Feb 12, 2020

This week's interview is with Doctor Nic Woods, Microsoft Australia's Chief Medical Officer. As we learn more about the use of artificial intelligence in healthcare, we see more and more similarities to the use of AI in education. And we have a fascinating discussion about the role of the professional alongside the AI systems - and what it means when we talk about a profession as a craft.

And in the "I didn't know that" moment, did you know that when you receive a doctorate in Sweden, you get given a sword. And yet medical doctors don't even get given their first stethoscope!

________________________________________

TRANSCRIPT For this episode of The AI in Education Podcast
Series: 2
Episode: 6

This transcript was auto-generated. If you spot any important errors, do feel free to email the podcast hosts for corrections.

Welcome to the AI for Education podcast. I'm Ray.
And I'm Dan.
And we're together, Dan.
Yay.
So, last week we went out and find somebody smarter than us. We did that the week before. The challenge this week was to do that again.
And did you?
You be the judge of that, I went off and had a chat with Dr. Nick Woods. Doctor as in the medical sense of doctor. So, he's our healthcare lead for Microsoft Australia.
Fantastic.
And I got a chance to talk to him about how AI is being used in healthcare. Well, thanks for joining us. Maybe it would be great if you were to tell everybody who you are, what you are, and what you do. Well, thanks very much, Frey. Great to be a part of your podcast series. So, I'm uh Nick Woods, the health industry executive and chief medical officer for Microsoft Health Team in Australia. My role is really to support a lot of our health customers and that's being hospital, state government, um and thinking about how they can leverage technology to improve the outcomes of health care, which could be around things like safe improving safety or even improving efficiencies. I've got a clinical background in emergency medicine and technology for a number of years and started off life actually doing engineering. So come full circle back into this intersection of health and technology.
So you are the kind of doctor that when they say on a plane is there a doctor on board you're that kind of doctor not an academic doctor.
I am that kind of doctor and fortunately I just had a recent example of that only three days ago.
Oh really? Wow. Gosh. Well thank you Nick for be for being the person you are. I I discovered recently that academic doctors in Sweden get given a sword when they graduate. So when you become an academic doctor in Sweden, you get presented with a sword as well as your gown. Do they is it true they give you a stethoscope as part of becoming a doctor in the healthare system?
Uh no, there was no such gifts given to me by my university. You were all on your own and you bought your own equipment from everything to stethoscopes from stethoscopes to uh odoscopes to even toes to take off blood.
Wow. Already we've started to drop into specific words in healthcare that I don't know because I've only spent my time in the in the education field. So I'm not sure what those other items are that you talked about, but I think that's no difference between your market and ours. So we deal with the education system. We've got a lot of really specific language. I guess the healthcare sector is exactly the same.
Absolutely. And and I think it's additive as well. There's the health terminology and ontologies. There's the technology terminologies and ontologies that are specific to health. So I think it's a big field. It's but it's a very exciting field of you know thinking about that intersection of of health and technology all of the time. I'm kind of starting to see now the stories about AI starting to be used in healthcare. There have been some stories over the last couple of months about different projects going on. So have the organizations that you work with have they started using AI yet.
So in Australia, AI in healthcare is starting to gain momentum. I think Australia is very well placed in many ways and we have availability of significantly large data sets because we have public health delivered services you know across hospitals for example across New South Wales. So there really is opportunity to think about how can we leverage some of that data which is necessary to train and and retrain some of these models to improve the uh outcomes of care.
And and one of the starting points for AI is having lots of data. I'm guessing the healthcare system has got lots of data,
huge amounts of data, zetes, and um it's thought that that data is increasing rapidly. So we're not seeing a decrease for example when things like radiology studies in fact radiology studies are are increasing every year and then there are new types of studies. What we talk about is data del huge and how do we actually make sense of that data. There are some areas which are still quite paper based in hospitals. So if we look in Australia the private health care system for example tend not to have made the investment in electronic medical records and digital hospitals as the public system has but over time inevitably some of that will change.
So as they start to realize the value of the data they've got one of the things we're seeing in education is the focus on BI business intelligence and reporting has started to change to oh we can use AI to help us make decisions rather than just giving us reports and information. Is that the same in healthcare?
Yeah, we have seen that with some partners and customers here in Australia. So I'll give you an example of the work we're doing with Victoria Ambulance where they have a number of different data sources that are in different systems and they really wanted someone to come and help them bring that data together but also start to help make more intelligent decision. decisions about most optimal routes to send an ambulance to a particular call to can we start getting a little bit more predictive of where we might need a number of ambulances for upcoming events. So it is moving into that world of of predictive and we certainly have seen a number of hospital-based health customers as well think about how they use their data for predictive models of things like length of stay and deterioration and and sepsis and some of the those things where we can use that information to earlier intervene and and hopefully make a more positive outcome on on the patient care.
So the ambulance example, I guess they're using a lot of data from different places because I'm guessing are they thinking about weather data? Are they using just healthcare data or that is it a a journey across exploring other data as well?
It's it's using multiple data sources. So they have a number of man and dispatch systems where they take calls, triage calls and and route to various ambulance teams that are allocated around the state. Then there are things like you know pulling in event information as well as some other information sources from other government agencies. You know it could be things like where there could be a um protest going on in the city for example. So it's really starting to think about how they need to respond to these types of uh events as well as as you mentioned things like Bureau of Meteorology data was the Victoria uh had the thunderstorm asthma outbreak a few years ago. Unfortunately, eight nine people nine people died. Uh so we're already starting to think about how how do we get a little bit further ahead of being able to forecast out that type of event and um then send more notifications out to people particularly people that are likely to be impacted by that.
So I guess in in that case it's about using AI to help you predict what resources you need and where you need those resources and then you know are there things you can do to minimize the impact of activities you can see coming up in the future and that that sounds like it's like population or state level stuff.
Yeah.
What about then the other example you talked about around clinical? So where where are people using it in the clinical? I guess the education parallel to that is in the classroom rather than in the management and organization of education.
Sure. So I think you mentioned before about you know predictive models, predictive analytics and that is true of hospital venues. So we've had quite a few hospital services that have tried building out models for predicting things like length of stay and also deterioration in sepsis. There's some good work going on there in New South Wales health. I think where we're at now, we're not quite into how do we take that into a clinical workflow to impact the care and then once you do that, what is the what is the result of having done that? So I think we're probably a little bit early in the journey in many ways. I think it's less about building the models. I almost think building a model if you've got data is almost commodity now. uh it's thinking about how do you bring it back into the workflow. How do you make that sustainable and you know whatever you do is going to require ongoing perhaps integration from electronic medical record data decisions about how you get that information back into clinicians. So it's natural within their workflow rather than having a look at a a different system to to get that.
So it's an exciting space to be. There's also some things going on in the image arena as well. So there's a company uh called Harrison AI that is actually based in in Sydney. They've done a lot of work in developing a an image based model to look at the likely success of a of an embryo in its early stages because when scientists have a look at embryo development for fertility treatment, they can tell by the way the cells are dividing and and the types of patterns and and pace at which they're changing, they can make predictions about which one's more viable. So, the Harrison AI team done a great job with forming a joint venture and making that a commercial entity and are doing similar just starting that journey with a lot of radiology data as well. So there's some entities in Australia that are are starting to do that. We've done some work having a look at children's ears. So an ENT registar or consultant or an outreach nurse looks at children particularly to straight Aboriginal children that have a high rate of middle ear disease. They often take a photograph of of that kid's ear and do a number of other tests. And so we're looking at ways can we classify that photograph a little bit earlier or get it assessed and triaged by an ENT consultant so they can make a proper decision around that. Again early days but it's great that people are thinking about how technology can streamline some of these care processes.
And it sounds like you have a similar situation to the one we have in education which is sometimes the algorithmic bit the bit that helps you predict something isn't the complicated bit it's the organizational response to it. I know that one of your colleagues that I was talking to was talking about preventing patient readmissions and being able to predict which patients might need to be readmitted so that you didn't release them in the first place which is the same for education predicting the graduation of students and the challenge was what do we do now we know this prediction how do we build it into the processes of the organization
yeah and I guess the way I think about AI and machine learning however we want to interpret AI it is just a tool There are many other tools that we have in technology and I think there has been a lot of focus on on AI and I think it's important to think that it's a really exciting tool but it is just one of the tools that we have in in healthcare and the way I tend to frame a lot of these conversations is think about what what sort of outcome are we after you know what sort of things can we change through technology and then what are the tools that we need to to do that because a lot of these things in healthcare really call on you know how do how do we get that data out but how do we get that information back back into those workflows. And I think that's been one of the challenges with some of the take up is is thinking about that integration between the systems that clinicians use. There's also a big opportunity and conversation that I often have with health hospital health systems etc is thinking about what's going on operationally and even financially. It's not a conversation I get into every day but certainly operationally you know we know that there's a lot of inefficiencies in hospitals. So it could be something as fairly bland is you know how do you get hospital porters to grab patients from the ward down for their radiology exams which can often be a fairly laborious paperbased system. Can we start thinking about using more smart technology to make that a little bit better or predicting intake into emergency departments during different times of the year etc. making sure that we're forecasting staff load etc. So I think some of those operational scenarios are just as important and sometimes easier from a regulatory perspective to to get start and get people working on than more clinical based ones which have have its own set of nuances and thoughts around are they still safe over time etc.
It's interesting you raised the regulation thing because presumably that's about making sure that you're doing no harm as well as as you're trying to drive an improvement in things.
Yeah and absolutely and I think most people are looking at machine learning and particularly where it starts predicting what the outcome of for example a um radiology exam might be or you know what sort of disease might underpin that people really starting to think about is this any different from looking at a medication. So when we start moving into that world of making recommendations based on whatever the output is of a machine learning model then we start opening ourselves into the world of of regulation and that is a fairly live discussion going on. There's a lot of work internationally thinking about regulation of software as a medical device and and what that might mean and where we where government entities like we have the therapeutic goods administration here in Australia where they should regulate what they should regulate on etc.
Yeah, it's interesting because I don't think there's been that discussion in education. We do have discussions around ethics but maybe not around the regulation and certainly I don't think I've seen a role for a regulator to start getting interested in that in the same way that they get interested in the physical delivery for example of education. So that ethics discussion presumably hits into healthcare but it seems to sit alongside the discussion around safety as well.
Yeah. The the regulatory focus is really based around how as a government agency do you show and demonstrate and assure safety around different medical procedures, devices, medications. So you know we have a number of mechanisms in Australia that regulate medications that regulate medical devices and And so it's in some ways it's it's no different to that. And so over time we're starting to see that there will be the need for things like clinical trials to look at the outcome to prove that for example if you go and have a an application or go into a dermatologist and have a bunch of you know sophisticated cameras that have a look at all your skin lesions and can make a prediction about you know chance of whatever types of cancer or melanoma on your skin. You know we need to know firstly that that is safe that it probably performs at the same level or if not better than either a consultant or a GP. And then we need to think about well how do we make sure that that is performing at that sort of expected threshold over time as well and that again is a bit of another nuance and if you think about one of the other challenges for uptake in this area because there has been a huge amount of work already done particularly for things like radiology computer vision output and looking at things like CT scan and and models making a a risk prediction of a bleed in the head for example in a CT scan but we need to make sure that those models when they're built they don't just apply to a particular population that might have a particular you know ethnic background they might be skewed to other coorbidities which might alter some of the outputs and as Microsoft we've even found some lessons learned so our health research team has even taken a lot of publicly available X-ray of the chest data and and built out predictive models for looking at early types pneumonia with pretty good accuracy. But when they took that same model and applied it to a population in India, the first set was based on a lot of data out of the US. Then that model drifted and the accuracy of the model was nowhere near as good as it was in the US. So it's thinking about how do we make sure that when we apply it to different populations, it performs in in the same way.
That's really fascinating because I know that it's well established that facial recognition isn't as good in certain populations, but you would think that, you know, something like a medical image would be the same around the world. And so it's making me think about it in education because somebody's built a model that works in the US education system, for example, doesn't mean it's going to work in the UK system or the Australian system or in China or Thailand or wherever. That's that's a really fascinating insight because a lot of the technology products we get in education tend to come over from the States. I'm guessing there might be a similar model in healthare
absolutely and I think at you know the population level particularly when there's image data involved you can easily understand even even things like looking at the back of an eye got different pigmentation in the back of an eye and there's a a bunch of companies now that are looking starting to look at the back of an eye or things like diabetic retinopathy so there is probably likely to be some variation as we apply those same models to to different populations so it's one of those areas that even once we've built a model and we apply it to a new population making sure that it performs to the to the same degree or where the variance might be or we might have to retrain to an extent and then over time just making sure that as their population changes it might be impacted that it performs to that degree. So thinking about that ongoing governance it's almost a little bit the analogy I use is when a pharma company lists a new medication here in Australia there is this process of postmarket surveillance so some of those sort of thoughts are going on around these models of how do we assure for the population particularly if it's a consumer basing tool that they're performing at the same accuracy that they they should have been and that's an interesting diversion from the conversation I was going to ask you about which was the challenge we sometimes see where technology makes a lot of things possible and there's a discussion certainly when it comes to the application of artificial intelligence about well we know that that's possible you can do it but should you do it I guess you've got the frameworks already in place in healthare industry to actually have that conversation.
Yeah, I think we've sort of had touched on this conversation previously around ethics of AI and and health care and it is is an important topic but I think people are probably fundamentally thinking about you know safety and and quality in the first instance coming back to to some of the ethical issues just how do we en ensure that models are performing across multiple populations and that's where I see tremendous value of AI and healthcare we about thinking about the opportunity that I mentioned with children with ears middle ear disease where there aren't a huge number of ENT consultants out in the Northern Territories for example you know that creates a great opportunity to perhaps provide a level of expedited clinical response faster than you know we may be able to offer today. But there is also another side which is starting to come into people's mindsets in in the industry and uh there's some work came out of Israel last year of actually being able to very subtly modify radiology image data to shift the response to an AI model. So you can think that you know there is even people are starting to understand that although it's probably not a big safety issue how do we protect around people potentially adversarially attacking images or modifying images so particular machine learning model may not work as well.
Wow that's amazing and a little bit scary to think about where things are going to go in the future. So, a lot of the conversation around AI has been about being able to use data to build artificial intelligence. You know, whether that's finding images, the bank of X-ray images or the data sources that are the national health databases around the world from different countries and being able to build big data sets. What is the most interesting or surprising piece of data that you've found being useful in the healthcare industry and an AI project.
Hard to say. I think for me this is all about impacts. And so to to give you an example of the NHS, which is a fairly unified health system in the UK and NHS England, each year they perform two million mammogram as a screening test to look at chances of breast cancer developing in in women. Each of those two million mammograms gets reviewed by two radiologists. So that's four million radiologists. So you start seeing the the size of an opportunity to perhaps look at how could some of that processing be improved where perhaps we don't need a radiologist to look at the ones that we are 99 whatever% confident are normal but we just focus our attention on the ones that need that clinical input to say this is just perhaps not quite normal and really uh start informing that decision. So to me when you start looking at its scale it's not just the scale of data it's the scale of impact.
So then one of the questions that comes out of that the mamograms is a really interesting thing because I know I've seen research that says that a computer can now spot uh can now read a mamogram more effectively than the human or more accurately maybe which is similar in education you know we know that AI/root can mark exam papers more consistently and more accurately than humans but you then get that debate between the human professional saying oh yes I know a computer thinks it can do this better, but I'm I'm a person that does all of these things. Does the same debate exist in healthcare where you're there saying we can solve these problems with AI and you've got human professionals saying, "Yeah, I know you think you can solve it." That debate between human versus machine.
It's probably an ongoing discussion. I think we've seen us sort of coming off a little bit of a hype curve and what AI might do. Uh I'm still very bullish about what it can do in, you know, the near midterm. and even long-term future probably more focusing on the mid to long from that example of radiology. I think that the college of radiologists here in Australia and New Zealand for example is actually looking at how do they build quality into this. We need more radiologists. There's more modalities etc. But I think what's excit so you know none of this is about getting rid of clinicians or anything because there's so much need out there in the first place and it's increasing all the time to me and most clinicians as well. This is about improving quality and improving efficiencies. So I think there is a bit of a shift now where people are starting to perhaps feel well this is not a threat. This is just perhaps going to change what I can do in a change for the good. And I think the people that see the opportunity are going to be the ones that naturally jump jump on it a little bit earlier. That's interesting. I think again there's parallels in education because if I think about for example automated marking humans marking things doesn't deliver value back to a student. You know simply marking a paper doesn't deliver value. Providing good feedback delivers value. So if you can release time by being efficient in one way in order to give a teacher more time to provide feedback, that's going to improve the outcome, not just get the task done. So it sounds as though there's similar discussions.
Yeah. And and I describe medicine as a c or health as a craft. It's it's not necessarily an art. It's and you know, there's still the element of human care. And to me, all of this is giving time back to clinicians to to become more human. That's that's the crux of it. That's the genus of it. And you think about if you got a complex patient, it's so complicated. You know, there is no going to be no master algorithm that could say if you've got an if you've got an elderly family member, there's just coming out of hospital with three or four conditions. You know, where do you send them? What sort of postf follow-up care do you have? What sort of family support should be there? No algorithm. Why would we even attempt that? But hopefully We can use the data to make better informed decisions more efficiently, but still provide that human care which is what we're good at and understanding what does that person really want? What can their family do?
Yeah. Same parallel with education. You know, I often say the people that believe that you can replace teachers with robots are so wrong because it misunderstands the role of the teacher.
Yep. Abs. Absolutely. And I see it exactly the same way.
Wow. I had expected there to be so many parallels between the two different markets. But I guess it shouldn't surprise me because ultimately both of the professions, both teaching and healthcare, is about a professional looking after a person and doing their best for them.
Absolutely. And it's interesting as you're describing it because I I absolutely get the the parallels between the the two areas.
Hey, thanks Nick. Thanks for your time. Maybe we'll catch up again in six months or so and talk about some of the new use cases and see where the parallels are to educ there.
Absolutely. We would love to do that. There's a lot going on. Uh I I often present and one of the topics I present on is AI and healthcare and I'm finding I'm having to sort of update my presentations on a on about a three monthly basis. There's so much going on. So look forward to that.
Great. Well, I hope to see you soon and I hope to be able to watch one of those presentations and then we can dive down into a deeper discussion. Thank you, Nick.
Perfect. Thanks, Ray.
Wow, that was fantastic, Ray. It was really fascinating. points that Nick brought up there. Something jumped out to me straight away was the parallel to outcomes.
Absolutely. The similar things that we've got in education. You know, it's all about people outcomes and it was all about patient outcomes.
Yeah. And also we got into a lot of conversation about efficiencies
which is interesting because I think different countries would see the word efficiency in a different way. But if I think about what he was talking about in there is well how can we use the data? How can we use AI in order to improve a service and and for that service to be more efficient. So the example was the ambulance one where they're using lots of data from do lots of different places in order to be able to work out the optimal routing for ambulances to work out where they should be. And that data included not just the stuff in the ambulance systems about well here's the expected load but also the the weather,
what is the traffic data
and what events are going on and putting all that in. And that to me was a parallel to education that was really strong because sure you can make some predictions with a small amount of data but so many other things that you would traditionally find difficult to pin down that would help you to understand a student better and to predict an outcome for a student.
And when you've got a subject matter expert like Nick who was a doctor and he has got a background in that clinical physician element, you know, you can really see that they're obviously applying a lot of this AI to the places that it actually needs to be. So obviously the outcomes for them are very much patient care and the point of patient care.
Yeah. And that was a good parallel to education as well when he was talking about How do you give time back to clinicians? How do you give people on the front line more time to treat patients and better insights for that? That has a parallel to education. We've often had that debate, haven't we, about technology of
are we saving time for teachers? Are there more things that we can do to save time for teachers? Can we help them to be sadly I'm going to use the efficiency word, but can we take away some of the things that they shouldn't have to do in order that they can put more and more of their time fixated on their students?
And I think what's fascinating for me as well taking your point on board there was that I think when we look at health I got this perception of AI looking at X-rays and analyzing the genome and all this kind of sexy stuff however the actual efficiencies they really need are below the surface and some of those headline grabbing stories that we see which are very important obviously around the breakthroughs that AI is making but obviously there's a lot of stuff underneath there in the processes which you get the same in education there's a a lot of processes that could be changed in edu and there's a lot of processes in health because we all so simil similar to education we all go to hospital one time or another or take our kids there so we all have seen the inefficiencies firsthand and I think that was something that really jumped out at me
yeah in a way it was a little bit boring wasn't it we were kind of focusing some of the processes and things like that but the headlines are very different now I used to work as a columnist for the times education supplement and that was in the day when being a journalist meant that things appeared in print not online.
Bear in mind for our podcasters says, "You've got a pen in your hand, right?" And I've got a digital stylus.
You're just showing off now. But in that world, it was about communicating a story. Whereas journalism today is about getting you to click on a story. And that is why I think that the conversation we had with Nick is very different from the headlines you read. Because the headlines you read are all about solving this world problem. Click on this thing to discover, you know, the mysteries of medicine and how robots coming to destroy your jobs and stuff like that. Because that's what we click on to read. But the reality dayto-day is that we're applying things in a in a way that is about efficiencies and outcomes that are not sexy headlines. And I think the same is true for education. We see some of the headlines about AI and education and they're very controversial. They're sometimes incredibly innovative, but they might only be happening in a small area and it's about the clicks because it's the clicks that pay journalists to write stories these days.
Absolutely. The other the other romantic comment that he made was something around health is craft, not an art. I thought that was great. What are your thoughts on that?
I think that's another great parallel to education because teaching is a craft. You know, it's a blend of a profession and technical skills, you know, science and art and those two things together. That's what a craft is, isn't it? It's a blend of technical and and artistic skills and and teaching us about that. I mean, I don't know. You you're you're the ex-teer in the room, Dan.
Yeah. Well, I think I think you're absolutely right. You know, it's it's a creative process. It's not just you've got your a kit bag of tools and technologies that you can use, but also it's about your emotional intelligence. It's about your creativity when you're planning lessons. It is a craft and and I think that was great the way you brought that through. I think when he mentioned an art, I think I think it can be an art as well. I think teaching can be an art, not in terms of the subject, but I think people who do it well can make it kind of flow. And similarly to surgeons and physicians who kind of really can do some fantastic things, I suppose comparing craft and art together, it's more of a parallel to I thought he was going to say health is um a craft not a science and that was going to worry me but yeah you know teaching is a craft and and it is got very much parallels to what Nick was saying there around healthcare
and linking back to last week's podcast if you remember that where we were talking to Lee Hickin about the projects going on in Kacadoo and he was talking about the indigenous rangers bringing their knowledge into it. To me it's the same thing. It's not this isn't just a robots are coming we can replace people with robots because if something's a craft, if something's an art, if it's not just straight ical science. You don't throw away the human in the process. What you do is help the human. And that's probably a growing feeling in this whole conversation around AI and education is the worst end of the story is the robots are coming to replace everybody. It's about that using AI and technology in order to help people achieve more rather than replace them.
And I think the speed of that you did mention right at the end there about there's so much going on in healthcare. He needs to do a a PowerPoint presentation every 3 months now. And I suppose you you do a lot of these presentations and so do I. And I think that's definitely the case in the things that I've been doing recently where you know about 3 years ago you could you spend a lot of time on one presentation and edit it the audience and keep the core messaging whereas now the core messaging is even changing very very quickly. So you can't rely on well I'll use the presentation I did last week. Often you have to change it you know in a lot more detail.
Could you imagine writing a curriculum for AI and expecting that you're going to write something now that's going to be delivered in three months time without having to come back and re The speed of change in this area is immense.
Absolutely. The final thing I'd like to pick up on and I'm sure you noticed it was this talking about the ongoing governance between the US and different jurisdictions for health care and the fact that some of those models that are applicable in India and other places were not applicable in Australia similar to education.
Yeah, that was interesting when he talked about that ability to read a X-ray image and the model that they used in the US didn't work in India which is weird because you think well humans are humans. you know, inside just
you think we're sorry, me in my uneducated, we're all the same inside, but it's really interesting then and and it made me think about education because we're originally from the UK and in the UK about one in 10 children go to private school. In Australia, nearly half of all children go to private school. In fact, I think in CRA in high schools it's more than half go to private school. So, if you were building a model to predict learning outcomes in the UK and then you brought it over here, actually, it's not going to work because the profile of students in private schools versus government schools is completely different. And I think we're going to find that across all kinds of things. You don't have snow days in Australia. You do have bushfire days. So there are all kinds of differences between the countries and and that's why I guess let's go back to that craft thing. You know, the craft is the understanding of of some of those things as well. And and you know, the same with the the Kacadoo indigenous rangers thing is sure we can apply this worldwide model about land use, but when we get into my territory, In my particular place, it's different. And I think the same would apply in my particular school, in my particular university, in my particular course, in my particular faculty. Ah, it's different in engineering, to social sciences. It's different in French to history.
Totally. Yeah. So, I can see your passports out to gain, Ray. That was a fantastic interview. I think you should go off and try to find somebody else. What do you think?
Well, let's keep seeing if we can find people smarter than us that know more about AI.
It won't take long.
It's not going to take long. See you next week, Dan.
See you next Okay.