime of if not our parents our grandparents that we were collecting under a veneer of scientific objectivity the shapes of people's heads right?

And that was really for racist purposes but we had a veneer of science on this saying “oh no we're completely objective about the shape of people you know size of people's heads and we'll import this to determine who's in a more superior” and so that is an example of how even the data before it's set for analysis can be biased, even in the collection of the data, can be biased. Who's making the choice about which data we're keeping, which data we’re getting rid of. Did we did we pay attention to every possible group? You know you it was for a long time that the difference between male and females was ignored. You know I was told a story recently about how the there was a drug discovered that had a 95 percent efficacy but for a particular ethnic group it had severe consequences. But that didn't come to light until later.

You know this is bias in action and it permeates our world. The difficulty in that conversation then is how do you motivate drug companies to invent a new drug that's 95% accurate, effective for the vast majority of the population, how do you motivate them to then go invent a new drug for the small population that is very poorly served and for which there's no obvious second choice? These are the types of conversations we're going to need to have going forward.

We know when you have a result and getting at our earlier conversation that says well I can tell the patient one thing and they're very likely to get it, or I can tell a patient another thing and they may or may not, but it but it's more accurate, you know what do you tell them? That's the output of data. If I tell them I'm gonna have a drug that everybody expects or I'm gonna tell them it's an experimental drug, how do I represent this? That's an output of the data? How is that objective or subjective?

 What if banks used algorithms to determine who got loans and what if it was it was learned that women were better at repaying loans than men? Just suppose, now that's a bias and they put that bias in the loan thing ,should that bias be taken out because it's a bias? Would that mean women would have an easier time getting loans than men? Now it's a fact they say. Or supposing they learned that if you applied for a loan on Monday you were less likely to default than if you applied on Wednesday, and you happen to apply on Wednesday and the algorithm turns you down, is that fair?

 My wife can even give a better example, actually, we returned as a trade-in 2 phones to Apple last year, for a mistake of apples the phones were returned. Now my phone was returned because Apple had become hypersensitive about frauds coming from overseas and changing the serial number of phones and this one got flagged as that and it got sent back. My phone had been replaced on repair so it was a different serial number, but the system just rejected it and sent it back and I had no obvious recourse. There's no hey try again there's no going to the store, there's no phone number to call, it just it ends. My wife’s phone gets returned because it's connected to my phone, it's just on the same account, my wife is somehow implicated in my fraud of the serial number of the phone.

And what we wouldn't know and this is where this begins to be even scarier, is the degree to which that trigger may have gone somewhere else, maybe we trust Apple, we trust them with all of our data, and some very well-meaning bureaucrat says gosh I don't know how to assess these credit issues, I don't know how to assess these trustworthy issues, but before I interview people for jobs in the federal government I'm gonna use the credit scoring system that Apple has right? And I'm gonna see that he was trying to pull a fraud and so even though I didn't do anything, right I could I could be rejected for even application into the federal government and I could have missed the opportunity to have been in the White House because of these sorts of issues.

And this can make you mad because you might then say wow I know of these connections and now I was pulled aside by the TSA at the airport, was that because the TSA also used this credit scoring system? And then you might say well that was linked to my Amazon Prime account, and then I go into the CVS drugstore and the doors opened a little more slowly for me, it can really begin to drive you crazy these linkages of data. And this is again where I say a systems intelligence view, we need to think about how all of these systems relate, because if you just add more information and if you just resist this information neither of those extremes are gonna be really helpful. We need to be thinking of these nuances of data where is it appropriate, where is it not appropriate, what are the guiderails, what are the standards that we can put in a place and have a conversation as a society? It's not for me to answer it's not for you to answer, but the more people that are engaged in this conversation, and physicians clinicians, researchers, all healthcare professionals are in a fantastic place to be engaged in this conversation. The richer that conversation is the better decisions we’ll make.

Fighting Bias in Data and AI

 I went from a big practice to a small practice and one of the arguments that was made to me was, well you were seeing 30 patients a day and now you're seeing 10, what about those other 20? What's going to happen to them? Well what I found out was that I didn't need to see the other 20, they didn't need to come at all, I had 30 because I had a schedule that filled 30 people and I needed to fill it. What I found out was using my cell phone a person could send me a picture of a rash, I could just write back they didn't need to come or they didn't need to come to get a referral to see the eye doctor, I could handle it on the phone. And so I realized that with technology, even simple technology like email, a lot of money could be saved and telemed, that is another example where a lot of money could be is wasted on stuff like that, doctors seeing patients that don't need to be seen for things that don't need to be seen.

: Emergency rooms are classic for this they're clogged with people that don't need to be in an emergency setting, they just don't have any other medical options, so that sort of domain is a perfect application of telemedicine.

 But no doctor is in a hurry to discourage that kind of business though, ERs are not turning people away.

 And I think how we might then begin to adopt telemedicine is other underprivileged communities, you know serving remote places that don't get as much care, or serving distortions in that in the healthcare system. I was part of this this program that was evaluating the efficacy of a great new hospital that was put in in one of these developing countries, it was put there by a very well-meaning philanthropist, but the unintended consequence is it sucked the life out of all the other hospitals in the area, and there was no real way to undo that. You can't then say well I'm going to create these other Taj Mahal hospitals.

So it's very at the time a well-meaning, a great effort, to put a big hospital into which everybody could be attracted, but if it then kills off some people's local health care you might have actually done a disservice. Telemedicine could have addressed, telemedicine efforts to just bring doctor expertise out into the country and that's a way maybe to even begin to correct some of these unintended consequences.

The structures we put in place around these technologies will last a lifetime or more. Our roads define our cities, we have the same issue in AI. You know healthcare professionals I actually think represent what I think is the framework how even we'll think about our careers. You know a hundred years ago we had just a man on a horse back, maybe 150 years ago, a man on a horseback quite literally and old drug drugstores right where they were prescribing opium and we wouldn't even fathom that today. In the medical profession we don't have doctors in the same way, we don't have just the family doctor that knows everything about medicine, it would be inconceivable that one person would know everything about medicine. And it's become segmented towards multiple experts. You don't have a healthcare expert, you might have somebody that knows about the system, about insurance, about patient care, about medical records, you have people getting degrees in social medicine. You have surgeons without without a sense of ego (or maybe they do have a sense of ego) but without a sense of ego you know they can bring it a medical ethicist into the conversation and they know they're well served to have that part of the medical experience done by a professional who spends their time doing that so there's multiple different respected professional experts in their domain in the medical profession and I think that's where AI is gonna go. You're not gonna have one AI authority, you're gonna have multiple AI authorities in these multiple domains in this system of AI.

 I'll tell you though one of the negative things about that in hospitals, I had a guy in a hospital and I said to the man I think you should be out of bed sitting in a chair today, I go down the hall I got a call “you can't do that until the physical therapist clears him to get out of bed, okay and since you're on the phone the insulin you ordered, we can't give them that until the diabetes expert sees them as well, okay all right and since we have you on the phone, the Quality Assurance people say that he should be discharged tomorrow.” And what am I? Who am I, what am I doing here, you know what's my role? I just sign off on all this stuff? What does the physical therapist know to get this guy out of bed, I've known him for 40 years.

 All systems can calcify, it's really our job to make sure we continue to bring vibrancy to these systems, and we continue to adopt them to the current realities. You know no one wants to be just a cog in a wheel and no patient just wants to be on an assembly line. I think that's part of an awareness of the totality of the system and our elevating the importance of that conversation, that system-wide conversation, I think can help bring dynamism to these systems and a dynamism to these conversations that if not entirely eliminate begin to at least mitigate some of these issues.

Telemedicine and Technology Infrastructure

 Well, Dr. Daimler, we talked about a lot of different things. What do you see for the future of medical care and AI specifically?

: I see incremental improvements coming. I see fantastic things happening with the innovations in data integration, with the ability to bring more data to analysis. I see incremental improvements in medical robotics that are beginning to have real big effects. And the terrific things about these little things we wear on our bodies, I think that’s going to bring us into the conversation more about us taking ownership of our healthcare. I’m really encouraged about the possibility of professionals in the healthcare system participating in the discussion around AI, they have a lot to say and they have a rich body of experience because of the analogies I see between the healthcare system and an AI system, to contribute to a conversation about what are the appropriate guidelines for AI in our society?

 I think with your unique background I think you’re in a perfect position to help guide us along those lines. I learned a lot, it was very helpful.

Future Opportunities for AI Integration

 presentation titled “Health IT: Exploring the Role of Technology in Healthcare.”

I'm Dr. Simon Murray, an internist from Princeton, New Jersey, and medical director at MJH Life Sciences. I'm joined today by Dr. Eric Daimler, a computer scientist and leading authority in robotics and artificial intelligence with more than 20 years of experience as an entrepreneur, researcher, policy maker, author, and investor. He was named a Presidential Innovative Fellow by the Obama administration and drove the agenda for US leadership in the research, commercialization, and public adoption of AI and robotics. Eric is also co-founder of six tech companies including Conexus. Welcome Eric, let's begin.

 Well, I'm really pleased to have you here today I've read a lot about you and heard a lot about the things you've done and I am really interested to see what you have to say about artificial intelligence and its application to medicine particularly.

 Good to be here, thank you for having me.

 I heard one of your videos where you said we should stop teaching trigonometry and algebra and we should be teaching statistics and probability. I agree a hundred percent. So much in medicine of what we do is probability, and so much is distorted in the misunderstanding of statistics and probability. You cited the issues of trying to treat blood pressure and then cited all the car accidents that kill people, you know that that was an analogy I used to use all the time. People would say, “I'm not going to take Lipitor, you know it might kill me” and I would say “you drive your car, well your chances of getting killed in your car are about a hundred times greater than taking Lipitor” right? But people don't understand statistics and a lot of doctors don't understand statistics.

There is an old adage that says: what's 1 plus 1—if you ask a mathematician, “2.” What's 1 plus 1, if you ask a philosopher, “it depends.” If you ask a statistician, what's 1 plus 1, “what do you want it to be?”

So with that, you could tell me what your definition of AI is.

 Those are good anecdotes. I have a good friend who went through medical school contemporaneous in my doctoral program and he, as a joke, wrote a paper on his comrades in life sciences, in health care. The title of that paper was “What Physicians Know About Computer Science” and then the paper was blank. That was him doing it a physician and as a computer scientist. But it's a funny relationship we have to physicians. I have been really excited to be interacting with the healthcare community from a lot of different angles and I often start with the definition: you know we think today of AI being electricity or you know “data is the new oil” and much of how then we talk about AI could be analogized to us talking about how the pistons work inside of an internal combustion engine or in electricity talking about Edison and Tesla debating alternating or direct current. You know those might be interesting and we might have a good intellectual conversation about that, but it really does nothing for us in talking about the system that is transportation, the system that is energy distribution, and that's what I'm working to do with bringing people into the conversation around AI.

There is a definition of a robot that I adopt for AI and it's anything that can sense, plan, and act. And then repeat that experience. That's a learning machine: it could be AI, could be robotics, but that's what I think: sense, plan, and act.

And inside that system is a lot of different touchpoints, it's a collection of the data from sensors that could come from any number of different ways, this could be software or hardware, and then into the computation, the planning, where we could be talking about the computation, the storage, the networking -- but not just the minor point of the learning algorithms.

And then we have the output which also can be disregarded or put aside, but the output is important because when we get a result from a computer how do we interpret it and how do we know whether it's right or wrong, you know it and if it gives us something that's 10% away from our expectations we might say well fine, but what if it's 10 times away from our expectations, but it’s right? How do we know? That's an output judgment. And that whole “sense plan act” I think is a great way to start the conversation and frame our whole engagement around AI and modern digital technologies.

 The MacDonald definition was way more involved: every aspect of learning or other feature of intelligence can in principle be precisely described as a machine, that a machine can be made to simulate it, an attempt will be made to find out how to make machines use language, form abstractions and concepts of blah blah blah blah blah…I like your definition a lot better.

 It’s towards the motivation: are we trying to teach you to be an AI researcher? Then we want to get rather precise and we can say that machine learning is a subset of AI and there are non-machine-learning AIs and deep learning that's in the public discourse as a subset of machine learning, that's all accurate but it's pedantic for the conversation of the general population. And if we if we focus on that too much it takes people away, in they end up just reacting to what the “high priesthood” of AI researchers produce and then we will present it to you and you will like it. I want people to figure out their place in this in this whole system, have a type of system intelligence around it.

Your definition also brings forward AI as the ability for machines to do tasks rather than take jobs. There's a difference. Right what we're trying to do is ask them to perform duties that humans do to make our lives better, we're not asking them to take our jobs because they're not going to take our jobs.

There's a lot of different ways to think about that. These augmentation machines have historically just improved our experience little by little. Tt's easy for us to think of cars in this metaphor when we think about AI and everybody can kind of join in this discussion: the cars we have today are very different than they were a hundred years ago. Little by little week we added brake lights we added turn signals we added airbags, more recently we've added ABS we've added traction control, and then even bringing forward in the last couple of years we've had Distronic cruise control, you know smart cruise control. We've had cars shake, give haptic feedback, and then force cars back into a lane. That's how AI comes into being in most industries, just little by little tasks are taken off of us and automated.

I don't think anybody's going to say their life is worse off for ABS. I mean ABS and traction control are credited with saving tens of thousands of lives just in the United States alone. I think that's how a lot of these technologies are going to express themselves in automated vehicles and I think that's a just a fantastic framework to think about all the applications of AI software or robotics hardware when we're talking about these technologies in the healthcare system.

Health IT: Exploring the Role of Technology in Healthcare