 like a run a run kind of memory lane kind of show today okay with Cliff Spradlin and Cliff Spradlin doesn't have a shot of Cliff Spradlin there he is that's Cliff Spradlin he's 31 30 30 sorry I'll make you older than you are and I knew Cliff when he was one week old and he was really a fantastic kid even then really yeah he was talking about code yeah that sounds right I remember that I knew you would yeah back when okay when his parents Cliff and and Phyllis we're next door to us in our law firm in the Davies building 31 years ago yeah and they were tax attorneys back then that's right now now Cliff is 31 he's got a life that we're gonna talk about he's got a wife and child outside the whole thing has come full-turn it's really fabulous yeah so nice to see you you too so I've changed yeah you changed you know what you were like that now look at you so you got into computers and that was right that was a thing that really had to happen that was a thing you were born for we all knew at the time yeah I have photos from when I was a kid from like when I say kid I mean when I was two or three like using computers actually Milo is now too but yeah well this Milo he's one and a half but I think I just I definitely fell into it I didn't really pay attention in school and I just played on computers all day and then suddenly people started paying me for it and so then I worked on video games which is like a kid's dream come to true and rocket software and most recently software for driverless cars or autonomous I'm talking about that so tell us a little about how you evolved into the into the world of software engineering I mean professionally well I used to play video games and I used to pry into them to try to figure out how they worked and I met friends online the internet was pretty new at the time but there was ways I could find people who are like MIT professors or just just programmers or software engineers on the internet it's the perfect place to find other software engineers right yeah sure well sure what better place yeah and I remember I remember you came to our office back when and you were either late teens early 20s and you told me you were working on the and I didn't even know what it was at the time the Android operating system that's true that was I worked on the first Android phone ever wow you had one you showed it to me yeah yeah it's it was so different it was different for me too and I had been working on phones previously yeah and now now it's like I don't know maybe even more popular than the iPhone so you're at the top echelon of you know a project software engineers that can come and do things that are disruptive I mean that the front end how did you get there I mean what makes it why can't I do that I think you could well I started just by learning how how computers work at a very low level like I wanted to really understand how video games work and as it turns out video games push the the limits of computers right in every dimension graphics and networking and just like simulation right like like Grand Theft Auto or something is a huge simulated world yeah and basically based on that experience I was able to work on rocket software and it's it's actually the same thing it's the same kind of math linear algebra and you need to simulate a rocket going to space before you launch it so it's rocket science really rocket science or it isn't rocket science you know I work to SpaceX and I still don't know who was the rocket scientist I would probably say the people in the propulsion department like to make the engines but then there's other people who just like you it's not enough to make an engine right it has to go somewhere so I don't know all right well just so I show that in what about Tesla you work with Tesla for a time yeah so Tesla is amazing Tesla is not like any other car company on the planet they have the hardware team and software team integrated they try not to buy parts from external companies because it'll slow them down huh and so of that yeah so it's like the perfect like incubator for new technology right and Tesla's are full of new technology whatever you think you can do yeah and I've seen Tesla release something from concept to like in customers hands on the in the order of a month or less that sounds like things that happen in China yeah maybe they're trying to copy it yeah so now it's somehow you evolved all of this into autonomous cars and I like oh and you're working in in the San Francisco that's where you live right now you have to live in Silicon Valley why why because that's where it's where all the energy is for new ideas and passion and just it's everybody has worked at every other company in that area too and so yeah they move around so all the ideas spread like wildfire whenever somebody has an idea everybody else starts working on too and the same thing is true with autonomous vehicles like there are dozens of companies in the Bay area separately working on autonomous vehicle not necessarily the big guys maybe little guys too little tiny guys you've never heard of two huge people like Apple secretly working on it well secretly but it's an open secret in the Bay yeah yeah so what does it mean to work on software for an autonomous feel I really want to spend time with you yeah because it's so important we know that it's going to happen right you know the people will demand it you know the American psyche is in love with the car they love with the kind of car but the macho car but you know electric will take over and part of electric my right part of electric is autonomous they go hand-in-glove I think it's nice because you know you can you can much more easily charge a car by having it drive over a pad there's less mechanical things like that I've actually seen autonomous vehicle technology on gas cars too so it's not necessarily it's not necessarily I'd say that they're too parallel but equally important steps for cars that hate so it could happen before we have you know huge explosion of electric cars could happen on gas cars yeah I would say that Tesla has made the first big step towards autonomous vehicle technology released to the public yeah and that's on an electric vehicle platform so that's why it's definitely got linked inside of people's mind yeah I would I would link it because in the end gas cars will go away and so much political pressure or they used to be until January 20th they used to be used to be political pressure for that now we'll have to see but anyway so what is it you know you have a brain yeah in the car and the brain what what is what input is the brain getting so the hardest part for for the software is actually the perception seeing the world seeing that there's cars in front of you where the lane lines are making decisions about changing lanes overtaking cars taking exits actually if you had a perfect view of the world then driving the car is actually not that hard it's it's it's actually uses very similar code to what you would write for a car in a video game it's it's you know the car can only turn by its steering wheel and accelerate and decelerate it doesn't have to do very much but so the really hard part is seeing what's around you and so that's what all the new developments been on and then as an extension of that where are you yeah that's actually another hard problem to you and so that there's a graph yeah perfect these are the these are the sensors of the the first Tesla car so what you see there is in the front of the car is there's a camera you can see in front of it yeah there's also a radar the camera radar have different pros and cons of them of what they can see in front of them and the kind of like peanut shaped area around the car is actually it's ultrasonic sensors which are the same sensors used for parking your car we repurpose them for sure just seeing what's around you just they don't work as well at highway speeds but they still work why all three okay well going straight the most important thing in autonomous vehicles is not hitting the car in front of you yeah that's as long as you don't hit the car in front of you you're kind of okay and so that the radar is exceptionally good at seeing metal objects in front of you and that's what you've seen in cruise control and things like that and you need the camera in order to see the lane lines radars can't see lines on the road well that is a huge problem and that's what the the last like the long tail of the development process is because most of the companies working on this already do very well when they can see the lane lines but when they start getting spotty and or disappear entirely that's where it it starts getting into the realm of artificial intelligence come to Hawaii because we have lousy land really we can what about bumpy road in the great environment to do the laboratory yeah I mean that's and bumpy road well so I mean you guys might not see it as as real as I see it because where I live I see autonomous vehicles all around me I see Google's driverless cars every day just driving around me now they're autonomous but they have safety drivers in them just like holding the stick like their hands around the steering wheel but the car is totally in control it stops in front of the traffic lights it makes turns I'm laughing because you remind me of a friend we used to have living in our house yeah his name was safety man who's a stuffed doll okay and we would sit him up in a couch while we were away and people would look into window and see this yeah six foot five guy sitting on the couch in a living room yeah he was safety man well in this case the safety driver is actually doing something because the software is not perfect yet and that's the last bit of development is what happens if like a cat jumps in front of you or there's a baby on top of a bed of leaves like our software might think it's a bed of leaves or it's a baby it's actually hard for it to differentiate let's say it's actually both yeah you have to teach it yeah you have to teach it and that's where you get in the artificial intelligence from so but let's so let's go back to the lanes I mean you know one question that strikes me it's always struck me about autonomous cars is that maybe we need to rebuild the infrastructure in this country and put in magnetic lines or you know reflecting radar devices and otherwise give the car some guidance as to where to go is that is that going to happen is that necessary it's not necessary but it will improve things we we just think that infrastructure will take too long it's gonna be too slow and expensive in the order of decades right yeah so we're just gonna preempt it by making software that can just see things as humans do and that's why we have to focus on cameras so because people see things with cameras or well cameras are like human eyes and we'll just have to make do on any road and later I think the most important infrastructure development would actually be for traffic lights because detecting a red light is actually hard for both a person and a car and a dog and yeah they'll just cross right but in the future you could imagine that each intersection will just broadcast its current status remotely and then you don't have to worry about the accuracy of seeing the traffic light well actually that's you know to me that's a thing whose time has come yeah if you look around this country you find very few modern traffic lights modern traffic sensors and signals you know we're gonna have to lay out the box we're gonna have to upgrade all our traffic signals I agree especially in Hawaii and this could be part of it so you know you've got to get to a point where you can hit that on the historical curve where where these municipalities you know find it necessary to put in new new infrastructure that's when you nail them for infrastructure that will help on autonomous cars I actually talked to Caltrans about this very topic and there's Caltrans is the California Transportation Department and they're so interested in this because they're always doing 20-year planning and they know that autonomous vehicles are the future so they want to know now from the companies that are working on it what does it take good for them yeah good for you for talking to them yeah and they'll set the standard for the rest of the country because once you roll this out and who's working on the traffic signals that the same people are working on the cars working and are you working on the traffic signals well I I didn't personally work on the detection of the traffic signals but it's it's the same group of people that are doing everything yeah transportation software engineers yeah yeah it combines a lot of different fields I've learned a lot just as a software engineer about the chemistry of lane lines on the road and the types of lines there are and there's some places where there's bumps instead of lines it's just it's kind of endless but computers have to learn every single one of those things I hope you're making no speakers this is the talk of the future these are the issues that are being resolved and will change your lives we're talking about major disruption in our world of transportation and energy for that matter we'll take a short break you can think about it you know make some notes come back in one minute hi I'm Tim Appichella I'm the host for moving Hawaii forward and the show is dedicated to transportation and traffic issues in Oahu we are all frustrated by sitting in our cars in bumper to bumper traffic and this show is dedicated to talking to with folks that not only we can define the problem but we hopefully can come to the table with some solutions so I invite you to join me every Tuesday at 12 noon and let's move Hawaii forward hello and Aloha my name is Raya Salter and I am the host of power of Hawaii where Hawaii comes together to figure out how we're going to work towards a clean and renewable energy future we have exciting conversations with all kinds of stakeholders everyone who needs to come together to talk about renewable energy be they engineers advocates lawyers utility executives musicians or artists to see how we can come together to make a renewable future Tuesdays at 1 p.m. we'll be back and we're back Cliff Spradler than me he's the software engineer part excellence going to change the world change the country change transportation you'll see remember the name Cliff Spradler I'm telling you now write it down okay so we got it we got a picture let's look at the picture of the van this is the autonomous van so to say right so this is Google's now they've spun off into a sub company called Waymo this is their collaboration with the Chevy Pacifica platform or Chrysler Pacifica platform and you'll see a lot of additions to the normal car right there's this huge thing on top yeah what is that it's a lot of different things and actually the insides of it are a secret and I haven't worked at google so I don't know anything but what they've told us is that there's a laser radar on top of lidar which is different from a radar there is the lidar is the housing difference it bounced back or not it does bounce back it's just it's like a laser that just comes right back okay and so you can get actually a very good 3d view of the whole world and it's actually something that's spinning around in the top continuously but it's safe for human eyes from what I understand I've had one right out your retina yeah there's a little eye warning on them but they say it's okay okay and then there's just a ton of cameras and all of these sensors are designed to handle like the corner cases of this this problem so like when you're making a turn on an intersection what's in your blind spot autonomous vehicles just don't have blind spots yeah that's great because I know I do always try to check my blind spot sometimes I don't even see what's my blind spot and that's the other thing is that one of their advantages is they can look everywhere at the same time yeah right and integrate yeah so but here I mean just just comes to mind I mean you know you can drive in New England and one all of a sudden there'll be a fog yeah you'll roll in from the ocean on a little cat feet in the words of Carl Sandberg okay and you can't see anything humans can't see anything it can't see anything right yeah so what happens well that's why our different kinds of sensors help out because our cameras won't be able to see just like people do but LiDAR and radar can see through fog just fine well depending on how foggy it is right but in that kind of situation you probably don't want to keep driving anyway so the important thing is to always pull over the side of the road in a safe way so that's one of the core aspects of this is to always have an escape plan yeah well you know but I also I also would imagine that there'd be a little voice in there saying you know Cliff do you think you ought to drive now it seems to be foggy past a certain percentage of precipitation why don't you pull over well that that's something that will happen in the new near future but in the far future I would expect there to not be a steering wheel and so what we're thinking about now is what it'll pull over for you yeah well it will pull over for you yeah but then what should it do are you stuck or should it be able to keep driving a little bit forward calls for an uber yeah right well actually uber is one of the huge players in this is that right yeah they spent hundreds of millions of dollars on this good for them it's easy into the future yeah well I think they just want to replace all of their drivers with autonomous vehicles no labor issues well yeah it's the most expensive part of a taxi service you know they're subsidizing the cost of their service heavily now and I think they just want to replace the humans with robots so that instead of having to subsidize it now it just costs that much and just put everyone else out of business it's a perfect idea yeah they'll advance this technology for sure yeah actually the uber cars in Pittsburgh if you if you reserve an uber there's a chance that you'll be picked up by an autonomous vehicle today in Pittsburgh yeah Pittsburgh is ahead of the game Pittsburgh is where Carnegie Mellon is and uber basically bought a team of roboticists and autonomous vehicle people from Carnegie Mellon and opened up shop next door and so that's where they are and that's where somehow Pittsburgh is like the forefront of this okay put me in Pittsburgh yeah I'm in Pittsburgh I'm standing in front of my house I I call an uber the uber drives up what's my experience like um well it's you just get in and it just takes you there backseat yeah backseat backseat there is no front seat no well there's no no okay so there's still a safety driver but all they're doing is watching right oh there is a driver but he's not he's okay we're moving with evolving into no driver so it'll probably take people estimate four to five years to get real approval to have no driver in the car and what the industry has to demonstrate that it's at least as safe as humans yeah right and there's a big conversation about is as good as humans good enough or should it be two times as good as humans three times as good you think it can be better it can be better but what's the minimum it takes before it's allowed to be on the road then this governmental issue well it will definitely improve some things right like it it's definitely gonna be better than drunk driving or driving if you forgot your glasses at home or like let's say I actually have heard of one case in a tesla where somebody was hurt and they had to drive themselves to the hospital and they used the autopilot feature to stay in lane on the highway on the way there and they did and they were just like dazed and confused and saved them and that's just a limited form of autonomy yeah well this is definitely going places yeah so i get in the backseat yeah i tell the uber where i want to go is it listen no no no no you put it the uber destination from the start on the app okay okay right the app you don't have to say a word yeah i don't have to say a word okay then i get in the backseat it knows it's me yeah it pulls out and it just goes i don't have to talk to the safety man no he's almost irrelevant yeah yeah it will he'll disappear soon enough disappearing safety man yeah okay so then it takes me to my destination and like uber now i don't i don't have to worry about money it builds me later my credit card that's it yeah that's it yeah so how do i feel about this i mean do i do i feel confident to the people in pittsburgh feel confident yeah very confident um they but then they always have the safety driver just in case but that's kind of the problem is right now there's a divide between assistance and full autonomy and unless you're a train safety driver it may be so good that you don't realize that you need to take over right in critical situations and so i think there's going to be a big jump where we we we intentionally don't release something that's in between those the assistance and full autonomy because it can be too dangerous yeah complacency that complacency is a real thing after about 15 minutes of perfect driving pretty much anybody will just stop taking it and get the co-confident on it yeah so you know you talked about the sensors and you talked about the car learning everything in all directions and all this and i suppose you know there's other ways other kinds of information that could be fed to the car for example um it's going to rain if the temperature is going to drop below it's going to be icy yeah reports of traffic jams all that stuff that we we know we can get yeah and it feeds into the car and the car now you know does something with its brain we call that routing decisions routing decisions and it makes those just like if you get in and have uh you know one of those what do you call it navigation systems right now or use the one on your phone um now the brain is is putting a lot more data into that and making decisions about well you can see how google is very well poised to make this happen because they already have all that information for just lump it up yeah and google earth my god you get everything yeah yeah so you know but the question is i mean are we are you confident right now the sensors i mean are the sensors ahead of the brain or is the brain ahead of the sensors because it's two separate things well i would say the sensors are ahead of the brain although it's important to know that cameras are not the same as human eyes there's a lot of deficiencies um compared to a human the human vision system but i would say that the problem is the software because if you look at an image from a camera you almost always know what to do right you could probably drive the car based on like a camera in front of the car and drive just as well as you do seeing the right yeah so it's all about the software now and cameras are getting better i mean yeah what do i hear that iphone has 800 people working on the camera all the time i think that's right it's getting better and better and better yeah and so you know it's like you just wait for wait for it to be developed somewhere else and then you know the coat tail on that technique okay so we're talking about the intelligence now yeah how does it work it's more than a case statement it's more than it's more than anything we can imagine right now it's it's it's running headlong into the future of software right what's it's actually very different from any kind of software i've written before because you don't write code you don't write say if this condition then do that you don't do that at all you train it just like you would train a person you have to design what the the structure of the neural network would look like um but then you feed it images like of roads and cars and you tell it at the same time you say this is a car on this road here is the lines and it learns that um by almost by probability it's like this is most likely this so like if you if you have something that's like a stop sign detector right a stop sign is usually red right and so if you're trying to decide is this a stop sign or not a stop sign then this neural network can quickly figure out that the presence of red in the image is a pretty strong indicator right and it builds on this but to a point where we don't even understand it it's just like a human brain we don't understand how it works sort of like space odyssey 2000 how how what are you doing right well it makes bugs really difficult right because oh yeah yes it does something wrong you don't know why and the only thing you can do is train it with counter examples of like here's situations that are similar you should follow this this different situation so you think it's just cool yeah you do so in the same way that kids and people do it has it's not perfect in its way you know when you write software it will always do the thing you tell it to yeah um but the problem with software and and the roads is there's too many situations to account for with software and logic you need something like a human brain to handle all the different right to learn yeah to get to account account for all those situations and and build it in i mean it gets it gets it's pretty big isn't it how big how physically big is the brain uh well so there's been a lot of work done to miniaturize the the amount of computing hardware necessary but the resources are massive um but actually we've been able to pick up piggyback on graphics cards meant for games because they massively distribute the problem amongst hundreds of little processors on a chip um and those work really well for this problem as it turns out especially because you have a lot of images that you gotta recognize exactly yeah right it's just images um but now that that uses a lot of power and for a car power especially an electric car power is a problem yeah so now a lot of work's being done to make that like special dedicated neural network processors that only do this right so you don't waste anything yeah yeah um so you know i mean i'm really wondering um how dependent the the car and the safety of the car is on the computer is there a failsafe here yeah if the whole thing goes down if there's a you know how about a bolt of lightning comes down and interrupts the the electrical system where somebody does it intentionally yeah um there's the car no enough to say what gotta pull over that's it we're finished now so the way this is usually designed is um there's something that's not been the neural network not the ai kind of like a lizard brain lizard brain okay that always primitive yeah that always has a plan of what to do in case the the higher level cognition fails really that's how we designed the system yeah and we also designed it to be electrically redundant um in case there's like a brown out or the lightning strike everything is redundant so um if the the higher level system just stops publishing information you'll always just pull over to the side of the road i love it i love it to say nothing about all the information you're going to get about weather and an ice on the road and who knows what about everything which can be integrated so one minute left okay what is your vision what is this going to happen how is it going to affect us okay so i think the thing that will affect us most is all these people who will no longer have jobs it's over three million truck drivers in america there's many states where the number one job is truck driver and trucking is the easiest thing for autonomous vehicles to do they just have to you don't even have to go to the end points right you can just drive on the highway and that would eliminate a lot of jobs just like just like uber yeah right just like uber so how will society react to um these jobs are going away and not coming back i don't think it's going to be like the industrial revolution or other times when technology has like deprecated old jobs and people have found new ones i don't think that's going to happen this time so are people going to get basic income are they how will we support people and how will they be happy without a job wow wow from from engineer to social engineer yeah we think about this a lot because we realize the impact we're going to have on the world and it seems like it's going to be in that plus but it's it's definitely going to change society fabulous i'm ready okay thank you cliff thank you wonderful to see you and to have this discussion stimulating