AI attempts to break the world record at Mario Kart Wii's Luigi Circuit in Time trials. Thanks to masq for editing this video! Join my Discord: / discord
No one will stop people playing Mario Kart AI haha
@Random_Guy-rm4tv
3 күн бұрын
@@aitango honestly it would be a cool idea if theres some CPU levels that are just a well trained ai,this would be very helpful in some games like smash bros since the cpu 9 is just broken (and broken as weird and not insanily good because it have some weird outputs)
@rembartx
3 күн бұрын
@@aitangoare you gonna make a rocket league ai😢
@CrossyChainsaw
Күн бұрын
@@Random_Guy-rm4tvon it
@NMeade
2 күн бұрын
This video was AWESOME, so cool to see the entire process for approaching this. Would be really cool to see this for other tracks!
@aitango
2 күн бұрын
Hey NMeade, I’m a huge fan of your channel! Would love to work with you if ever you’re looking for a new video idea, including trying some other tracks
@haiden9170
2 күн бұрын
@@aitango maybe on Maple Treeway , it's is her favorite track ( if i remember correctly ) or Mario Raceway i don't remember
@aitango
2 күн бұрын
I’m good for either of those if he wants to :)
@Scalene17
2 күн бұрын
NMeade versus the ai on his big bad castle
@kazii_the_avali
2 күн бұрын
@@aitango ngl i would love to see maby some fanmade tracks
@masqmp3
3 күн бұрын
really enjoyed editing this one!
@aitango
3 күн бұрын
Really glad you enjoyed it, amazing job with the editing, couldn’t be more happy
@Kat21
Күн бұрын
whats the song at 0:37 i think ive heard it before
@masqmp3
Күн бұрын
@@Kat21 The swimming mini game on Wii Party 👍
@xiaolin867
Күн бұрын
So you're the one that added the unnecessary, attention-span subtitles...
@campbellblock3061
3 күн бұрын
It’s so cool to see how far the AI project has come. It’s even starting to use preliminary quickhops! But one thing that would save an enormous amount of time is timing the chain wheelies. I noticed that the wheelie input is just being spammed every other frame on straights and that seems to result in half chains every time. Timing the chain 181 frames after the first wheelie would improve this considerably on LC, and it would also help to not fill up the input file on longer tracks later down the line.
@craftyawesome6830
3 күн бұрын
He mentions at 7:49 that it only inputs every 2 frame, so it is possibly not wheelieing for multiple frames per chain. (in fact missing 3 frames every time if I understand correctly)
@clouds-rb9xt
2 күн бұрын
@@craftyawesome6830Why not every frame then
@craftyawesome6830
2 күн бұрын
@@clouds-rb9xt Probably performance. 30hz inputs instead of 60 means half the amount of times you need to compute the inputs the AI chooses.
@clouds-rb9xt
2 күн бұрын
@@craftyawesome6830 Need a super pc tbh
@lorenzosotroppofigo1641
2 күн бұрын
@@clouds-rb9xtHe already has a beast pc. Neural networks are expensive
@Loki_SR
3 күн бұрын
Been waiting for an installment like this into this little series since the beginning. Commenting to let you know I would LOVE to see more
@aitango
3 күн бұрын
I’m pretty confident this won’t be my last time trial attempt :)
@aitango
3 күн бұрын
I’m pretty confident this won’t be my last time trial attempt :)
@NathanWoyessa
3 күн бұрын
This is who I though TAS was as a kid
@alansmithee419
2 күн бұрын
In a way this is a TAS. It's very different from other TASes but still.
@UrMumLuvsUHun
2 күн бұрын
Your way of getting the AI to do shortcuts is really smart, it never clicked with me that it wouldn't just do that on its own
@aitango
2 күн бұрын
Yeah that's pretty much expectation vs reality in AI... if it can take an easier route, it almost always will
@LuckyHamburger
3 күн бұрын
Personally I would have opted to get the ai to take tighter lines by making the checkpoint reward proportionally to how close it gets to the wall. I'm not sure how feasible or useful that would actually be but I'd love to hear your thoughts.
@FuzzyJeffTheory
2 күн бұрын
You have to be careful introducing extra rewards, as it can lead to suboptimal behavior. The AI might slow down a lot to get as close to the wall as possible
@Pyrite3
2 күн бұрын
@@FuzzyJeffTheory is right.
@fluffsquirrel
2 күн бұрын
I've been so excited waiting for this! Fantastic execution and thank you for explaining your methodology!
@aitango
2 күн бұрын
You're very welcome, really glad you enjoyed it! I'm far from finished with time trials too
@fluffsquirrel
Күн бұрын
@@aitango Epic! I'm so glad, looking forward to it! Amazing work as always
@mango10152
3 күн бұрын
another banger! good job
@aitango
3 күн бұрын
Appreciate it!
@8_chips296
3 күн бұрын
ai: insanely good at the game logan: still somehow better
@aitango
2 күн бұрын
Logan is too powerful, but the gap is closing…
@ImJustCath
3 күн бұрын
That slick subscribe insert made me laugh and got me to, GG.~
@knownas2017
2 күн бұрын
These videos are so enjoyable to watch. Thank you c:
@aitango
2 күн бұрын
Glad you like them!
@AndyLundeen
2 күн бұрын
I'm looking forward to watching the AI mogi and clan wars after they get TTs conquered. Incredible work!
@aitango
2 күн бұрын
AI vs Human clan wars would be so much fun, even way harder than time trials
@DDoig1
Күн бұрын
The vocal fry is crazy :D
@Renardeau
3 күн бұрын
still , now it seem possible to go beat the wr
@aitango
3 күн бұрын
Feels so close… one day
@14sirs
3 күн бұрын
W masq moment
@aname4390
2 күн бұрын
Based video. Grumble Volcano no ultra sounds like a challenge.
@atomicstorm7006
2 күн бұрын
Is there any way you could train ai by letting it analyze other people’s inputs? For example you could let it watch the top 100 times on a course over and over and let it train that way.
@lorenzosotroppofigo1641
2 күн бұрын
Doesn't make any sense to train an AI on the same dataset "over and over again". Also it's a completely different model and the dataset is too small. What you are asking is so infeasible that the first time this was done with good results was like 6 months ago by OpenAI. I mean they tried before, but it wasn't that good
@rodafy9110
3 күн бұрын
was looking forward to seeing it develop this far. it's good enough to beat me on LC now... would be cool to see the AI eventually learn how to play online!
@TomMKW
2 күн бұрын
haha the ai actually learned how to do TAS hops. that is so cool.
@aitango
Күн бұрын
Yeah it was really cool that it actually figured that out by itself
@preacherm4n
2 күн бұрын
good video! now have the ai learn how to superhop lol
@aitango
2 күн бұрын
Thanks!! That would take quite a bit of training haha, although still possible I think
@elvsrbad2
Күн бұрын
I would have liked to see if it could have found that shortcut on its own without the out of bounds area, but overall good job!
@kiunthmo
2 күн бұрын
you could pretrain on ghost data. would be interesting to know more about how you were able to actually get the data, whether there were tricks to make it run faster, etc
@SSLuis98
2 күн бұрын
This is quite impressive, huge advancements were done. I suppose you could create agents for each track so the AI would depend on them whenever it plays, this would he useful when playing against other players / cpus?
@athehybrid
3 күн бұрын
awesome video as always!
@aitango
2 күн бұрын
Really grateful for the support!
@manfredi19
2 күн бұрын
such a good video, keep it up
@aitango
Күн бұрын
Thanks, will do!
@29lovvan
2 күн бұрын
Never heard more vocal fry in a commentary before
@samuraimaster321
2 күн бұрын
A sense of proudness lol
@aitango
2 күн бұрын
It’s quite weird but it’s definitely there
@lucascowdenisawesome
2 күн бұрын
Amazing video your RL models are really great to see in action. Do u by any chance ever share your repos or have a vid walking through the basics of your code? I’m trying to learn RL more and would be great to learn through exploring it in a context/game I love
@aitango
2 күн бұрын
I currently haven't made the repo public, however will do once I reach 100k subs. I'm currently trying to publish the algorithm I used for this video, and if it gets published I'll for sure do a video on it
@breezyx976
3 күн бұрын
What would happen if you made an AI to calculate the expected remaining time to complete the lap, based on the current state, so the main AI knows what to optimize? (I forget what this is called). Or is that what you do already?
@alansmithee419
2 күн бұрын
This sounds like a form of proximal policy optimisation (PPO). Have a second AI that estimates the time it will take the driver to complete the race if it performs a certain input at the current time, and then learns how good its guess was based on the actual time it takes, so it then adjusts the best input the driver should have taken, and the driver constantly adjusts to be more likely to use the move that the second AI predicts is best. Is this what you meant? He is using a form of PPO, but I believe he's using it on the reward function, rather than on "time to complete the race", as that often results in an extremely difficult learning process for the AI. This is why reward functions are used in cases like this.
@Goomey
2 күн бұрын
This is extremely impressive wow, there is very little left to optimize at this point. The only immediate things I could think of would be a delayed drift (that way you don't have to hop to align for the back straight) and then obviously chain wheelies
@aitango
2 күн бұрын
Technically kind of I guess
@Goomey
2 күн бұрын
@@aitango I'm just nitpicking at this point, the vid is insanely well made and easy to follow, please make more! :D
@chaoscodytv
3 күн бұрын
Yay a new post!
@aitango
2 күн бұрын
:)
@epicawesomesauceguy
2 күн бұрын
i love this. have you tried using continuous actions for the steering instead of 5 discrete ones? it may not be necessary but i'm curious
@aitango
2 күн бұрын
The method I’m using doesn’t actually allow continuous actions, however is far more sample efficient than those methods. I’m not sure how big an improvement adding more angles would actually give it, after like 7 you’d probably get diminishing returns
@Hélène-b9t
3 күн бұрын
I never seen such a great video!
@aitango
3 күн бұрын
Wow, thanks for the support!
@alganpokemon905
3 күн бұрын
great vid. the ultimate would be an agent that can play unseen tracks lol. how much compute would that take?
@aitango
2 күн бұрын
Glad you liked it. Would probably take ages, especially if I individually design the rewards for each track.
@alansmithee419
2 күн бұрын
He has tested things like unseen tracks before, but they've never been very good at them. He did however find that the AI could adapt to new tracks very quickly via further training if it had prior training on a few others.
@mkwpatsfan
2 күн бұрын
Can you make the A.I time trial DK Summit next?
@_Dxnny
Күн бұрын
I have no idea how any of this works but could you get the AI to watch how the WR plays the track to copy and improve on it?
@lonelyPorterCH
2 күн бұрын
Dang thats impressive :D
@aitango
2 күн бұрын
Thanks! Still want to get that wr though
@BillboMC
Күн бұрын
I’m surprised it didn’t master chain wheelies? Or is that a side effect of it not being able to input every frame?
@ruhenot6682
2 күн бұрын
Very interesting video. I'd suggest to cut down on the subtitles though and only use it if you want to emphazise what youre saying. You have a very soft spoken and clear voice. Constantly having text on screen is stressful and in your case just unnecessary.. :)
@aitango
2 күн бұрын
Glad you liked it! Yeah might try using them slightly less and see how the videos do
@andy02q
Күн бұрын
TAS is <a href="#" class="seekto" data-time="62">1:02</a>:446. If it can't beat that using TAS tools, then not much point to it.
@GarkadeXD
3 күн бұрын
Let’s see your ai do n64 sherbet land glitch lol
@RazorBird
3 күн бұрын
Thanks!
@aitango
2 күн бұрын
No problem!
@samuelzackrisson8865
3 күн бұрын
Really interesting video. I have a quick question though how do you play multiple ghosts at the same time in tas tools?
@tascreeper
3 күн бұрын
that is insane.
@aitango
3 күн бұрын
Yeah pretty crazy, I hope one day I’ll get the world record
@WizardOnYoutube
Күн бұрын
We need to take the ai and the game and subtract the rendering so it can play many many many instances a second Run the game such that it does like 10-15 races a second and let it run for a year😊
@aitango
Күн бұрын
There’s a cool project doing something similar called kinoko which is looking to reverse engineer the physics engine. My ai uses images as input though, so sadly I still need to render the image
@サビン-w5j
3 күн бұрын
Do you think it will be possible to teach an AI a tech track like CM or bowsers castle?
@aitango
2 күн бұрын
Definitely possible, the main problem would be my knowledge setting it up!
@spacebarium
2 күн бұрын
goddamn how did it learn to do TAS hops
@aitango
2 күн бұрын
Honestly no idea how it figured out how to do the optimised shortcut, really blew me away
@CodeChrisB
2 күн бұрын
Hey any way to get some information on what tools you used . I certainly have the skills to do so I have a degree in theoretial AI design but never touched AI gaming. Even better would be perhabs soms source code? I know thats quite a lot to ask for but any help would be appreciated. Im a retired software developer (23 yrs old lmao) and think creating an AI of my favourite mario kart would just be amazing
@ДаниярКадамбаевич
2 күн бұрын
Мне кажется Или это искусственный интеллект скоро появится сингулярность где он бесконечно усовершенствовать себя в игре Mario Kart это будет безумие а потом он захватит весь мир я смотрю твоё первое видео и вначале он был таким то что ошибался в начале трассы😂😂😂😂
@ДаниярКадамбаевич
2 күн бұрын
Те кто не знает что такое сингулярность Это такая вещь то что у искусственного интеллекта есть сингулярность а это значит он будет обучаться до бесконечности то есть усовершенствовать себя Он играет в одну игру как например Mario Kart а потом он переходит на другую игру потом создаёт своё ии потом заканчиваются все игры переходят на приложение потом на телефон потом на роботов а потом захват мира их невозможно задержать даже если вы запихнули самую защищённую во Вселенной компьютером рано или поздно он просто Придумай как выбраться из компьютера
@flambambam
3 күн бұрын
Now I'm curious what this ai could get with shooting star. I hold the outside drift bike bkt of <a href="#" class="seekto" data-time="72">1:12</a>.108, which I think is pretty dang good for the combo.
@nobafan7515
3 күн бұрын
This is amazing. Do you know where i can learn how to program an ai like you did? Is there a youtube tutorial that sets this up? So far i seem to find out of date tutorials or none that tun while the agent is shown playing as the ai trains. Any help would be really appreciated.
@14sirs
Күн бұрын
What are the specs on the PC that you are training your AI on?
@gabrielechioda7069
3 күн бұрын
Impressive
@aitango
3 күн бұрын
Thanks
@ObeisantBug3808
2 күн бұрын
do you have the ghost for this still? I would like to race it because it beats me by .1 and that end time is about my short term goal on the track
@aitango
2 күн бұрын
If I remember I might post it on my discord server sometime soon
@NotK41
3 күн бұрын
This is so awesome! I wonder if one day it could be used to discover new glitches and optimal setups for shortcuts?
@bl3icom
2 күн бұрын
the hardcoded mrbeast style hype subtitles with the soft gentle voice and.. idk maybe is just me but it makes it hard for this boomer to watch. nauseating even. that meta, is a nope from me. but still gg on the improvement but i just cant watch this lol. please dont take this comment too harshly, gotta deliver to that audience 🙏
@bytebeamer
3 күн бұрын
Can you make an ai to play taiko no tatsujin?
@mkwpatsfan
2 күн бұрын
How did the AI learn how to quick-hop?
@aitango
2 күн бұрын
It taught itself through trial and error, I just gave it the controller.
@relt1554
2 күн бұрын
Is it faster to hop to cancel a wheelie or just cancel the wheelie via the "tilt Wii-remote back down" input?
@khirek5335
2 күн бұрын
They're equal but hops are more reliable and you only really stop a wheelie when you want to take a turn and start drifting anyway
@Grofvolkoren
3 күн бұрын
Awesome videos, but you have a lot of vocal fry in your voice. Are you aware of that? Still a fan of what you are doing.
@aitango
3 күн бұрын
Sorry about that, I didn’t notice anything myself. Glad you could still enjoy it though. I’ll try and sort it out
@masqmp3
3 күн бұрын
hi, we used adobe's podcast ai to enhance the microphone audio. the original audio doesn't have the 'vocal fry', instead it's the podcast ai emulating a high quality studio microphone 👍
@CrossyChainsaw
Күн бұрын
Read memory from the ram? How?
@Ifx-
3 күн бұрын
is it possible to make an ai that can play any track its given, even if its never seen it before? so that instead of getting really good at one track its getting good at the game as a whole and applies that skill to any new track?
@Esto12
2 күн бұрын
What did you use to be able to get the values from dolphin ram into python?
@aitango
2 күн бұрын
I’m using a fork of dolphin which allows me to interact with the emulator in python. Check out my video evolution of my Mario Kart AI to see more
@Samed_2706
2 күн бұрын
Average lap time is 33.222 seconds lol
@alansmithee419
2 күн бұрын
That would imply its final time was ~1:40. It was 1:10 (rounding up), so 70 seconds. 70/3 = 23.333... Or are you referring to something else?
@noahcheng1895
2 күн бұрын
can it play bowser's castle?
@aitango
2 күн бұрын
It can play any track in theory, I just need to help designing the reward function!
@craftyawesome6830
3 күн бұрын
<a href="#" class="seekto" data-time="460">7:40</a> Were these stick values carefully chosen? AFAIK there are optimal spots for soft drifting.
@LLInnovA
3 күн бұрын
tfw your pr is slower than an AI generated run :( I want to see it try all the tracks. You should also provide the ghost so we can race against the AI
@korben669
2 күн бұрын
you didnt teach it how to softdrift or delay drift or most of the basic driving fundamentals thats the main reason why the time wasnt that great
@MurdogYT
2 күн бұрын
Satan.
@aitango
2 күн бұрын
I did think the AI getting a .666 was a funny touch
@JEdits1
3 күн бұрын
I feel like if you did this to where it could actually learn to do each track optimally, I think it would be a cool way to calculate the theoretical fastest time on each track
@Stabidistabstab_PBP
3 күн бұрын
Are you going to put in more rewards and penaltys or is this it?
@aitango
3 күн бұрын
Since posting I have a few more ideas to improve things, but that’s for next video haha
@Stabidistabstab_PBP
3 күн бұрын
so there is a chance, pog
@ragazoub
Күн бұрын
Kart gpt
@HeeeeeeyHowAreYou
2 күн бұрын
Now do it on a track with more variables such as Shy Guy Beach
@tubbysza
20 сағат бұрын
Everyone enjoy his replies while it lasts, when he's bigger, he will never reply!🥰
@Makabert.Abylon
Күн бұрын
Spoiler alert🚨 It can’t
@carolinefell4505
3 күн бұрын
Under 30 minutes gang
@anonym5160
3 күн бұрын
Can you please make AI plays Taikyoku Shogi?🙏
@aitango
3 күн бұрын
I think Deepmind made an AI to play shogi called alpha zero
@anonym5160
3 күн бұрын
@@aitango Whe‘re can i watch the games?
@Saxofome
2 күн бұрын
AI or TAS.They're not similar to each other
@aitango
2 күн бұрын
I think world records will soon need a new category
@Winter_Fan_01
3 күн бұрын
What if you teach the AI that it's going to get a massive rewards for every 500ms of improvement, then every 100ms as it becomes better. So that risk is worth it. Then 20ms. Then 1ms.
@lumpylumpyloo
2 күн бұрын
Ai generated video
@kupecxa9843
2 күн бұрын
lol choosing 2 frames as the minimal interval of changing inputs and only 5 different direction imputs to choose completely defeats the ai's advantages, no wonder the ai wasn't getting good times
@marvsky4845
3 күн бұрын
I cant stand your "vocal fry"
@edenalexandriab9120
2 күн бұрын
Remember - you are loved! Jesus loved you and gave his life so that you could be free because of Him! He died for our sins so that we can be reconnected with our Creator God. He wants us all to repent of our sins and receive a new life in Him! A life of strength, faith, and the ability to truly love others as God has loved us. He wants us to follow Him, and let Him be the Lord of our lives. Why is that part important? Because through Christ we will be given the power to overcome sin in our lives today. Christ offers this blessing to all who choose to believe in Him, trust Him and follow Him xx He will never leave you nor forsake you xxx
Пікірлер: 157