@@endiii27 They may have autonomous drone swarms already, but I doubt they are as fast and agile as the new system here, as this is cutting edge research.
I think it's important to keep in mind that the drones were compared to world champions, not your average human operating a drone :D so the vision based drone is already insanely good
Yeah, I wish I could take a look at the AI Architecture. You see, Artificial Intelligence works best where Humans are involved in the least. It takes a lot of computing power, which most schools do not have, but the Best AI just needs DATA, and a library of known regressive curves, and let the Artificial Intelligence create its own Algorithms, to set up Predictive Modeling which it then can refine by LEARNING. I think that an AI System that can't do as well as Humans must have Human Contamination, that the only LEARNING comes from Human Students getting in to tweak their own algorithms, meaning that it really isn't AI, is it? Real AI is just setting up the Architecture and then hitting RUN. It's got to be Data Rich at the beginning to save time later, during the Learning Phase, but the AI should be able to optimize itself. Really, the entire point of AI is to eliminate Human Beings. So if Students are there, then it isn't Artificial Intelligence, no, it's just a very complex and fallible human designed Algorithm full of the same Errors that Human Beings always drag along behind their Monkey Azzes.
@@Yusuf-ge5gq good morning Yusuf. Whoever created the term AI was NOT THINKING. It was probably some Computer Science Idiot and not a Philosopher. You see, ALL Intelligence is inherently ARTIFICIAL. Only Stupidity is Natural. But Intelligence can be either Biological or Machine. Do you understand what I am saying? I am saying that I refuse to be Stupid because everybody else doesn't seem to be fighting this NONSENSE. Also, "Machine Intelligence" SOUNDS better. Are you convinced yet?
It's a program to help AI drones detect targets faster. Yes, sold as rescue targets but I think we can all expect the military complex is already stealing this data.
Every time a researcher doesn't know how to justify their research - "let's say you need to search for survivors in a collapsed building..." :) Jokes aside- really interesting video! i've been following this group's work for several years and was always waiting for such a challenge!
Yeah indeed, I'm pretty sure that this technology will be bought by the military to develop kamikaze drones that will fly at +100 km/h and avoid any obstacle automatically until they reach the target!
This is great. I read the university's press release a few weeks ago, but your video does a fantastic job of covering the whole event and the details of what was accomplished. Thanks for sharing!
Very interesting video. I assumed part of the data collection from the neuroscientists was about learning how you think about maneuvering your quads in races? Obviously, there's a long way to go in stand-alone vision drones - not just being able to map what's ahead in 3d but understanding the materials, risks of 1 route vs the other... which is a colossal amount of computer task to stick on top of a quad right now, but it's a very interesting prospect. As they continue on their racing theme, I think it would be a good exercise for them to start flying multiple vision drones at once and start thinking about making decisions about when to back off and when to go for it... as well as being able to correct their lines.
You're 100% correct, there is a long long way to go for this, glad to see someone recognises this 👍. Realistically it's only the first baby step. In fact the university didn't even really understand drone racing.
I was so looking forward to this and you didn't disappoint, I only wish it were a longer episode! Thank you for sharing the experience and fantastically documented Paul. It was great to see you with Alex too! I'm sure I said it to you some years back and I'm 100% by no means the first to say, but we sooo need to see the 2 of you with Minchan. Best wishes from the UK 🇬🇧 👊🏼
@@MrE. Thank you so much for giving me the heads up, very much appreciated. Looks like it was done using multiple accounts with slight variation in the number of spaces in the name. Anyway I'm pretty sure I managed to block all those accounts for now
Fantastic video, guys! Thanks!!! And amazing flying there, Thomas! 😃 But yeah, they're training SkyNet earlier than I thought. 😬 Anyway, stay safe there with your family! 🖖😊
Fantastic video! Having been there on the day and seeing the races was just incredibly impressive! Hopefully, this competition between humans and ML will continue to push both sides and ultimately help in real world applications.
These drones are going to carry a grenades to kill targeted people. They need to work out that ability to race side-by-side, because they are going to be sent out 12x at a time until the target is eliminated.
very interesting, and well documented video. What's most impressive to me is that it's even possible to steer the drone vision based and at that speeds. Well done pilots and engineers
They probably know where they are from the gyroscope data, but use the ai vision to recognize more precise where they are and use that to correct there pre programmed route.
THAT IS AWESOME.. After flying 6 years FPV who would of thought that they would be AI computer controlled to do corkscrews tighter than I've ever seen.. That was crazy and awesome.. I see a military contract all day long.. Good luck and thank you for showing us this.. This could have easily been considered classified..!!
This was never anything but military. I notice their "rescue" bots never bother carrying any search & rescue equipment through the course. All they need is a few grams of nerve agent and crash into somebody's face, mission accomplished...
Impressive. We are underestimating the ability of machines to learn from humans and share their knowledge. We are still missing critical infrastruttures to allow knowledge transfer at high speed. Love your content mate, a Fpv drone pilot and robotics engineer
the fact that automating movements in 4 d space, unlike car/boat automation in real life which is exceptionally tough to get it precise, and these makers made an automated complex flightpath at great speed with great accuracy is incredible. I wish to be part of this kind of team someday and i will make a competition for this amazing auto/AI drone
Fascinating the thought of this technology once it has matured being able to get an AI drone with race pilot level flying into the hands of first responders is going to have massive impact in saving lives in dangerous situations
This is such a cool challenge, technically. We should have no illusions about the fact that the search and rescue use case is a niche scenario next to the military application as fast, hard to hit scout and assassin drones though. :-(
I saw the other day what I can only describe as a flock of drones all flying in like a formation almost. They were not lit up or have lights really. But they looked to move together almost. Like 1 controlled them all sorta. Hard to explain. Super cool to see. Only lasted about 2 min they fly away fairly quickly. They've very small too. Maybe even smaller than ones in this video.
I would bet that the so-called "vision-only" drone in reality uses vision and gyro and accelerometer data, paired with a database containing the exact dimensions and geometry of that particular track. By using gyro and accelerometer data, the drone can calculate in real time exactly where it is in respect to its starting location. If it also knows the track's dimensions and geometry, it can combine that information to navigate the course. Inertial navigation. I don't think the vision drone is flying on video alone. Still, I am thoroughly impressed at how fast the ai drone was. Thank you so much for sharing this, I had a blast watching it. Best regards from Brasilia, Brazil.
Awesome video! I'd like to know how much of the processing capacity the autonomous drone was using. In other words, does it need more processing power or just more AI learning and optimization?
I feel as though the AI needs optimisation. As they said, it can’t correct for anything it’s not programmed to correct for. But then that raises the question about whether or not it has enough processing power to account for these new optimisations on top of what it’s already doing.
The AI processing for these will be pretty intensive. It'll have to be doing pretty heavy inferences at a fairly high frequency to give the drone enough reaction time at those speeds. The tests I did a few years back on a fairly new iphone for the time, doing image recognition on a comparatively slow loop, would chew through the battery and heat it up heaps. This won't be done on your standard flight control processor, they have additional hardware for the inferences.
@@droneys1348 training an AGI (Artificial General Intelligence) which allows human like general learning isn't going to be feasible currently. But sounds like they have a bit more training of the AI model that could be done to make it more robust. It'll still have to be limited to generalisations that it can learn from their dataset though and won't handle truely unique situations it's not familiar with.
@@BMSWEB Today you guys are absolute heroes but in a Skynet scenario you guys are going to be looked at as the absolute worst 😂 BMS Drone death swarms 😳 I couldn’t help but keep imagining that thing chasing me! I mean the thought is extremely brief but on the upside at least it’ll be quick 😂
Love how excited and positive you kids are about creating attack drones that can chase down a human easily and avoid multiple obstacles....dont think about it just geek-out and enjoy freedom while you can.
exactly, maybe they need to look at an episode of black mirror. That's the whole issue with ...they only ask them self, can we do it...not should we be building this...
I feel like some future version of this tech could bring ghost-runs from velocitydrone into real life, like an FPV race AI trainer/opponent with different difficulty levels and both pro and famous profiles. You could even have a mode with a specific laptime that you wanna beat or a nemesis mode where it keeps getting faster and faster. That way you could get real life race training even when you don't have anyone to race.
Well done Thomas nice narration and story line it felt like watching a documentary. I'm sure its just a matter of time before the logic in the AI drones is improved to do more than just try to get through the course at the fastest time then you may have to watch out. BTW what a great experience for you !!!
One thing that struck me watching you and Alex fly from 3rd person is how wide your lines were while doing looping elements (conceding that you might not have been flying at 100%). When you watch the AI drone (when it works), the loops are super tight and compact. As a computer programmer and pilot, this was all very interesting. One thing that is very clear is that it might not be this year or even next year, but I'm pretty sure that in the near future you and Alex are going to get absolutely smoked by their machines. Scary stuff!
I'm reminded of when computers were first pitted against grandmaster chess players. At first, the humans beat them easily. Then, it was a good match. Now, humans are at a disadvantage. The things that make us able to perform impressively also turn out to be our downfall. AI, once sufficiently trained, doesn't suffer from fatigue or from forgetting things. So in the same way, the things that make AI perform impressively also turn out to be its downfall. It's a delicious irony, to me. But in all cases, the fact that we are able to train machines to perform on par with humans is itself a very impressive accomplishment. Let's hope we continue to improve how we do this so as not to put ourselves in unnecessary danger/risk. Great video. Thanks for sharing!
@@BMSWEB I've formed that opinion after doing some AI research - it's insane what the next couple of decades will hold. But - it'll only happen if people focus on making it happen... which as you say, may be a while off
Very cool and a great job by everyone in the team! Yet this is what the horror short film “Slaughterbots” warned us about... I don't think that this technology will only be used for "search and rescue"
wonder if the wake of a leading drone if throwing off the AI. large thrust column affects distance traveled per input. does it fly like 10 ms full throttle then turn... or does it measure movement via accelerometer and gyro? and the AI is running on a computer and interfacing with a Tracer TX?
The vicon system is a normal drone where all the processing and controls is managed on a ground based computer, while the vision drone does all the processing onboard!
I predict us Human Pilots will be around for many years to come. It will be a while before they will come close to flying in an uncontrolled environment
When the pilots were wearing the DJI Goggles, they were racing the Vision Drone. They were flying the drone provided by the University. However when racing the Vicon drone, they used their own goggles and drones
THIS IS TO ALL THE PRO PILOTS!!!!! vanover and such : have you guys discussed with the artificial intelligence designers about how you can recover the drone after it impacts the gate or an obstacle???? I think that that could come in very handy. Especially in the field when they actually apply this technology. Maybe give them a little jumpstart so that they don’t have to crash in the stuff so much. What I mean by this is try to convey the technique of saving a drone after an impact. If you have ever seen Mr. Steel fly in person you know what i mean. I have seen him save his drone from the most certain of impacts. When I watched him the first time I was completely amazed how he could just bounce off things and whatever technique he was doing he was able to keep the drone in the air and save it. That is what I mean if you could help them with that. Just a thought thanks for the video amazing 🤩 !!!!!!!
Great vid and interesting experiment - but I must say that with all of the limitations cited, this didn't feel like "AI" or even autonomously controlled flight as much as simply executing a pre-programmed flight path...
This is so cool. It's kinda scary to think about Vanover/Thomas tier AI droneswarms on a search and destroy mission.
*rescue ;)
@@rambozo_fpv176 trust me the military will be in on it before anyone else. so destroy it is.
@@Ezio470 China got the tek down
@@endiii27 They may have autonomous drone swarms already, but I doubt they are as fast and agile as the new system here, as this is cutting edge research.
Kinda? You mean absolutely bloody terrifying.
I think it's important to keep in mind that the drones were compared to world champions, not your average human operating a drone :D so the vision based drone is already insanely good
si señor tiene mucha razon esos vuelos que muestran no lo hace un piloto comun ,,el dron fue superior en realidad ,
Yeah, I wish I could take a look at the AI Architecture. You see, Artificial Intelligence works best where Humans are involved in the least. It takes a lot of computing power, which most schools do not have, but the Best AI just needs DATA, and a library of known regressive curves, and let the Artificial Intelligence create its own Algorithms, to set up Predictive Modeling which it then can refine by LEARNING. I think that an AI System that can't do as well as Humans must have Human Contamination, that the only LEARNING comes from Human Students getting in to tweak their own algorithms, meaning that it really isn't AI, is it?
Real AI is just setting up the Architecture and then hitting RUN. It's got to be Data Rich at the beginning to save time later, during the Learning Phase, but the AI should be able to optimize itself.
Really, the entire point of AI is to eliminate Human Beings. So if Students are there, then it isn't Artificial Intelligence, no, it's just a very complex and fallible human designed Algorithm full of the same Errors that Human Beings always drag along behind their Monkey Azzes.
@@leovolont No, this is an AI
@@Yusuf-ge5gq good morning Yusuf. Whoever created the term AI was NOT THINKING. It was probably some Computer Science Idiot and not a Philosopher. You see, ALL Intelligence is inherently ARTIFICIAL. Only Stupidity is Natural. But Intelligence can be either Biological or Machine. Do you understand what I am saying? I am saying that I refuse to be Stupid because everybody else doesn't seem to be fighting this NONSENSE. Also, "Machine Intelligence" SOUNDS better. Are you convinced yet?
Their ability to adapt quickly outside their AI algorithms is still an issue. Throw more ever changing variables you can overwhelm them.
This is so impressive. Thomas (and Alex) you must be very proud to have assisted in this project. Looks like you had a lot of fun too. :)
Thanks Graeme 😊
It's a program to help AI drones detect targets faster.
Yes, sold as rescue targets but I think we can all expect the military complex is already stealing this data.
Thanks Graeme!! It was awesome to be there in person!
Every time a researcher doesn't know how to justify their research - "let's say you need to search for survivors in a collapsed building..." :)
Jokes aside- really interesting video! i've been following this group's work for several years and was always waiting for such a challenge!
Yeah indeed, I'm pretty sure that this technology will be bought by the military to develop kamikaze drones that will fly at +100 km/h and avoid any obstacle automatically until they reach the target!
...to find last survivors..
This is great. I read the university's press release a few weeks ago, but your video does a fantastic job of covering the whole event and the details of what was accomplished. Thanks for sharing!
Thank you 😊 for watching
Very interesting video. I assumed part of the data collection from the neuroscientists was about learning how you think about maneuvering your quads in races? Obviously, there's a long way to go in stand-alone vision drones - not just being able to map what's ahead in 3d but understanding the materials, risks of 1 route vs the other... which is a colossal amount of computer task to stick on top of a quad right now, but it's a very interesting prospect. As they continue on their racing theme, I think it would be a good exercise for them to start flying multiple vision drones at once and start thinking about making decisions about when to back off and when to go for it... as well as being able to correct their lines.
You're 100% correct, there is a long long way to go for this, glad to see someone recognises this 👍. Realistically it's only the first baby step. In fact the university didn't even really understand drone racing.
Fantastic video Thomas! Cinematography, editing, writing, it was just wonderful and I’ve been looking forward to this one.
Thanks so much buddy! Miss Ya heaps! - Paul
Thanks so much Gil!! I'm glad you enjoyed it 🤩
blows my mind to think i've been following the channel for 8 years now! what an epic journey.
Yep, who would have thought. I guess you never know where life takes you. Thank you
I was so looking forward to this and you didn't disappoint, I only wish it were a longer episode!
Thank you for sharing the experience and fantastically documented Paul.
It was great to see you with Alex too! I'm sure I said it to you some years back and I'm 100% by no means the first to say, but we sooo need to see the 2 of you with Minchan.
Best wishes from the UK 🇬🇧 👊🏼
The same offer was open to Minchan but he declined. We've got some more content on this, so may be able to show some more if people are interested.
Glad you enjoyed it! Thanks for watching :D Ha ha that would be a fun team to fly with 😍
Thank you for taking us on this journey with you, just amazing all round!
Thank you
@@BMSWEB Spammers replying to comments on here masqurading as you, see below.
@@MrE. Thank you so much for giving me the heads up, very much appreciated. Looks like it was done using multiple accounts with slight variation in the number of spaces in the name. Anyway I'm pretty sure I managed to block all those accounts for now
@@BMSWEB No problem!
@@BMSWEB is this" dji mavic 3 win" part of the spammers issue?
Amazing, been waiting for this! Thanks for sharing, I'm gonna get the popcorn and enjoy every last second :D
Thank you 😊
Fantastic video, guys! Thanks!!! And amazing flying there, Thomas! 😃
But yeah, they're training SkyNet earlier than I thought. 😬
Anyway, stay safe there with your family! 🖖😊
Thank you buddy, stay safe too
Always thanks for watching MC! Tech is moving quick 😅
Fantastic video! Having been there on the day and seeing the races was just incredibly impressive! Hopefully, this competition between humans and ML will continue to push both sides and ultimately help in real world applications.
And hopefully competition will remain peaceful. Otherwise I for one welcome our new AI drone overlords.
These drones are going to carry a grenades to kill targeted people. They need to work out that ability to race side-by-side, because they are going to be sent out 12x at a time until the target is eliminated.
🙏🏼 Wow! Thank you for sharing more of this amazing experience.
Very interested to see how things progress in this space.
Brilliant edit too! 👏
Thank you 😊for watching David
Thanks David!!
Amazing to see where this sport has and is taking you guys ! What an experience!! Super job
very interesting, and well documented video. What's most impressive to me is that it's even possible to steer the drone vision based and at that speeds. Well done pilots and engineers
They probably know where they are from the gyroscope data, but use the ai vision to recognize more precise where they are and use that to correct there pre programmed route.
That is fantastic. I can't wait to see this again in a year or two.
It's going to be fascinating to see where this all goes hey
I remember watching you run into the field after school to review a new plane with your father. I have enjoyed watching you grow in your abilities.
Thanks Ryan!! Honestly still kind of mind blown where this jouney has taken us!
Thanks Buddy, it's amazing how far things have come hey. So good hearing from followers from way back
THAT IS AWESOME.. After flying 6 years FPV who would of thought that they would be AI computer controlled to do corkscrews tighter than I've ever seen.. That was crazy and awesome.. I see a military contract all day long.. Good luck and thank you for showing us this.. This could have easily been considered classified..!!
This was never anything but military. I notice their "rescue" bots never bother carrying any search & rescue equipment through the course. All they need is a few grams of nerve agent and crash into somebody's face, mission accomplished...
Impressive. We are underestimating the ability of machines to learn from humans and share their knowledge. We are still missing critical infrastruttures to allow knowledge transfer at high speed.
Love your content mate, a Fpv drone pilot and robotics engineer
Thank you 😊
I love how confidence pours out of Alex. Good stuff 👍
Alex is a pretty cool character for sure!!
This is so awesome, love to see the frontier being pushed
Great work! besides racing i see big potential for "scripted" cinematic flights
This is right down my alley, thank you for producing this
Thank you 😊
It was awesome to see you guys flying in Zürich! Incredible performance from everyone!
Thanks Thomas! You and Vanover are incredible!
It's pretty cool how well those to get along!!
@@BMSWEB like minded people. That's why this hobby is the greatest!
That was an amazing video... So much interesting information
Awesome stuff. Love how fpv is evolving into something greater!
Thank you 😊
Glad to see someone take on this game I once visualized.
the fact that automating movements in 4 d space, unlike car/boat automation in real life which is exceptionally tough to get it precise, and these makers made an automated complex flightpath at great speed with great accuracy is incredible. I wish to be part of this kind of team someday and i will make a competition for this amazing auto/AI drone
So happy that fpv has been getting so much love past year
That was very cool guys! Christian is such a nice guy too, I've flown with him and Elia in Zurich. Amazing to see you all doing such nifty stuff!
So Cool!!! Zurich is an amazing place that's for sure
Wow I had no idea you knew those guys!! Thats awesome. Glad you enjoyed it! See you at Australian Drone Nationals :D
@@BMSThomas @bmsweb Yes lots of catching up to do this weekend!
This is a great video. Solid state LIDAR could really help real-time navigation.
sure ? do you think they just totally forgot about that ?
@@mm-hl7gh It's bleeding edge so expensive to get modules small and light enough for drone use.
Oh DAMN! SKYNET LIVES! THE HORROR!
That onboard vision drone is a thing of science fiction lore come to life.
It's pretty crazy how quickly tech is evolving hey? I wonder what the tech industry will look like in another year!
"... gives us more time to do bigger and better things ..." yeah , like waiting in the line to get something to eat.
Just commenting to support this awesome channel!
Thank you for the support
Impressive video Tom - I can tell this took a lot of effort.
And so damn interesting! Thanks for sharing!
Thank you 😊
Thanks dude!!
Fascinating the thought of this technology once it has matured being able to get an AI drone with race pilot level flying into the hands of first responders is going to have massive impact in saving lives in dangerous situations
really like this format - nice Vid
Thanks!
Incredible a person can fly like that!
What a great video! You guys killed it
Thank You 😊
This is such a cool challenge, technically. We should have no illusions about the fact that the search and rescue use case is a niche scenario next to the military application as fast, hard to hit scout and assassin drones though. :-(
Great video. Very interesting to see all this innovation. Well done guys.
Thank you Siggy 😊
Thanks Siggy!!
Well done Thomas and team this was so well documented and narrated
Thank you 😊
I saw the other day what I can only describe as a flock of drones all flying in like a formation almost. They were not lit up or have lights really. But they looked to move together almost. Like 1 controlled them all sorta. Hard to explain. Super cool to see. Only lasted about 2 min they fly away fairly quickly. They've very small too. Maybe even smaller than ones in this video.
I would bet that the so-called "vision-only" drone in reality uses vision and gyro and accelerometer data, paired with a database containing the exact dimensions and geometry of that particular track. By using gyro and accelerometer data, the drone can calculate in real time exactly where it is in respect to its starting location. If it also knows the track's dimensions and geometry, it can combine that information to navigate the course. Inertial navigation. I don't think the vision drone is flying on video alone. Still, I am thoroughly impressed at how fast the ai drone was. Thank you so much for sharing this, I had a blast watching it. Best regards from Brasilia, Brazil.
Awesome video!
I'd like to know how much of the processing capacity the autonomous drone was using.
In other words, does it need more processing power or just more AI learning and optimization?
I feel as though the AI needs optimisation. As they said, it can’t correct for anything it’s not programmed to correct for. But then that raises the question about whether or not it has enough processing power to account for these new optimisations on top of what it’s already doing.
The AI processing for these will be pretty intensive. It'll have to be doing pretty heavy inferences at a fairly high frequency to give the drone enough reaction time at those speeds.
The tests I did a few years back on a fairly new iphone for the time, doing image recognition on a comparatively slow loop, would chew through the battery and heat it up heaps.
This won't be done on your standard flight control processor, they have additional hardware for the inferences.
@@droneys1348 training an AGI (Artificial General Intelligence) which allows human like general learning isn't going to be feasible currently.
But sounds like they have a bit more training of the AI model that could be done to make it more robust. It'll still have to be limited to generalisations that it can learn from their dataset though and won't handle truely unique situations it's not familiar with.
Nice Info | Great Tech,.....!! Thank you for sharing with us,....!!!🙏
Thanks Buddy
The most mind blowing thing in FPV since FIRST DISCOVERING FPV PERIOD!!! 😱
They still have a long way to go, but it's one heck of a start
@@BMSWEB Today you guys are absolute heroes but in a Skynet scenario you guys are going to be looked at as the absolute worst 😂 BMS Drone death swarms 😳 I couldn’t help but keep imagining that thing chasing me! I mean the thought is extremely brief but on the upside at least it’ll be quick 😂
Love how excited and positive you kids are about creating attack drones that can chase down a human easily and avoid multiple obstacles....dont think about it just geek-out and enjoy freedom while you can.
exactly, maybe they need to look at an episode of black mirror. That's the whole issue with ...they only ask them self, can we do it...not should we be building this...
@@panobirds correct, cyberdyne will be perfected by college kids, cause its cool
Crazy fast! Nice video team BMS.
Thank you 😊
I watched some of this live on Instagram, pretty crazy stuff they are doing there 👊
Es impresionante la velocidad a la que vuelan esos drones y la habilidad de los pilotos es simplemente espectacular.
Awesome Video! Andy very nice explained.
Thank you 😊
Great video! Thanks for sharing!
Thank you 😊
Very interesting! Lots of speed on a small track, wow! Good luck with your adventure's.
This was incredible. I hope to see leaps and bounds over the next 12 months
would love to see updates on this and see the dvr footage
this is just awsome
I feel like some future version of this tech could bring ghost-runs from velocitydrone into real life, like an FPV race AI trainer/opponent with different difficulty levels and both pro and famous profiles. You could even have a mode with a specific laptime that you wanna beat or a nemesis mode where it keeps getting faster and faster.
That way you could get real life race training even when you don't have anyone to race.
Hi! We actually spoke to the guys about this!
-Thomas
@@BMSWEB very cool!
This is AMAZING!! well done to you guys. :D
Thanks
Thanks Huey!!
Yes for sure, come visit South Africa
This was awesome! Thanks for sharing!
Thank you 😊
Thank you for sharing the experience
Thank you 😊
Nice performance you got there. thank you
Thanks!
impresionante como puede volar por IA ESE RACER ,ES EL COMIENZO Y FELICITACIONES PARA ELPILOTO HUMANO ,SALUDOS DESDE ARGENTINA
Amazing video Thomas!
Well done Thomas nice narration and story line it felt like watching a documentary. I'm sure its just a matter of time before the logic in the AI drones is improved to do more than just try to get through the course at the fastest time then you may have to watch out. BTW what a great experience for you !!!
11:07 Nice to see Orqa googles on Vanover :) Cheers from Osijek :)
One thing that struck me watching you and Alex fly from 3rd person is how wide your lines were while doing looping elements (conceding that you might not have been flying at 100%). When you watch the AI drone (when it works), the loops are super tight and compact. As a computer programmer and pilot, this was all very interesting.
One thing that is very clear is that it might not be this year or even next year, but I'm pretty sure that in the near future you and Alex are going to get absolutely smoked by their machines. Scary stuff!
WoW that's fricn insane 😳😱 Nice flying guys 🔥👍🏻✌️
So excited for this AI to leap out of the lab and have real world applications! From automated camera chases, to emergency response.
Way to go Thomas and Vanover...
I'm reminded of when computers were first pitted against grandmaster chess players. At first, the humans beat them easily. Then, it was a good match. Now, humans are at a disadvantage. The things that make us able to perform impressively also turn out to be our downfall. AI, once sufficiently trained, doesn't suffer from fatigue or from forgetting things. So in the same way, the things that make AI perform impressively also turn out to be its downfall. It's a delicious irony, to me. But in all cases, the fact that we are able to train machines to perform on par with humans is itself a very impressive accomplishment. Let's hope we continue to improve how we do this so as not to put ourselves in unnecessary danger/risk.
Great video. Thanks for sharing!
Great video! And nice wins against Alex and the AI. Eventually, AI will be completely unstoppable 😭
I honestly think we're still a far way away from that. This was an extremely controlled environment but who knows hey
@@BMSWEB I've formed that opinion after doing some AI research - it's insane what the next couple of decades will hold. But - it'll only happen if people focus on making it happen... which as you say, may be a while off
Great video Thomas 👍🏾💪🏾🔥
Thanks Wiggan!! Miss you man!!
Big THX for this Content .. i love it.
Glad you enjoyed it! :D
Very cool and a great job by everyone in the team! Yet this is what the horror short film “Slaughterbots” warned us about... I don't think that this technology will only be used for "search and rescue"
wonder if the wake of a leading drone if throwing off the AI.
large thrust column affects distance traveled per input.
does it fly like 10 ms full throttle then turn...
or does it measure movement via accelerometer and gyro?
and the AI is running on a computer and interfacing with a Tracer TX?
The vicon system is a normal drone where all the processing and controls is managed on a ground based computer, while the vision drone does all the processing onboard!
It was awesome to watch!
great video dude !!!
Thank you 😊
and now you have to fit this to an X Class rig :-).
Great video bro, nice information
Can you please attach the relevant papers to the description? Thanks
Amazing, glad to see we humans are not phased out yet :P FPV=life
I predict us Human Pilots will be around for many years to come. It will be a while before they will come close to flying in an uncontrolled environment
Mind Blown!
awesome video :) impressive work
Thank you 😊
Great video
Great video! Thx!
This is the next level of Alpha Go.
Were you guys actually racing DJI or was that just to spectate the AI drones?
When the pilots were wearing the DJI Goggles, they were racing the Vision Drone. They were flying the drone provided by the University. However when racing the Vicon drone, they used their own goggles and drones
Great stuff Thomas! I missed you at IO. Hope to see you and your dad soon!
Thank you 😊Same here, missed doing the AI, shame this was on the same time. Fingers crossed will be back in 2023
Thanks Gator! Fingers crossed we make it next year!
awesome work guys !
The AI doesnt trash talk or make excuses either, thats refreshing😄
This was amazing!
so sick
Thanks so much man
Thanks dude!!
Thanks bro that was really helpful
THIS IS TO ALL THE PRO PILOTS!!!!! vanover and such :
have you guys discussed with the artificial intelligence designers about how you can recover the drone after it impacts the gate or an obstacle???? I think that that could come in very handy. Especially in the field when they actually apply this technology. Maybe give them a little jumpstart so that they don’t have to crash in the stuff so much. What I mean by this is try to convey the technique of saving a drone after an impact.
If you have ever seen Mr. Steel fly in person you know what i mean.
I have seen him save his drone from the most certain of impacts. When I watched him the first time I was completely amazed how he could just bounce off things and whatever technique he was doing he was able to keep the drone in the air and save it. That is what I mean if you could help them with that. Just a thought thanks for the video amazing 🤩 !!!!!!!
Wow, great video.
Thanks :D
Great vid and interesting experiment - but I must say that with all of the limitations cited, this didn't feel like "AI" or even autonomously controlled flight as much as simply executing a pre-programmed flight path...
Overfitting to the track. How does it perform when the track is modified?