Who’s Buying Skydio? – A Nerds Perspective
The sky do now. I called this clip who's buying sky. Do I promise you that wasn't clickbait I've spent a lot of time studying the company going all the way back to their Gen 1 prodotto? I thought that was incredibly disruptive. The first drone they put on the market right through their latest release, the sky do, and what I'm here to tell you is that the brilliant engineers that built this product out in California are using some of the most advanced leading edge technology available on Planet. Oggi. To create essentially a robot that has an advanced optical system and imaging system, so it can detect its environment. It'S got neural networks built inside with some of the latest chipsets the process. All and information and it's got incredibly complex. Software that's really really smart to process that information and make decisions on behalf of the drone. Ora, stranamente, that robot, that autonomous robot they've built just happens to fly so in a lot of ways all that high powered technology behind what Scott EO can do with this drone it's a bit of a parlor trick, because that same technology can do so much more. In another industry and I'll get into that in a few minutes, but I do love the fact that I can fly this thing full speed at a tree and it looks at the tree as an obstacle and finds multiple paths around it and chooses one of those Paths so it doesn't crash into the tree so from an obstacle avoidance and an autonomous flight perspective, there's, nothing even close to this on the market today, detto questo, that's one thing that this kind of autonomous behavior can do and again I think inside the drone It'S, wonderful, its whiz bang technology, but I think the IP behind it is something other bigger companies would snap up in the heartbeat, so I'm going to spend a lot of time talking about the technology, and I have to warn you I'm going to get super nerdy Today, because I get really excited about all three of these topics – I'm going to talk about because they really are leading edge.
So if you're, younger and you're in school study, any one of these three you're gon na do really well when you get out and if you're curious like me, if you're a nerd, you're, probably already looking into things like this, but this is the future. This is the next 10 20 years of our evolution around electronics that are in this drone today, so I'll get into them in a second. But let me talk a little bit about the company. The reason I think that they may be a target for a bigger company to grab up is that they're, a small company. There were a start up out in California, very small group of engineers working on their technology, they've grown a little since then they've got another venture capital funding, but there's still a small company compared to other drone companies or other industries outside of the drone community. So for a company that size it wouldn't take much from an investment perspective to buy them and by the IP and they've got investors. Ora, investors I've been involved with a couple of venture caps before companies. I'Ve worked with venture capitalists when they get their money back and either want a license. The technology or they want to sell the company. One of the two or though IPO, which means they're gon na, go out and put out public stock they're, non gon na. Go any one of those routes they're either gon na sell the company or they're gon na keep investing hoping that they're gon na make enough money off selling these drones.
Ma ancora una volta, Ci credo, even if they sold a million of these drones, the profit they make off selling a drone is nowhere near what either licensing or selling that technology to another company could generate. So I think they've got some pressure from the venture caps to say: let's, look at other opportunities to make money off this core technology, because there are companies out there that are sweating. The details – big name companies, Amazon, Google, self driving car companies like Tesla – that are trying to make this kind of technology work today in a bunch of different scenarios and they've got armies of Engineers that are coding it trying to build those systems together to make them Autonomous and predict things in advance of them happening, these guys have done it in a drone. I just think that's, absolutely incredible. So I'm definitely gusting about the engineering behind this and I'll get into those three different topics in a second, but essentially what you've got are three things that make this incredibly special and again, all three of these are different than any other drone in the market. Today the first is they're visioning system, the optics they've got trinocular vision and I'll talk about that first. The second is they're using advanced artificial intelligence. They have deep learning, networks or neural networks inside the unit that can gather data from a bunch of different inputs. It can weigh that data, make decisions and then control the drone to make a good outcome happen on the drone.
So it's smart enough to not only see obstacles but assess those obstacles find ways around those obstacles and bick pick the best path to get around those obstacles. So it's thinking and unlike a lot of other drones, because I get the question all the time about well. How is the obstacle avoidance compared to the others? That'S a different term? Obstacle avoidance is kind of a dumb thing. Obstacle avoidance is just a couple of sensors that are bouncing some kind of IR signal or radio signal off of an object coming up in front of it, and a drone will react to that. That'S kind of a digital decision like hey there's, something coming, stop don't go into it. This is predicting so it's, not just avoiding or reacting it's predicting so it's, looking at the environment and it's almost a 3d space, because it's got so many cameras on it and it's drawing a map of points in that environment in advance of being a threat. It'S already looking for the holes that it can go through to avoid those threats, so it's a whole different level of analysis going on inside the drone. So this is an incredibly smart product, so the first one is trinocular vision, which is the way it gathers. The data, the second is neural networks and the processing behind that they're using the jetson chipset, which is one of the latest internet of things. Chipset that has neural networks built in it's, got learning curves built into it.
It'S got the ability to train it in advance and add that to its experience when it's out there flying and the third thing, it has it's incredibly complex software, because programming normally anybody that this program will tell you that a normal program is typically a linear program Or you're executed command. The result comes in. You execute another command, so you're following a food chain, if you will decisions, neural networks are different. There they're a bunch of things going on simultaneously that end up in concert, so it's sort of like in a I guess. In an orchestra, listening to a flute player play the song on their own you're, only getting bits and pieces of it in a neural network. You'Ve got the entire orchestra, so there's a lot of things going on together to produce that symphony and all that's going on inside this drone. So again it blows my mind how far they've come with this advanced technology. They'Ve got inside the sky. Do too now, I'm gon na go a little bit deeper at each one of these topics, because I think each one of them are incredibly exciting and each one of them on their own. The way they've perfected it in the drone are valuable outside the company and that's. Why I'm saying sooner or later, somebody's gon na realize they've built this brilliant autonomous system that would be applied to other technologies and other industries outside of the drone and they're gon na snap them up whether it be the trinocular vision, the neural network programming inside or The software that makes the decisions about what to avoid and finding holes.
You can imagine immediately if, instead of propellers on this thing, you put wheels on it. You'Ve got an autonomous self driving car now course. There'S gon na have to be code, written and other sensors built in for safety, but the quarter of it is inside this drone. Allora, as a drone guy, I love flying it, but as an engineer and a nerd, I am absolutely blown away by the level of technology inside the sky. Do two drone – and I can assure you there are large teams of engineers in the government, in private sector and in universities all over the world that are spending tons of time and treasure developing systems just like this, and that small team added sky do nailed it In this drone now next I'm going to talk about the underlying technology that makes a product like this possible, as I'd mentioned, the optical system is trinocular vision, that's a really interesting concept and it's pretty important too, because I think we're going to see more and more Of that coming along the next is the neural networks. I'Ll talk a little bit about the Jets and chipset behind it and what neural networks are all about, and third I'll talk about the software and some of the packages that are used to write programs like this and the kind of problems that they're confronting writing a Program for a neural network now I know it's really super nerdy stuff, but trust me when I tell you it's, really interesting and I think that you're gon na see this type of technology moving to other drones over time, but probably more importantly, the IP inside.
This drone is definitely a valuable commodity that I could picture some other big company coming along and writing a gigantic check and gobbling these guys up and then applying it to their industry on their own. So I'll have to see where that goes, but first I'm going to talk about trinocular vision and man is this cool stuff, so stay tuned for that Music, the first leading edge technology, the new sky do to employees is incorporated into its visioning system they're using a Technique called trinocular vision, which may be a new term to a lot of people out there. I'Ll explain why it's important in a minute, but essentially they've, mounted three cameras on the top of the drone and three more cameras on the bottom of the drone which allow the drone to have a much more complete view of a 360 degree environment that it's flying Through but more importantly, it gives the drone a much more accurate perspective on obstacles in front of it and how close those obstacles are to each other it's, an extremely accurate way to gauge an environment in front of you. Looking for threats and safe areas to fly through now, it's almost impossible for us as humans to understand what trinocular vision is about because, as creatures were binocular creatures, we have two eyes and if I think back to the early days of enhancing images, I always think Back when I was a kid, I used to watch a lot of pirate movies on Saturday morning and the captain would always come up on deck.
He pulled a spyglass out of his pocket and extended looking for boats on the horizon. Now that gave him a really nice amplified view of what was out there, but it didn't give him a real good perspective of how close that new boat was to him or how close that boat was to other boats out there. The next big step forward was a set of binoculars where you actually had two lenses. You put them up to your eyes and you've got this binocular view. Now that gave you not only amplification of the image but distance perspective between those images. This takes it to the third level. So now I have three lenses that more accurately enhanced that distance between objects and a real good test just to illustrate that is look at the wall right now or find something on your desk to look at and close. One of your eyes, you immediately lose depth perception. You can't, really tell how far away that object is from you, the minute you open your eye. You now have a perspective between those two eyes and your brain does all kinds of calculations to tell you this one's closer than that one it's the reason we can sink a basketball shot I'm, not real good at the habit. I know a lot of you probably can, but binocular vision gives us perspective between those objects, and it keeps us from bumping into walls. If you had a third eye, you'd have a much more accurate view of the world and that's exactly what they're doing with trinocular vision now that's, not something that occurs normally in nature.
There is one odd creature called the mantis shrimp that in each eyeball has three separate regions, so it's got dual trinocular vision and looking through those three lenses, the brain of this mantis shrimp can actually determine a much more accurate view of its environment. Whether it's chasing prey or it's, avoiding a predator and that that's an incredibly cool concept from science. That really has intrigued me for quite some time. I can tell you that that trinocular vision is something that all of the artificial autonomous robots driving cars, robots and factories really need to get that accuracy down to tiny little dimensions so that it's not going to bump into things in a warehouse where you're not going To drive into a curb if you're, having autonomous car driving around the city, so trinocular vision is leading edge technology that a lot of those advanced labs are working on, and here it is in the sky. Do two now the next section deals with what do I do with all those images, because each one of those cameras is a full HD resolution, full screen and camera that's firing a lot of digital information back at some central brain inside that copter that has to Process all that information and not only process it read that data and make decisions on that data to alter the flight path of this drone to get around those objects so stay tuned. Then I'll talk about the brains of it next, because man is that cool stuff.
The next big engineering innovation that sky do is introduced on this amazing flying platform is around the processing unit they've chosen to put inside the quad. Now you can think of that as the brains of the quad, and maybe the cameras are the eyes of the quad, but that central processing, unit or CPU, or in this case we're gon na call it a chipset because there's a collection of chips that actually work Together to make that possible, it has to keep track of a lot of stuff, so you've got six high resolution cameras that are firing high definition, video back at that CPU at 30 frames a second that's a lot of information to process and that main chipset has To not only accept that information, it has to correlate it put it together and draw this incredibly complex picture of the environment around the quad. So it knows exactly Rob. Stickles are, but in addition to that, that CPU has to identify targets that it doesn't want to run into and then find holes that are big enough for the quad to navigate through to avoid those obstacles. Oltre a ciò, it has to keep track of its flight characteristics, it's ulema tree the GPS, coordination of words sitting the control signals and video feedback to the controller that you're using and the phone that you're using to view the camera so there's a lot of Stuff going on inside of there and honestly most chipsets out there, can't handle three cameras, let alone six cameras, aggregating that data and be able to make decisions on it.
So with the brilliant engineers at Scott who have chosen to do is to go with the latest. Cutting edge chipset out there and it's called the Jetson XT it's a chipset from Nvidia that was purposely designed to handle these kind of complex tasks in autonomous environments. So you'll see that used in robots that make decisions for themselves a lot of that core technologies used in self driving cars. Anything that needs to take in a lot of information process that information and then make decisions on it now, in addition to that, that chipset also enables something called neural networks now that's a complex term there's a lot of studies going on out there about how these Neural networks are developed and what they can do, but essentially what a neural network is it's it's, a framework sort of like our brains that can handle multiple inputs at the same time, correlate those and then based on the software which we'll talk about next, make decisions That have positive outcomes, so in our brains, we do that all the time, if we're in an environment, we're looking at something or listening to it, we're touching it with our hands. All those inputs are correlated together in our brain it's, a beautiful neural network and the way I'd reckon it out in the field is if the best example, I can give you if you're at a party and there's 50 people in the room and you're talking to Somebody we're kind of we're kind of single threaded beasts where I can keep a conversation going on with one person at a time.
The other 48 people in the room are just noise. I can't listen to somebody else. Unless I take my attention away from that. First person, so I can kind of fake to conversations but I'm not really giving either of them my full attention. A neural network can keep track of 50 conversations at the same time and give responses back to input that it got so it's. A very complex interweaving of connections inside the product to actually be able to process all that data simultaneously at speed and make decisions. Now the software that's laid on top of that is where things like artificial intelligence comes from machine learning, deep neural network interaction, all that stuff happens inside here and I'm here to tell you that doesn't belong in a drone. That technology is on the leading edge of most scientific research going on today, around autonomous BOTS, and when you hear things like computers, beating chess players, those are neural networks that learn from previous behaviors and apply those patterns to future behavior. Allora, in a lot of ways, the drone is predicting what's going on and I've had friends that fly this and I agree with them that almost feel the drone is treating the pilot as a co pilot it's just that smart it's, smarter than the person fly in It and it's because of the chipset they've chosen the software which we'll talk about next and those neural networks that can actually process more information and react more quickly than we can as a pilot it's, absolutely brilliant.
The amount of technology that's going on inside this product. So I know I'm gushing, but the nerd in me is lit up when I look at stuff like that and think neural networks belong on the leading edge of science, not at a drone for under a thousand bucks. Quindi, di nuovo, sky do is packing a lot of advanced technology inside here that I think there are other companies in the space that when they realize what this can do are going to be salivating to get their hands on that technology. And I think this is a company that other companies and different industries will to talk about sort of at the end of this are going to snap up in a heartbeat, quindi rimanete sintonizzati. Then I'll talk briefly about the software and then I'll give you a couple of I guess bets of who I think may be, making a move on Scottie, Oh down the road in this third section. I just want to talk briefly about the amazing software that's required to actually make all the magic work inside this quad, because you've got cameras that are feeding back information. Those are fairly pedestrian and they're off the shelf. Cameras you've got a CPU set in there. That'S advanced beyond anything that quads are flying today, that's important, because it can process that information, take all those inputs and generate an output, but you need software to actually make those decisions for you, so the brains of this, the artificial intelligence in this is really controlled.
By whatever patterns are recognized as good versus bad and that's all done through the software now and days gone by programmers used to write in a fairly conventional fashion, I'm, a programmer myself from way back and you'd write a program that was kind of a linear track. Where you'd have inputs, you'd evaluate those come up with an output and make a decision based on that or maybe you've got conditionals, where you do this or that based on an input in this kind of environment, where you've got neural networks, you've got to aggregate all Of those inputs simultaneously, compare it to an experience grid or something you've done before and tell it when you see this, do that so there's a lot of inputs that are happening simultaneously, which requires a whole new framework for for coding. So you've got things like tensorflow and pi torch, which are brand new programs that allow all the wizards that are programming in this environment to help this get smarter and in some ways, it's kind of scary, because the drone and this kind of artificial intelligence, especially with Neural networks involved can actually be smarter over time, so it gets better at dodging objects and get better at following you. It gets better recognizing a person over time, so it becomes better at its job and a lot of ways that can be a little bit. Creepy it's got a bit of a terminator feel to it where it's smart enough to find me if it loses me in the woods, especially if I'm carrying a beak and it's.
Looking for my GPS coordination but visually it's. Recognizing me, as a person and the more it sees me the better. It knows me it's someone like a baby growing up where at first it likes you because you're smiling, but over time it recognizes your face is somebody that wants to pay attention to. So the software in here is really the guiding principle by what this quadcopter does. When it's confronted with an input – and I guess the only analogy I can use is if you've got kids you've raised kids, you know what it's like you have a lot of influence over those kids when they're young and you can set patterns for them to teach In the moral compass of what's, good versus bad, maybe not put your hand on the stove because you're going to get burned, but you develop patterns with the child that, as they grow, become intrinsic to every decision they make going forward and they learn on their own. And they kind of incorporate those into the basics you've given them well, the software actually sets the base level of what decisions are going to be made, E poi, on top of that, the neural network and that Jetson chip set makes it even smarter over time. Così bene, you've got in your hands. Here is really an advanced robot that can make really complex decisions based on the inputs that I was talking about a few minutes ago. Gps coordination, the streaming coming back from the camera, the telemetry being sent back from all the sensors inside of it and give you the most positive outcome, ma, ancora più importante, gets smarter over time.
Quindi, di nuovo, all the tech that's built into this just happens to fly. I think it's great, that they built a drone around it, but the drone itself is fairly pedestrian. I mean drones are something you can put together in your kitchen table. If you buy the right amount of parts, the fact that sky do built all of that intelligence and neural networks and artificial intelligence, deep learning all that impressive stuff inside there. This is something that another company would look at and go the intellectual property that developed this. The IP is something we need to own because to develop this from the ground. Up is gon na. Take us years of work, hundreds of engineers – and I still can't, get past the fact that that small team and they were a mighty graduates. I get that they're brilliant people, but that small team kind of a startup and I'm rooting for him put this together in the face of all the competition out there. They put this brilliant drawing out in the marketplace, so love it or hate it. You can't deny the fact this drone doesn't belong in the market today. This drone should be five years from now, ten years from now based on the tech that's inside of it, and I think when you build something like this I'm gon na call it a unicorn when you build something that's unique. Other companies are going to take notice and they're gon na want to write a check to own that IP, quindi rimanete sintonizzati.
Then I'll give you some predictions and I think three different market segments that are going to be really craving this kind of technology very soon and that's. Perché? I think there's a good chance that sky do hopefully will continue as a drone company, but the IP behind this is going to be licensed or sold to somebody else so stay tuned for that. I hope I've explained why I'm so excited about the engineering that went into this new sky. Do two and again any one of those innovations would be enough to pay attention to the strong, but together the orchestration of all those things I've mentioned really makes this way. Bigger than a drone, it just happens to fly like I said before, but the robot inside they're, the brains, the neural networks, the processing, the cameras, the trinocular vision, all that stuff is absolutely applicable to other companies today in other industries. Oggi, now I'm going to talk about three different industries that I think should be paying attention to this and, Credo, are gon na make a play for this company. The first one is an obvious one: it's the drone guys right. The drone guys that are competing with this are looking at it going. You know we have crash avoidance today, but that's dramatically, different than obstacle avoidance and automated navigation like this does so most drones will see an obstacle and in some cases the Dumber drones will just stop and it won't go any further, che va bene, you're, not gon.
Na crash into a tree, the smarter drones will actually see it and try to find a way above it or maybe even a way around it, but it's, not a sophisticated environment where it's constantly looking for holes, it's just scanning whatever areas in front of it to Find solid objects versus non solid objects and it's heading for that non, solid objects, so they're kind of dumber, but they're not autonomous flight. This is autonomous flight, so I think the drone guys could pick up this IP in a heartbeat because, Onestamente, if they charge your engineering teams with developing something like this they're years away and that's a huge market gap, so writing a check for the IP that's. In this and applying it to their next generation drone, I think it's a no brainer, so I think the drone people are very excited so, whether it's DJI, oh I tell X, Dynamics you name the company. I actually think I think that I shouldn't say this, but I think GoPro ought to be looking at these guys, because GoPro is one of those companies that I was so rooting for when they came out with that karma product and they had some issues out of The gate early on with the engineering – maybe they didn't, put it through enough beta testing – they didn't, send it to me. I would have found some problems with it, ma comunque, GoPro could be a company that would snap them up, throw the go puller label on this thing and have an autonomous drone.
That'S used for action, filming like GoPro karma, was designed to do, but any of the drone companies would certainly be interested in this, so let's think about other industries car guys there are a ton of car companies right now that are racing to that golden place of Automated driving so Tesla's working on at GM's, working on it, lyft and uber both have programs going on. You really have the basis of an automated driving car right here. The ability not so much that we're going to use six sensors. Maybe we were to use 36 sensors and a couple of different chipsets, ma al centro di esso, having the ability to aggregate or let data find pedestrians, find obstacles and navigate around them or stop. The car is something that would immediately apply to that space, and I know there are billions of dollars being spent on autonomy in that driving space. So I think the core guys could make a play for this. This third category I'm just going to call the other guys – and that includes every other major industry that currently is doing research and artificial intelligence or autonomy for products that they produce, and you can probably pick the winners in that space. Companies like GE Samsung, mela, Microsoft, Amazon. So all those companies today have large teams working on advanced technologies like artificial intelligence, neural networks, machine learning, autonomy and they're. Looking for this kind of technology to put into their products and move forward with that, and if you're in college right now, they're hiring people left and right that can do programming on the software.
I was talking about that have experience in neural networks or that kind of decision making tree so that's a great place to go if you're in college right now, but all those engineers that work for sky do if that company is bought, work for the new company And they're, bringing all the intellectual property that's developed into this product, the programming skills, the technology skills behind it. Quindi, di nuovo, if you're a company, the size of GE, buying a company this size that's around here for them, it's not that big a deal or somebody like Amazon, that's, running warehouses, all over the country, wouldn't it be cool if they could take this technology and immediately Incorporate it into the robots that are running around those those distribution, centers picking stuff off the shelves to do visual indications of yes, I've got it or the box is damaged and oh that's, a human don't run into them because I'm going to whip down this down. This row with this robot, so all of those companies have a need for this type of autonomy, crash detection and decision making that this product brings to a drone that we're flying. Quindi, di nuovo, I don't want to sound down on this. I really hope that Scott EO stays independent but they're, backed by venture capitalists which again invest in a company, let it grow and then either IPO that company license the technology or sell a company to somebody else. So the exit for these guys, I don't, think long term is going to be a drone company.
I could be wrong on that and I hope I'm wrong on that. But I think that the technology inside this product is going to be very appealing to other companies that are going to be willing to buy them and buy them soon before they become too popular. And start dominating the drone market and become too expensive so anyway, that's my opinion. Everything I've talked about today is pure speculation on my part, but again as an engineer and a nerd, I love this kind of stuff and I'm, hoping that you've stayed with the clip. This long – and you guys love this stuff too, because for me flying a drone is fun and I love it. I get out every Saturday put the drone up when it's not raining and just have a great day out there flying, but I also want to understand the engineering behind the technology because drones in general shouldn't be where they are today. There should not be this sophisticated. So there's a quantum leap between RC products that I've flown. Since I was 11 years old and drones. Drones are space age stuff. They came out of nowhere. So the fact that they're, this sophisticated in a machine that can fly to navigate can avoid obstacles can recognize. People to me is just mind blowing, so I hope you guys are as excited as I am anyway that's it for today. I'Ve got links below. If you have any questions betting, I covered drop down there and leave me a comment.
I'D love to have a conversation about any of these topics I talked about and if you're interested in this kind of stuff, I can spend more time talking about it at a future clip but it's got me jazzed and I love this technology.