Learn more about Beyond Limits, Asia Pacific at: http://www.beyond.ai
Find Leonard Lee on LinkedIn here: https://www.linkedin.com/in/leonard-lee-9646473/
JC: Well, hello everybody. And thank you for listening to the future of BizTech. I am your host, JC Granger, and I’m here with Leonard Lee, president of Asia Pacific with cognitive AI. This company is solving the world’s most complex problems, and I absolutely had to have Leonard on the show here today, so he could tell you all about it. I think you’re going to really like it. Leonard, thank you so much for coming on. Maybe introduce yourself a little bit to the audience and tell us about yourself and what you do over at cognitive AI and then we’ll kind of jump from there.
Leonard: Thank you, JC for having me on the show. So a little bit about myself. I’ve been in the tech industry for probably close to 20 years. My background is in engineering avionics engineering, but I’ve since moved into it and digitization. So apparently I run the Asia Pacific business as a president for Asia Pacific for Beyond Limits. And prior to this I’ve had you know, experience 12 years with Microsoft in sales transformation and global expansion, and also a digital ventures with Airbus and the last couple of years in the startup world. So with beyond limits, I think I’m really excited to be part of the company and the pioneering, a leadership team, right to drive our global expansion. So, you know, what we do is a very unique branch of AI, which we call cognitive AI.
Leonard: And we combine conventional models that most people know about like the deep neural network, the learning models, the machine learning models. And then we add on top are symbolic AI IP blocks. And the net result is we are able to create a AI model that thinks like a human being. So if you’re in a refinery, they will act like your best operator. That’s been on the job for 20 years and we’re able to codify that expert knowledge into the cognitive AI model. So that’s kind of a quick summary of what we do.
JC: So that’s fascinating to me because, my background is in psychology. I mean, I’ve been doing marketing for 20 years,, but my degree was in psychology and I always found cognitive psych, for example, to be something that I always took a lot of interest in. There was this one course I took called Sensation and Perception very specifically. And it was learning about how, you know, the brain kind of codifies you know, the things that it hears touch, taste, smell, and then how it relates that to his thought process. So when you’re talking to me, you know, people, people listening, everyone’s heard of, of AI, you know, they understand the idea of maybe machine learning, for example, right. You know, the computer starts off like a baby, you know, takes in all the information and it works it from there. But when you say cognitive AI, I don’t think a lot of people really understand how that relates in the tech field. Like, you know, how is it that, you know, what’s the difference, I guess, is my point from what people traditionally understand is AI maybe machine learning versus the cognitive AI that your company specializes in.
Leonard: Yeah, sure. And the best way to explain it is through an example, right? So one of our core focus in terms of product and strategy is working with the oil and gas industry. In fact we have a great investor plus commercial partner British petroleum is one of our key investor and a big customer as well. And so what we’ve done with BP and we’re extending the same product to other oil and gas company is we’re able to combine a lot of their data analytics and ML models that they may have used or have worked with other startups or other AI companies. And basically we have now added some expert knowledge coming from the operator of the refinery that’s been there for the last 20 years. So in a lot of operational situations where you have a large complex you know, facility like a refinery, think of this as a super complicated factory, that is four stadiums, large in size and footprint.
Leonard: So how do you manage the scale of complexity and operation? And it’s, it’s tough to use just the numeric models that are out there, where you pump in a lot of data. And one of the key reasons why a lot of these models fail at a certain point is you just don’t have enough data or a complete set of data you need to build that specific model to then, recommend the decision or give you an output. So when there’s broken data, when there’s unstructured data, a lot of the current AI models fail because it just can’t function. It just needs lots and lots of data for it to work and to give you a recommendation. So because of that limitation we’ve branched out into a symbolic AI where we can build in the expert knowledge, like literally the veteran who’s been there for the last 20 years and he, or she knows exactly when I see this kind of conditions, I do this.
Leonard: And then, because the temperature is such today, or, you know, depending on other factors and, and senses that, that he or she can read from, you know, I would do a search to take a certain action, which may or may not be a sort of intuition, right. That I’ve been here for 20 years, I know how this refinery works. And so we’re able to codify that kind of expert knowledge, intuition, and experience into the AI model. And maybe just to share a little bit of heritage on the company – the technology itself was originally basically matured in the NASA Mars Rover program. So our CTO Mark James, was the head of AI for the Mars Rover program. And the reason I bring that up is because if you think about the first piece of the Rover, you know, going on Mars and trying to navigate around Mars while doing experiments that it was told to do. So at that point, you, you think about, there is no Google maps from us, right? No one’s been there. So it has to take limited data like temperature and tilt angle and the direction of sunlight, and yet has to make decisions based on very, very limited data. And so that’s how the whole cognitive AI came about is through that NASA and Caltech program with NASA for the past 20 years.
JC: Yeah. I mean, that makes sense, because I remember too, I remember watching that when the Rover, you know had landed. And I remember that one of the things that they had talked about was, you know, we lose communication and we can’t just remote control this thing all the time. And it has to be able to be self-sufficient in case we can’t give it commands. And so is that kinda what you’re talking about? Like that self-sufficiency of being able to take it that step further and saying, yeah, I’m not getting a command to do this, but I know that this is my goal and what I need to do. And so here’s what I’m going to do next. Is that the kind of cognitive AI you’re talking about with absolutely. Yeah.
Leonard: To be exact the communication lag is seven minutes, right? You’re talking millions of miles between Mars and Earth. So in seven minutes, there could be a lot of damage done to the Rover if it tumbled down the slope or fell in a crater. So that’s exactly why this independent thinking and decision making based on very limited data is so critical – mission critical.
JC:. Oh yeah. That’s fascinating to me. I had no idea that you guys had kind of a NASA, you know, background or history and whatnot. And one thing I did see on your website as well on and by, and by the way for the audience here, if you want to go to beyond.ai if you’re following along so to speak, that’s their website. But I do see, you know, right now we’re going through a Corona you know, COVID pandemic right now. You may have seen it on the news once or twice. And I see that you guys are talking about predictive modeling with this. And I didn’t know this before we got on the show. Can you tell me a little bit about how your company is using its technology to help or to analyze what we’re going through right now?
Leonard: Yeah, absolutely. And yes, we were all living through this pandemic together and I hope everyone’s staying safe and doing all the right things. So we as a company we also have big ambitions in the healthcare industry. And we were fortunate to have had the opportunity to work with, actually, Cleveland clinic. And we actually had a press announcement a couple of weeks ago where we worked with the Cleveland clinic to build a COVID-19 prediction model. So basically what the model does is it takes in again, a lot of numeric data, like the number of hospitals in the area, PPE supply and volumes of that doctors and density of people, mobility factors like how many people drive cars versus public transport versus other modes of transport. So it takes in all of this data. Plus of course, the epidemiologist infant expert knowledge, the doctors and the clinicians, and it combines all of that into a prediction model and what the and right now, as we speak, there are 16 hospitals in the Florida area run by Cleveland clinic that are actively using this model as we speak. And what the model provides is it’s able to predict based on the input and the expert knowledge from the doctors and the hospital system to predict what would be the curve of number of infections, number of ICU admissions how long a hospital stay per patient, and therefore able to help hospitals basically prepare the entire supply chain, whether it’s hospital beds, number of ventilators, ICUs and PPE, right, which is so critical. So they use that in their daily operations of the hospital to prepare the supply chain and at the end of the day basically give patients the best quality care possible.
JC: Well, that, that I find fascinating because a lot of things that I’m seeing on the news right now is, you know, a lot of people assume that because the pandemic is hitting, like, for example, in America, they’re saying, well, we don’t have enough PPE, but it’s not that broad of a reality. There are hotspots that don’t have enough. And then there are cold spots that have too much, like I thought, for example, that nurses were, you know, in supply and that’s true in hotspots, but in other places, you know, they’re being laid off. Right. And so what I think is interesting about your model is that it sounds like any way that it helps create efficiencies within that. And especially with supply chain, which means, tell me if I’m wrong here, but does that mean that essentially when you have this issue of too much in one area that doesn’t need it and not enough in another one that your model could basically predict, analyze that and reroute resources to make sure that the right amount of resources are getting to the places that needed it and, and taking away from places that don’t quite need it just yet. Is that how it’s kind of working into the supply chain, more of an efficiency algorithm?
Leonard: Yes, absolutely. That’s definitely one of the outcomes. And one of the benefits of using this, and if you think Cleveland clinic obviously is a huge healthcare operator and on the East coast, so in the 16 hospitals, so they would be exactly using it for that purpose, right. Is to optimize the supply chain that if my prediction is I’m going to have a swell of patients, you know, in the next two weeks, whereas another area, I mean, I’d have such a steep spike. Then obviously I will move my supply around and personnel and ventilators, you know, whatever the related equipment is to the spots where they need it most. So absolutely.
JC: Are you struggling to generate warm leads fast for your B2B tech company? Do you wish you could hand up the entire process to a team of professionals and have leads flowing in daily to fill your pipeline, convert to big ROI sales? At Infinity Marketing Group, that is exactly what we do. We are lead generation masters, who specialize in the B2B tech industry. So whether you’re in hardware, software, or even tech services, we here at Infinity Marketing Group can turn your company into a lead generation machine. So for more information, visit our website at www.infinitymgroup.com. Email us at infoatinfinitymgroup.com or give us a call at (303) 834-7344. We look forward to helping you realize your revenue goals.
JC: What about the human factor though? So I learned how to fly planes at a young age. I was flying planes about the same year that I was driving cars. I was 16 when I had to fly. And one of the things that we were taught is if you get into a situation where you can’t see out of the cockpit, so maybe you’re in clouds, it’s called the IFR instrument flight rules. You have to trust your instruments. Yeah. IFR. So you have to trust your instruments. But one of the things that they talked to us about is that if you get into a situation where you feel like, for example, you’re like, I feel like I’m steady, but my instruments say that I’m leaning. You have to trust your instruments. The reason why I bring this up is because my, you know, cause when you don’t trust your instruments, that’s how planes crash because your internal balancing mechanism in your head thinks one thing, but it’s wrong.
JC: Your instruments are basically never wrong. Essentially is what the lesson is. My question to you is if your system can tell people, listen, you need to reroute, this does PPE to this place because you don’t need as much here or there. Do you find though that the human factor of making that decision changes the outcome? So your system is like, no, this is right. We know just on these stats that this area needs more, this area needs less, but what do you say about the people staring at the data and then maybe trying to override it? Cause they think they know more like, have you seen something like that happen and how do you work with your clients to help them trust their instruments, so to speak?
Leonard: Yeah, that’s a great point. I think, especially in the world of AI that you know, the school of thinking is that it could automate everything, then will take over humans. So that’s one sort of a line of thought and that’s, you know I think it’s far from the truth. But to your point first of all, you know, avionics background, I worked for Airbus still love those big big toys.
JC: Oh yeah, absolutely.
Leonard: First of all, I think there are probably two parts. One is it is also part of so when you build an AI model, whether it’s a symbolic model or a conventional AI model, it has to be taught and learned, right? So some of this expert knowledge that you were talking about there has to be a way to feed that back into the AI model and for, for us that expert knowledge and or industry guidelines or regulations all of that is very useful knowledge base that we build into our symbolic model. So as the machine just gets better and better as you provide that feedback into the model, so I think it’s, we can call it reinforced learning, or basically, you know, if the machine, for example, today, based on certain circumstances, it’s recommending decision A, but the operator chose decision B. So we then have to have a way for that feedback to go back into the model. So that the next time, the exact same situation arises, then the machine knows, Oh, previously I said, I recommended a, but they saw on the human input and expertise is actually decision B. So that’s how our model..
JC: How do you get the humans to trust it though? How do you con it could, because it’s a gradual learning thing, just like all AI is at what point, how do you get the humans who are having to make the actual decisions? How do you get them to trust that the AI is smart enough to make the right decisions? At what point, you know, do you work with your clients? Do you have a system in place where you say, listen, it’s going to learn for about this long, but now at this point you got to trust what it says. If it says send 10,000 gowns and facemasks to Boston and, and reroute it from Kansas, then you have to do it like at what point? Yeah. Do humans know that they say, okay, this is smart enough now. And now I can trust what it says.
JC: That’s really, my question is, cause it’s such a gray area, right. It’s not a matter of that. It’s not that people don’t trust that AI is very smart. I think the problem is that most humans don’t know when it got smarter than them. Right. So like, I guess I’m just curious. Has your company solved the human factor to when humans should now say yes, I trust the software. If it says, do this, I’m going to do that. You know, have you, have you discussed with your clients at what point they should make that transition to allowing the system run itself a little more?
Leonard: Yep. That’s a great point. I just, there’s an analogy. I have a Tesla and then there’s an autopilot mode, right.
Leonard: Of course in the beginning, no matter what you do you put two hands on the wheel, just, you know, more and more as you drive the car, you figure it is pretty good. And it’s probably better than me when I’m sleepy on the wheel. So to that point what we do is, first of all for a project deployment we have a period of POC, proof of concept. So where we ask the customer for data, whether it’s three months of data or six months of data, and then we put it into the model and then show them what the model has predicted and they can do a so-called back-testing on some of the decisions or the results they were expecting manually versus what the AI model predicted. Right? So in the case of the COVID prediction model we were working with the doctors in the Cleveland clinic on a daily basis to fine-tune the model based on some backtesting. And I think at the end of a three to four-week period of testing and tuning they were very happy with the accuracy of the model. And I think from that point onward, that’s how we were able to deploy across 16 hospitals. And they’re using that on almost a daily basis.
I like that analogy of the Tesla, by the way. Cause I think a lot of people can relate to that idea of right now of, you know, automated cars. Right. You know, especially when you hear about, and unfortunately, you know, you hear on the TV like that, you know, one went wrong and it crashed and they’re like, Oh look, we can’t trust it. And it’s like, yeah, but you really can’t trust yourself because about 30,000 people a day, you know, like crash their own cars, you know, so it mathematically speaking, you’re far safer, you know, letting the computer do it, but you know, again, it’s a human factor thing. Let me ask you this, what are some of the big milestones that beyond limits have reached that you are personally proud of?
Leonard: So I think as a company we were founded about five, six years ago here in Glendale, California. So without I think, rich technology heritage from Caltech and JPL and NASA, we were able to quickly breakthrough in the oil and gas industry, which I think a lot of AI companies out there, you know, there’s a lot of facial recognition AI, a lot of NLP and language processing and chatbots but to go into an oil and gas industry and go real deep to solve a very complex problem, I think that’s what we were able to achieve with some of our early customers in oil and gas. So we’re pretty proud of that. So the next couple of big milestone is clearly we are at the finishing stages of our series C investment round.
Leonard: We’re very pleased with the results so far and we’ll be making some announcements in the next couple of weeks. But in terms of operational milestones we went to do a couple of things. One is to expand globally. So Asia Pacific is a key market for us for many reasons, right. We all know about the growth in the different economies. And also there are certain countries in Asia that are leading the way in FinTech, for example, or healthcare, for example, right? So market expansion is one and then also because of our oil and gas, you know, you gotta be in the Middle East. So we are also doing some, some good expansion in the Middle East and in Europe as well. And then the second would be around new verticals.
Leonard: So the healthcare that we just talked about, that’s clearly, that’s something we wanted to break into in 2020 and 2021. And then also because of our presence now in Asia, specifically in Hong Kong which is a big financial market and big IPO market. So we have some fantastic strategic partners in FinTech that we’ll be working with.
JC: Is there any that you can tell us about right now about any of those partnerships or are those still under NDAs?
Leonard: Yeah, I think right now in the NDA, but two or three weeks, we can definitely say there are large financial institutions. One is a bank..
JC: I had to try..
Leonard: So certainly and to share with you also, we’ve only been on the ground for 90 days. So we’re very proud to have really set up the infrastructure across four countries, Singapore, Hong Kong, Taiwan, and Japan. So now we have companies and we have teams boots on the ground, engaging customers and close a couple of deals already. And this is all in the midst of a global pandemic. So we’re pretty proud of 90 days on the ground in Asia so far. That’s, that’s amazing that that’s, that’s a huge feat.
JC: I have two more questions for you. One is a personal question. And then one is a business one at the end. I’m curious what, you know, you know, what you do now, of course, right. Obviously, but what did you want to do when you grew up? Like when you were a kid, what was your, when I get older, I want to do this. So what, what was your childhood dream?
Leonard: So it goes back to airplanes. So literally when I was you know, three, four years old, I wanted to be a pilot. So at the time as I went through college, first of all I think the family didn’t have the means to put me to flying school at the time. And then also I didn’t have 20/20 eyesight nowadays, you know, they’re a lot less stringent than, you know, I have had Lasik surgery since then. So then I thought what’s the next best thing is to be an avionics engineer. So I’m still in the cockpit, just at the planes on the ground, right. So I still get to see the plane touch everything and even better, I made these avionics equipment. My first job was with Honeywell or allied signal. If those in pilots in the industry, the Bendix King range, that was my first job.
Leonard: I was making the GPS and the radio and the transponders. And so I had a lot of fun. And then from there I moved to it and then business. And but the last I had the chance to work with Airbus and there was at totally, you know, different level as part of the leading some of the digital transformation efforts. And there, I got to see, you know, the A380 and A350, and being in these massive factories that are as big as Disneyland building the A380, A350, and even things like the Eurofighter, for example. So yeah, you know, as a little kid growing up that’s just a dream come true to really see in person how these magic machines put together.
JC: That’s awesome. You know, I’m the same as you. I mean, I couldn’t afford flight lessons when I was a kid either. I got lucky we had this. It’s the weirdest thing. Cause I went to high school in Tacoma right off near McChord air force base. I don’t know if you know that, that Seattle area. So they had this program where they took like 20 students out of the whole district that they would give flight lessons to for free. It was like for lack of a better term and kind of a pun, it was a pilot program, but in both sense of the word, you know? And so I got really lucky. I got to do it for free. Otherwise, that’s definitely not something I was gonna be able to afford.
JC: And then I ended going to Embry-Riddle which I’m sure you’re very familiar with Embry-Riddle being in the avionics industry out in Daytona. So, you know what I was doing with my major, cause we talked about psychology. I started out in the human factor, psychology. We were helping to design the cockpits, not in the instruments, but where the instruments go. So a lot of what you did, you know, creating these panels and whatnot, we would put pilots through these cognitive tests and for example, track their eye movements while they were doing a flight SIM, because if you ask a pilot what their top 10 most important instruments are, but then you actually track their eye movements of what they actually look at the most. They’re not the same.
JC: So we would help redesign cockpits based on what pilots actually subconsciously thought we were the most important. So that’s just interesting. I just I’d like talking to definitely avionics people like that. It’s pretty cool. Alright. So my last question for you here, before we wrap it up again within the idea of the title of the podcast, the future of BizTech, I mean, you talk a little about healthcare, but I’m going to leave this open to wherever you want to go with it. What part of beyond limits do you think is going to be the future of either of any of the industries that you go after? You know, what is that 5 or 10-year plan where you see beyond limits really, really contributing to innovation in one of those industries?
Leonard: Yeah. I think so as we look mid to longer-term I think there’s two things. One is, you know, AI in general, it’s still a, let’s say an emerging technology in the sense that businesses are still trying to figure out if I invest X amount of dollars in AI, what’s the return that I get, and this is kind of the same transition. And I was at Microsoft at the time the transition from on-premise computing to cloud computing, right? So back in the day, people would ask, Oh, you know, why should I do cloud? I’m perfectly fine with my server under my desk. And, and then, you know, what’s the advantage of the cloud. And so I think we are going through that kind of a transformational shift in, in technology and, and the benefits that it can bring.
Leonard: So I think we, I firmly believe after seeing a few of those technology transitions that AI is, is definitely going to hit home on ROI and benefits and efficiency and all of that. So for us, clearly we want to drive kind of the cognitive AI industry and how that impacts the verticals that we’re in. So AI industry as a whole, and then secondly, to really make an impact for the verticals and industries that we work with. So oil and gas is a key one with our oil and gas customers like, like BP and the like, and then second in healthcare, as we just talked about, I think there’s still so much more we can do with AI in healthcare both on the clinical side, as well as like a hospital management you know, how do you do billing in the hospital? How do you do bed management in a hospital and so on? And then third is the financial services and FinTech. So I think we really want to drive the whole AI industry forward without cognitive AI IP blocks and then in these three industries that we really want to go after and together with our strategic partners in those industries.
JC: Well, that’s fantastic. Listen, this was a fascinating talk and I want to thank you again for coming on. If anyone listening wants to engage your company what’s the best way to go about that? Is there a phone number, email website? I mean, what’s the best way to get a hold of you guys?
Leonard: Yeah, sure. So first of all the company website as you mentioned earlier, it’s Beyond.AI, my personal email is firstname.lastname@example.org. So you can kind of reach out to me personally or kind of hit our website with some great you know, use cases and white paper that talks two to how cognitive AI is a huge differentiator versus the conventional AI feel, feel free to chat.
JC: Thank you so much again for coming on, sir. I appreciate your time.
Leonard: All right. Thanks a lot, JC.