AWS DeepRacer TV – Ep 5 New York

Posted by


(drum beats) – [Emcee] AWS, New York Speedway! We’re looking for winners. – [Blaine] These are your babies! – [Martin] Yes. – [Emcee] His own homie beats him! – [Mike] Unless I win this, I’m not getting to the biggest convention. – [Atchima] I’m just going
to give this a shot! (laughs) – [Daniel] Our only car is the Deep Racer. – [Joe] I built this huge track, I’ve got to have something to show for it. – [Emcee] Yeah! – Out of nowhere! – This is an unprecedented finish for us. – [Emcee] Uh oh, what happened? (drum beats) (upbeat music) – [Narrator] Welcome! To the world’s first
autonomous racing league. Developers of every skill level, from all corners of the globe, compete for the ultimate
prize in machine learning. Welcome to DeepRacer TV. (Upbeat music) – New York City. It might be easier to list what it isn’t, than what it is. (rock music) The cultural, financial, and media center of the world. The most populous city in the U.S. The best tech city in the planet, according to a recent study, and home to more job
openings in machine learning than anywhere else. DeepRacer, New York City, on the road to the championship
cup at AWS re:Invent 2019, And the clock starts, now. – Head on over to the paddock, where you guys can sign
up, sign in, compete. And represent New York City! – What do your parameters look like? I cannot show you my reward function, but I can show you the action space. I don’t turn too widely. (techno music) – I’m Martin Paradesi,
and I’m a data engineer. And I work on data pipelines, and streaming data. I was born in India, and I came to America for my masters in machine learning. The field of machine learning has changed and I’ve been trying to stay up to speed with the field. And DeepRacer seems
like a natural evolution of my learning process. Welcome to New York! We have about 10 to 15 people from Capital One. And Nelson is here with me, so I’m hoping that he
does well on the track. All of the machine learning
has stopped last night, now it’s all about human learning. And the best driver on the track will probably win. – You’ve been prepping for this race? – I’ve been prepping for
quite a few months actually. – Okay, wow! – I have tried several
metrics using waypoints. Using a lot of speed and, as long as there is one wheel on the track I’m fine with it. – So, you’ve got more than one model? – I do. Actually, I have a bunch
of them right here. – These are your babies! – Yes. – You’ve got your babies! They’re color coded! – Yep! – Oh my goodness, which ones are which? – I think the bottom ones are waypoints. – That’s awesome! And which one are you racing today? – The speed one. – Just turn it loose, let it run? – Just let it rip. – I like it! All right, good luck! – Thank you. – [Emcee] We have Martin getting on down. He’s knocking it out right here. We’re trying to find the fastest time, once again here at the AWS. DeepRacer! – [Martin] That was good. Can we get it for Martin? (crowd cheers) He is our leader! Our new leader! Look at that smile on his face. Come on man, we can make some more noise! Who potentially can
represent New York City! – I’m thinking, oh my word I’m actually first place right now. And I never expected to be first place, but now that I’ve won it, I want to win it all. – 10.54 seconds. Are you just going to sit
back and let this one ride? Are you going to race again? – No, I’m going to try my other models. – Good luck with that. – Thank you. (upbeat music) – [Emcee] Joe showed me a video of him with his very own track. That he has at his job. (techno drum beats) – I am Joseph Nielsen. I work for Slolom Consulting. I help do their software engineering. I make APIs, search engines. You know, the office nerd, or whatever. – On your mark, get set, go! (electric whirring) – I never really got into
reinforcement learning, until I started playing with DeepRacer. I asked my kids, hiking
or arts and crafts? So here we are. It’s been fun learning about
reinforcement learning. My oldest is supervising
the machine learning. – [Interviewer] How’s it going? – Good. – [Interviewer] Have
we reached convergence? – Um, no. – Going in to the competition, I really got to do well on this race. I built this huge track in the office, on 61st floor of the World Trade. It’s 37 meters long. I didn’t even ask my boss, I just started building it. Keep on going! No, no! (yells) No. Ah! Oh, you (bleep)! I’ve got to have something to show for it. (drum beats) – [Emcee] 20! – Oh! – [Emcee] What position will Joe be in? All right, they’re going to tally up his fastest lap! Oh! Joe! New finisher! Tearing! Center, good job! Joe, I believe you can fly! – [Joe] Two nights ago,
I made a breakthrough. I just need to do speed
squared times progress. – So, what Joe is doing, is
combining existing data points, like position to center,
next waypoint, and speed, into a single value he’s
using to train the DeepRacer. Called variable future waypoints. (upbeat music) It’s like skiing. What you’re going slow, your eyes only need to focus
a few meters in front of you, to adjust to changes in the course. But, if you’re speeding downhill, you need to look farther
ahead to plan your moves. When it’s moving slowly, Joe’s DeepRacer model focuses
a short distance ahead when deciding whether to turn. But at higher speeds, the car looks at a waypoint
farther down the track, and that’s variable future waypoints. (guitar music) – To get on that leaderboard I think we have to stay on
the track the whole time. – [Interviewer] Where’d you get that? – re:Invent 2018. He started coding. I thought models. – Mathematically, if
you stay in the center, it’s going to take more
time to get around the track than if you’re in the left. – That’s right. So, mark the left line
as your center point, that’s not a bad way to do it. – [Daniel] I’m Daniel. I am the Head of Product at Logicworks. We specialize in
providing managed services to AWS customers. This is my son. – [Derek] I’m Derek. And I’m 10. (techno music) – [Daniel] He’s a very good chess player. We program Legos. We make Legos do automated stuff. – [Derek] We both share
an interest in coding. – [Daniel] We wanted to learn python, and when the DeepRacer got announced, I was like “Oh, this is perfect”. We live on the Upper West Side. I bicycle everywhere. – [Derek] And I walk. – [Daniel] Our only car is the DeepRacer. (Laughs) – Yeah, that’s it. – [Emcee] Here we go, we got the Pohls family
trying to get it all down. AWS DeepRacer League. Sponsored by Intel. Right here in New York City. (drum beats) Let’s see where they
come up on that board! Good job! Danny Pohl, with 14. fourth position. (drum beats) – [Blaine] How are you doing? – Hi. – Your name is Nelson? – Yes, it is. (upbeat music) – I’m Nelson Yang. I work for Capital One, and I’m a software engineer. Growing up in a tech household, the main thing that
really got to motivate me, was wanting to understand what my parents were talking
about at the dinner table. I dabble a little bit in
video game development. At Capital One we brought
in a physical track. Martin was really the
driving force in this team. It would be insane if we
were to win the competition. – [Emcee] Joe is in second place, Martin’s on first. We’re about to have
our next racer, Nelson. So, let’s see what happens. Capital One represent big time! – That was fast. – That was fast, I hope
we had a good time. (audience clapping) – Suddenly, I hear things
go wild on the track. – [Emcee] What? What? – There’s a new first place
and I’m like, “Oh, wait. It better not be someone else.” – [Emcee] We have a new runner! – And then I see Nelson, is happy and dancing over there. And I’m excited because he’s my colleague. – [Emcee] So, here we are right now. Martin, life was good to you. Life was good to you at one time. His car has beat you. Where is he? – He’s getting all the limelight. – [Emcee] There, he’s
getting all the interviews. Oh yeah! – We are really good friends at work and he just brought this beef between me and Nelson. – [Emcee] His own homie! Like, if my homie did that, it’s like East Coast, West Coast beef. – Oh, I’m 100% East Coast. – I prefer all genres of music. – [Emcee] Yeah, Nelson, what’s up bro? You got some beef, bro! – [Blaine] Intel creates
the Atom processor, which is used in the DeepRacer. They’re also responsible for OpenVINO, which is the platform used by developers for machine learning. Giving developers the opportunity
to manage their models, in the cloud, and on the car. (techno drum beats) – [Emcee] Race for prizes and glory! We want to see this little boy change! Now, I have a lot of faith in Mike! – [Mike] I’m Mike Pellecchia. I work at a company
called Applied Visions. I’m a senior software engineer there. DeepRacer is probably my first foray into serious machine learning. I’ve always been interested
in machine learning because I like games. Usually playing board games with friends. Put the sand on the mountain. My daughter is turning four in August. I put her to bed around
eight o’clock, I’m exhausted, and then I have to start programming. I put in some time training, and I would love to win it. I would love to go. It’s one of the biggest conventions. Unfortunately for me,
the funds aren’t there, so unless I win this,
I’m not getting there. – [Emcee] Mike! Definitely, is trying to bring the pain. (DeepRacer whirring) (energetic drum beats) – We, we’ll count it. I
thought that was good. – [Blaine] Michael, 11.9 seconds. How’d it feel? – It was a little scary. It started off a little slow. Slower than I would have liked to, but it kind of picked up. – Now, are you going to get
back in line and try again? – For sure, for sure. – All right. – This is your first
time getting it going. – Yeah. – Good. So you’re confident. He is game face. He has no time for nonsense. Look at that hair pin, she did it like a pro! He’s got a very stressed
out smile on his face. You got to believe in yourself some more! What’s up, young man? How you doing, you all right? You got something to say, say it to my face, young man. How you doing? Watch this. She’s going to be amazing
in front of your face. You good? This is my Aunt. I told her to come here. I snuck her in. Why not? Bring your family out, and let them see how great you are! If anyone has a Lambourgini, on the license plate it says Blaine, make sure you guys move it, it’s going to get towed. Bro, I’m done! (bass guitar) Three hours left, so make it all happen. Joe, make us proud, Joe. – I put my model in for training. I set my look-ahead
distance to point-five. – [Emcee] Here we go! – [Joe] Maybe there is hope that I can end up in the top three. – [Emcee] Look at this guy! Boom. – [Joe] Oh, 10.5! And sure enough, I get 10.5 seconds, and get back into third place. – [Emcee] The Pohls family right now. On the track. All right. – [Blaine] The Pohl family
improved their positions. – [Emcee] Yes. – It’s not going to
make them on the podium, but they could be up for a top 10 finish. – [Emcee] Martin, he
needs to redeem himself. – [Martin] I’m just waiting in line, I had only one bathroom break all day. So the conditions were tough. (laughs) – [Emcee] Let’s see if he can
beat his colleague’s time. – [Martin] This time I’m
using a combination model. A combination of waypoints and speed. It hasn’t been overfitted. It’s been trained just right. – Regardless of his run right now, he’s a good person. – There’s less than an hour left. I hope I get another shot at this. – [Emcee] His name is Nelson. He is trying to top himself. – Capital One took one more
swing at the leaderboard. But, Martin and Nelson are
unable to improve their time. – It was a little discouraging. I hope that I can get
one last shot to race. – All right, they’re jumping
back on line right there. The top three have not moved. In a while, so we would definitely like to see a shift in energy right here. – Are you going to change anything before this next run, or are you going
to stay with the same model? – Yeah, so the first model I ran, I was running it rewarding hard on speed. – Okay. – Now this one, is just staying
more true to the tracks. – A perfect line might be a good call. (laughs) Good luck with that! – Thank you very much. – All right. (dramatic drum beats) – So it was tense. I’m watching my car run, – Nice turn. – And I’m not seeing the general wheels, I’m seeing it run pretty smoothly. The crowd was starting to get into it, like I was still, head down,
playing with the tablet, I start looking up, and it was, you know, really ripping around. – [Emcee] He’s good, he’s not playing! – I felt like it was fast. – [Emcee] Let’s see how that looks on that leaderboard right now. Are you ready? (cheering) Oh wow! AVI. Mike! (fast guitar music) AWS, New York Speedway! We’re looking for winners. – [Blaine] Michael seemed
to be sitting pretty. Martin and Nelson from Capital One, and Joe from Slolom, they want to race again, but we’ve just been told
there is no more time left. – Michael’s going to win. And I could sense the
disappointment in Nelson. And I felt really, really bad for Nelson, because he didn’t get a second shot at redeeming himself. – This right here, is our final competitor. She has been flown in from the mean streets of Alabama. (techno drumbeat) – My name is Atchima Klomkaew. And I am going to start working at Google as an engineer resident. This Monday. (laughs) I was born in Thailand, and I moved here when I was 12, to go to school in Alabama. I guess I’m, like, really outgoing and very adventurous. So, I like to have things
on my plate all the time. I went to China for four
years to study abroad, and learn it’s language. I think programming
language is pretty similar to human languages. And then I got back four weeks ago, and I had to move all
of my stuff to New York to be in the same city
as Michael, my boyfriend. He really motivated me, like, “try more. Come to New York”. If he hadn’t told me that, I probably would just
not be here. (laughs) – Three, two, one! Here we go. This is her first time. – I don’t know, I’m just
going to give this a shot. We’ll see what happens. – [Emcee] She is zooming, on by! – On the first lap, I
had like a decent speed. – Three minutes left! (dramatic drumbeats) She’s fast! (crowd cheering) Oh, right behind! Yo! Out of nowhere! – And then, within the
first minute or two.. – [Emcee] Uh oh. What happened? – I think it ran out of battery, and I couldn’t race it afterward. – [Emcee] And that is time! – I was like, oh okay, I guess that’s it. I can’t do any more race. We’ll see what happens. – [Emcee] All right,
she did have great run. (dramatic drum beats) (crowd cheering) – You got it! – Are you serious? You had individuals who had
the whole track in their house! They threw out all their furniture and they put this down
in their living rooms. And their parents hate them! And you just came here out of nowhere! – This is an unprecedented finish for us. One racer to go, a racer who hadn’t raced all day, comes in, places the top time. But that’s how it goes in DeepRacer. (Upbeat music) – [Emcee] Third place. This individual has stood
in the line all day. He brought a lot to the table. Definitely an amazing haircut. Our third place, give it up for Nelson Yang! (crowd cheers) – It was a team effort all in all, I would say. – I mean, this is as
much his as it is mine. – Yeah. – Take it. (both laugh) – The second place winner, give it up for Michael Pellecchia. – Unfortunately, first
place was my trip to Vegas, so I haven’t looked yet
to see if there’s another event on the East Coast I may see if I can make it out to that. Try again there. – But, right now, it’s all
about this wonderful storyline. She came here, from the streets. I was like “Hey, you should enter”. And she’s like, “where at?” At the Javits, why not? And she came on down. Literally, the last individual. Atchima Klomkaew! – [Atchima] I did not
expect this to happen. It actually feels like a dream. I guess this is real,
because it’s really heavy. (laughs) (upbeat music) – So that does it for AWS New York. Out of the blue, the very last second, Atchima comes out of nowhere, takes home the trophy, and she’s going to Las Vegas. But there’s still a chance for you to win your place to
compete there as well. In the virtual circuit, and at the remaining AWS summits.

4 comments

Leave a Reply

Your email address will not be published. Required fields are marked *