Listen to Stop & Go F1 on Spotify - spotify.link/PAEa58BGDyb Follow me on twitter - TheRhysieB?t=Fq93... Follow me on Instagram - therhysieb?igsh... This week on Stop & Go F1 we take a look at the AI Powered Racing Series #f1
@@DanielDorn-tr7tw yeah but without a human to protect, some of the safety concerns probably could be relaxed. Obviously teams want to protect their investments but just think how fast these cars could go without the weight of all the life saving bits like roll cages, halos and the like. Given time to mature, I'd love to watch this alongside humans racing
Actually DARPA put on the first autonomous vehicle race. I helped start Axion racing back in 2003 when the DARPA Grand Challenge was announced. It was a race in the Mojave Desert from Barstow to Primm Nevada. They gave out a 1 million prize for anyone that would have completed the course as well as DARPA contracts to go with. Also my vehicle, Spirit, was number one qualifier at California Speedway for The Big Race. We also, with a 94 Jeep Cherokee and $750,000, were the first autonomous vehicle to climb Pikes Peak in Colorado 8 years before Google did it.
It's not even the first open wheel autonomous race, either. IndyCar has been doing this with university teams with the Indy Autonomous Challenge for a couple years. They're a long way from a proper full race, but watching this, it looks like Indy is a little bit farther ahead than A2RL. They even unveiled the first purpose-built race car chassis for AI cars this year at CES.
@@leprechaunbutreallyjustamidgetGiven the number of times Tesla Autopilot has chosen to steer towards a wall and stomp the accelerator? No, it really wouldn't be better.
Why, do you think it could`ve gain consciousness and evolve rapidly and then decimate anyone on the track or he could`ve traveled back from the future in order to kill Kvyat`s mother?
I remember probably close to 10 years ago now I randomly came across an Autonomous Vehicle Challenge competition in Vegas or something which was broadcast live, where different universities/companies were competing to drive around a road network, with dirt roads, rough roads etc. Every car had essentially a giant roof box of sensors, cameras, and everything needed to make it move by itself and they were all either stationary, rolling an inch before braking and repeating over and over or driving 3km/h. The fact that the technology has reached the point where they can be "raced" on a track is nothing short of amazing, the fact that this has all been compacted down into the size/weight of an F1 car is even more brilliant. This doesnt really detract from the technology, it gives the technology a much larger audience, and it gives universities and companies the ability to realise that maybe it is something they can accomplish better than the current competition. If they werent allowed to race until every bug was worked out it would possibly be another decade, however, letting the wider public see what we can do right now drives interest.
It is really cool from a technological and engineering perspective, but for racing, this is nonsense. I couldn’t think of anything more boring than watching a bunch of robots drive around a track autonomously. If humans aren’t involved then there are no stakes for anyone. The technology is great but it should be used for automating dangerous and monotonous things such as loading/unloading cargo ships or driving freight trucks. Racing is a wasted application for this.
@@michaeletzel4877 I think you’re missing quite a lot of things. If humans aren’t involved it doesn’t change anything; greyhound racing, RC racing, pigeon racing, marble racing and even battle bots. People don’t care about a human in the car, they will watch regardless. Rather autonomous racing focuses on the team itself as they need to not only build the car but create an AI capable of racing, these teams are putting hundreds or even thousands of hours into the car, and they all want to get the win, just like F1, NASCAR, WRC, all have massive teams in the background. And yes autonomous vehicles are best suited to roles like mines, trucking, buses, taxis. But like with all other forms of motorsport, the technology that is developed by these teams will be repackaged and used in a wide range of uses, and racing these in competitions will see this technology develop faster as people have something to look to achieve.
@@LCaddyStudios I’ll agree that autonomous racing will probably find an audience but it will be minuscule compared to the number of people that enjoy (human) F1, just like the audiences of all of those alternatives you mentioned. Perhaps autonomous racing could become its own thing but if human drivers ever disappear entirely I think motorsport as a whole will be much less popular than it currently is.
that is not UAE strategy. they go full steam. probably blowing they host the first F1 AI race ever in the world, and the 2 month notice is probably cause Saudi announce they would the the same 3 month later. Couldn't help but notice how they were so much "F1 car" but going way slower than expected.
A robot may not injure a human being or, through inaction, allow a human being to come to harm. A robot must obey orders given it by human beings except where such orders would conflict with the First Law. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law. A robot must not learn how to drive on a closed circuit unless it crashes into another robot except if it conflicts with the first, second or third rule.
I think perfecting the ai and having them race the best car you could build, that isn’t held back by the limits of g force a driver could handle would be pretty cool
People who think LLMs are a road to machine sentience dont understand two key and basic concepts 1) statistics, bc thats what models like Chatgp are based on, and 2) the point of diminishing returns. Its like putting of 10lbs of mussel when you start training for a year and thinking in 5 more years your going to put on 50 more lbs.
Yeah a LLMs (Large Language Model) is probably not driving those cars, although based on how they did it could be a chat bot driving 😂 It'll be some other deep neural network based model that didn't have enough training data
@@wiredvibe1678 Very simple: we can learn from experiences. LLMs use large data bases to predict what the next words will be based on the previous ones. If you feed it a text that explicitely states that and why a particular sequence of words it used is false, for example because the LLM is unknowingly referencing something like a meme, it will not be able to learn from that at all. It can not contextualize information, it can not learn what sarcasm, references or parodies are. Sure it can repeat definitions for those things, but it can't recognize them and therefore can not distinguish between true and false statements, it doesn't even have a concept of what that difference means or why it might be important. Also if consciousness was as simple as this, we would have figured out consciousness decades ago, so that's how we know that.
@@creativedesignation7880 I'm pretty sure we can contextualize an experience as a statistic, and then "learning from an experience" could be defined as adjusting the output of your behavior based on the input... I don't see how you have shown that we are not just a really complicated statistical model.
OH MAN! 🤣🤣🤣 So many one liners that I cant pick one fav. I like your style so much I subscribed halfway through the video and Ive never seen anything by you before. "To brag in BINARY" had me rolling around the floor. Hilarious. Some true RU-vid Excellence in this video well done you. Abs BRILLIANT!!!
Just stunning and brave watching F1 cars driving 120 MPH slower than than can go. I mean he passes coming out of a corner yet neither car looked like they were accelerating. Just put remote controls and have million dollar RC cars. Would be more enjoyable than whatever this was.
AI engineer and professional driver here - Why do you think that? Because I can tell you that no matter how much you know about racing, 2 months is not enough time to deliver an AI capable of racing. Heck, it’s not even enough time to train a driver capable of racing..or a car.
@@TheOfficialOriginalChad So don't have a race. Call it a demonstration or an exhibition. Like at last summer's Festival of speed. Calling it a race implies certain elements that were clearly lacking here.
@@Syolaar After your demonstration of racing knowledge, engineering competency, and impulse to form an opinion without knowledge of the aforementioned, I think it might be wise to sit down before sharing your event planning and marketing strategies…especially when you realize your definition of “race” is not the same as any dictionary.
@@TheOfficialOriginalChad Was thinking I've made better driving and navigation stacks for my cnc vinyl cutter robot a scalpel on a servo on omni wheel robot could cut vinyl stickers out from gerber which is trickier than it sounds step tracking correction and keeping the scalpel oriented for the cut was a b*ch :) . but that took 18 months to get the model accurate and the bulk of that was with the use of 4 free a100's going 24 hours a day for nearly the whole of that 18 months a while was pretty cool stereoscopic vision and lidar fore and aft with 4 microwave meat detectors and a look down sensor for the table edge.
@@leonmusk1040I have tried and tried and tried to decode your nonsense that’s where I’m drawing the line. I can’t tell if you just forgot a bunch of punctuation or you’re using terms without knowing their meaning or a combination of the two. “My cnc vinyl cutter robot a scalpel servo on Omni wheel robot”… Forget everything I said. All I want to know is: wtf is a meat detector?
I disagree that the project is a "joke", I think it could actually end up being the fastest racing series to ever exist and be incredible to watch but I absolutely agree that they wanted to make it into a joke because eventually the engineering teams would get all the kinks out of the system and the AI, once it figures out how to properly race, would utterly destroy any human driver and seriously degrade the F1 racing scene if not completely kill it.
The project in itself is no joke, because if people start taking it seriously there would be many improvements in self driving, also there will no longer be a g force limiting factor anymore and the cars can be pushed to absolute limit until tyres complain. But it's the fia that's the problem and also software wise, don't you think racing games that are 20 years old have better racing AI than these cars?
to be fair, atleast its interesting this way. if the ai was good there would be no point to the series. its just watching ai drive around the track at decent speed, probably not overtaking.
@@brettjames5061 well I imagine if they were to make it an actual sport (if you can even call ai racing a sport, more an engineering challenge I guess) that it would be a little more than just a straight forward normal F1 race, I imagine things would be added in that would never be aloud if human drivers were on the track, idk what exactly but there’s no way it would just be a normal race cause as you said, that would get very boring very quickly
YOOOOO, that fireworks show was the BEST that I’ve EVER seen on video. Can’t imagine how awesome it must have been in person. The “race” was a joke, but that display was absolutely amazing! 😎
That was funny, embarrassing, and impressive all in one go. 2 Months to program a car to do that when you consider all the lights, extra lanes, and everything else that's going on around the track and cars actaully made a full lap, that is impressive. Thanks for showing this, keep up the good work. I would imagine here will be a good leap forward for the next race, the teams will have gathered a lot of information and feedback from this race, which they will them use to improve the control of the cars. I wonder how well a Tesla would do around one of these circuits?
Honestly, as hilarious that whole debacle was, it's still kind of impressive they managed to build something that can finish a lap and even overtake with that kind of time crunch. Definitely have the potential, but suffering the same disease of overpromise underdeliver.
@@03056932 well that's why i mentioned the overpromise and underdeliver. I'm not well-versed in either side of the automation but i can only assume that they both have unique challenges in which the autonomous racing scene has not yet matured as opposed to the robot fighting scene. At least, i believe within the novelty and time crunch of the scene, it's still quite a feat.
@@LRM12o8 I mean the difference is one is a robotic driver, the other isn't. I think what killed the fun is knowing that it's an AI driving the car even when most of what i believe the entertainment on the race track is the car go whoosh part of it, and AI "kills" the vibe. But ultimately, i still feel the entertainment part of it is still seeing the cars go racing. Or maybe i'm just overselling the idea to the wrong audience.
90% of cars on roads do not, they just commute from one point to other. And they are huuuuuumans, they do stupidiest errors, make huge pileup in 100% safe conditions, squash the whole school class when trying to reach water bottle (yes, Greenacres school) and so on. So theoretically autonomous cars have great advantage in everything - speed, traffic dencity, safety, independency of human stupidity, booze and drugs, bad mood and fatigue. If my van would have ability to switch to autonomous drive on highways and other main roads it would be great advantage, saving me lot of time, money and risks. Drive Sydney-Adelaide, 1400km, takes 24h instead of 15 and incures significant risks as I can not be 100% vigilant without any errors and distractions. And kangaroo problem easily solved with proper bullbar. The problem is that our beloved government 110% will make cars not autonomous, but government-controlled. Do not go there, do not accelerate here, your carbon emission in this month exceeded so no ride at all, you are obliged to give a ride to this vulnerable illegal immigrant and so on.
@@antontsau So why not just have trams or trains? We don't need a load of 'AI' powered machines from rival manufacturers all sharing the same space on a busy road.
@@SilverfoxJB because tram or train can not travel from point to point, only from station to station, and can be private space which is always near user, ready for his personal use. Car - can. 30 km on highway train can do even better,, 5 km in suburb and 5 km in city - no way at all. Including public transport, not only cars. See Adelaide OBahn, "rails for buses", stoneage version of such a self-driving car. The same bus, without any transfer, crawls on city and suburb streets, but 15 km in the middle it flies 90 km/h with driver do not touching steering will at all. Yes, its suitable for buses only, but if make similar electronic control on dedicated highway lanes cars could do the same.
I kinda like the idea of a coding battle, but that's because I'm a software developer. It doesn't have to be racing, but racing offers interesting problems of computer vision and physics in a relatively controlled environment, so I'd probably watch this more than, say, a robot football match.
@fabioguggeri325 I agree. I must admit, considering they only had two months to do this, they did a good job. It was a massive mistake to make this a public event though. They just shot themselves in the foot when it comes to events of this kind and scale.
Competition accelerates development. A Motorrace is a pretty good controlled environment to test real life application of software and robotics. Every team having a different AI is like have different drivers, it is not like everyone has the same code or problem solutions, so if this is going to be fleshed out more, it can get really interesting.
It's a technology race, just like F1. The difference is that they are developing the racing AI, where in F1 they're developing the car and the technology around that. Same with Formula E, developing battery and other EV tech.
It's almost as if the technology is still half baked but everyone's desperately trying to push this as the next "big thing" in order to secure more VC to keep their money pit research companies afloat.
2-3 months is criminally not enough to produce a half-decent product, and the fact that they even managed to do this says a lot about the kind of talent available that wasn't allowed to show itself. The code for something like this would be tens or even hundreds of thousands of lines long, and even writing something like that is an immense task in itself, not to mention debugging it and going through integration hell to put it into the car. Here I was shitting on it myself until I heard they did this in 2 months, then my opinion completely flipped. 2 minutes round a track that F1 cars do in 1:25 isn't too far off, especially in slower cars. Hopefully the organisers hold it again properly in a year, this time giving the teams EVERY SINGLE SECOND THEY CAN to make sure that it's actually a good event to watch.
You can do 2 minutes in this track using a 100hp hatchback. Theres nothing impressive about it. Specially when 2 minutes was just the best qualifying lap, and there were others lagging 8 minutes behind
If they can't get a handful of cars going the same direction around a track, how could we possibly let autonomous cars loose in our cities with everything that goes on day to day, it'll be a blood bath.
Self driving technology has been around for some years now, so this is just building on something that already existed. Here's a video of a Roborace car from 6 years ago. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-QtVbch-02Fs.html This farce of a 'race' shows that this self driving racing is as crap as I expected it to be. Utterly pointless.
@GraveUypo if you can get me a video of someone doing Yas Marina in 2 minutes in a 100hp hatchback I'll edit my comment and add whatever you want me to say to the end of it
@eroffroad5438 because self driving technology today isn't done in 2 months, and has to deal with cars that are either travelling in roughly the same direction at roughly the same speeds in predefined areas (which makes the code a lot more simple) or going at less than a quarter of the speed, giving the computer way more time to make an accurate decision. And again, a decade of work compared to 2 months.
I don't know if anyone has mentioned it in the comments yet but this is the petrol version of Roborace. That started off as a bit of a joke, but it actually started to show some real potential and at times was exciting to watch. Just a shame that it run out of money and went under. - Scott
@@M3rVsT4H I think the rate AI is progressing is both impressive and scary. We’re still a long way from a true sentient digital consciousness yet. But this Autonomous Motor Sport currently, makes it look like we’re going backwards lol.
In 2004 the DAARPA Grand Challenge was created as a contest for autonomous vehicles to race across 132 miles the desert - not a single vehicle finished, and many crashed or simply shut down in the first few minutes. One year later the Stanford team completed the race in 6 hours, and every other team finished. That was 2005... Take home message : these cars have terrible AI, but that will change very quickly.
It’s actually great for engineering.. help with autonomous véhicule to write the best program within f1 rules to win a races given many race circumstances.. was pretty fun
This was pretty hilarious, but admittedly pretty cool as a proof of concept. All of the software stuff can be worked out over time and if it is, this could actually become something pretty interesting to watch in the near future.
Compare the results of the 2004 DARPA Challenge with the results of the 2005 edition. It is not easy to guess how slowly or how fast things will improve.
Or compare Tesla's FSD from 2023 with Tesla's FSD now in 2024. Years of slow-and-steady improvements in AI driving, and then suddenly with 12.3.x, a huge leap forward.
hahaha this is exactly how i felt watching this live and trying to explain how sad it was to my friends. thank you for capturing the experience perfectly
I think it has promise. Though not in lower formula cars, but in purpose built machines that can go well in excess of f1 cars. I think the idea of cars going in excess of 300 or 400 mph ( or more) around well known circuits might be very appealing to viewers and sponsors alike. Thus also teams willing to make the monsters in the first place.
@@tiberazur3 just free styling here. What if we ousted the personal and had robots to clean the track and what not ( how hard could it be.😆 )Or maybe it could be like old Leman and if a car crashes or fails just leave it there and have the other cars have to navigate it. As for fans saftey you could always move them back from alongside the track. Or I get the feeling a barrier could be engineered to contain app debris or cars from entering the stands.
I always imagined the next step in racing was to make cheaper disposable cars, relax the safety rules and have them driven remotely, then its still human versus humans but they can take insane risks and with no driver the cars could be made much cheaper. Watching robots race robots is pointless, like trying to play a computer game that plays itself and ignores your inputs, it's not sport.
From a technical POV this is just amazing. I imagine race cars without ALL the FIA's safety rules just blazing at incredible speeds around race tracks. Imagine the spectacular crashes, lol. Cars with spikes and banana's to eliminate the competitor you say ? lol, so much could be done.
They should flat out run races like this with Civics or Corollas. The slowest current models. The cars could then actually be holding at the edge in the turns vs going 1/8th their actual capabilities.
2 months in not fair timeline. In only 2 months AI could finish the whole lap without crashing is an achievement. Give it a year and train the model properly and see the result. How AI could make an overtake off the racing line when it isn’t even trained for that.
Exactly. These teams had 2 months to develop a code they could only properly run for the first time during the event weekend, having to perform tasks they had no way to train the code for in advance. Just the fact they managed to get laps done is impressive, let alone the fastest laps being at the 2 minutes mark. People keep throwing sh1ts at the participants only because they have zero idea what it takes to achieve some meaningful result here, but I'm impressed and looking forward to much improved round 2, as the teams now finally have some relevant data and experience to improve their AI drivers.
Good question. How do you train an AI to take the risk of a ‘fatal’ crash vs not bothering to go fast at all and adapt to different track conditions and other drivers. Extremely challenging
The race might have looked hilarious, but it still is an amazing feat of engineering and the only way to go is up. I'm excited to see what happens next.
I am no AI hype believer at all, but I find the experiment interesting. Something I would love seeing was if it could be possible to have real drivers successfully controlling these cars remotely on rigs or something like that.. that could be fun.
I'm not here to pee on anyone's techno-bonfire or spoil the fun, but it seems to me that, however good AI racing gets, the thing it will always lack is passion. A human driver CARES about his task. He's struggled up the motor racing ladder to get where he longed to be as a kid, and sitting behind the wheel really, really matters to him. A car can be programmed, but it can't be programmed to care with any kind of authenticity. Red mist, anyone? Revenge? Grudges? Personality conflicts? Ego? Pride? Motorsport is all about passion, and you can't fake passion. Or, at least, that's what my ex-boyfriend believed. 🤭
@foersterjunior but you, the audience, don't get to see any of the drama and struggles the programmers went through to make their faceless bot race, so why do you think the audience will care?
by far the most exciting F1 race ive ever seen period, way more exciting than the entire last decade of F1, if this is what it takes to make F1 exciting im 100% for it!!!!
Oh, that last lap and an overtake at turn 5 is so funny! LOL And that footage just frame-in-frame repeeating Max vs Lewis 2021 Abu Dhabi GP 🤣🤣🤣🤣🤣 Nevertheless, it was pretty interesting. Thanks for the video!
In all honesty, what they accomplished is impressive. The issue is they hyped it up too much. Its nit ready for a stage like this. Now, it will take even longer because it will probably get pulled from existance. Lol, as soon as I typed this out, you said the same thing in the video 😂. Cheers mate
I also watched this and felt that the cars/technology just wasn't ready for prime time. I do think that they will improve as time goes on an the AI gains experience.
They can’t even make this work on a closed track. Meanwhile, people are doom scrolling on their iPhones while their fully autonomous Tesla rolls around on public roads!!!
This isn't AI and that's the big lie. This is just basic script. There's no AI here. A true intelligence will recognize a goal for their needs/desire as a motivation and will act sapient as a result. As it stands, these AI Engineers are just bad script writers who couldn't code a Mod on Warcraft 3 using 2005 mapping tools.
That is a good experiment! I see no point even developing it further! Who would watch a race by computers? Who would date a AI girl? AI is a great helper and this is where it needs to stop.
Was a first ever and was a test. What did you expect? Some amazing raceing? The technology is amazing to even get a race car like that around the track. If it is so easy then why hasn't it been done before. Has a Long way to go but for the first time out they would have lurned a lot.
What did we expect? Well, not this. I would have expected this years ago. Nowadays we are used to reading about self-driving cars doing mostly ok in traffic, so they should be able to get around a racetrack easily. Maybe Tesla should've just sent one of their standard models, adapted to know the rules of racing. I think it would have won by a mile.
@@karlbucannon youre right. Its much more complex on roads dealing with traffic patters, other cars, different lanes, people not obeying traffic patterns, speed limits, weather, etc . But we've been able to do that for years now.
I've seen a self-driving car suddenly stop on the highway and very slowly try to navigate and swerve around between the highway reflectors on the lane lines. Just like the tow truck being driven by a human: humans can creatively figure things out (like which lane lines to pay attention to in a road construction area). AI can't do that kind of problem-solving.