AI vs AI 100m Dash (deep reinforcement learning)
ฝัง
- เผยแพร่เมื่อ 21 ต.ค. 2023
- AI Competes in a 100m Dash!
In this video 5 AI Warehouse agents compete to learn how to run 100m the fastest. The AI were trained using Deep Reinforcement Learning, a method of Machine Learning which involves rewarding the agent for doing something correctly, and punishing it for doing anything incorrectly. Each agent's actions are controlled by a Neural Network that's updated after each attempt in order to try to give the agents more rewards and less punishments over time. Check the pinned comment for more information on how the AI was trained!
Current Subscribers: 264,870 - บันเทิง
![Genshin Impact | กาชา Clorinde ไวฟุนั้น ควร C6 [เทสตัวละครต่อ]](http://i.ytimg.com/vi/SOVaBvL7K20/mqdefault.jpg)
It took me months to make this video and it took my computer over 3 days straight to train/record the agents, I hope you enjoy it:D
After teaching Albert to walk in the previous video, I read a lot of comments asking about what would happen if I used a more human way of punishing and rewarding Albert, so that’s what this video is about! Each agent starts off the same, the only difference being the design of their body. They’re each rewarded for moving forward and punished based on the efficiency of their movements (based on a muscle fatigue system), so by the end of the video they each should discover a movement that works efficiently for the body they were given.
NOTE: Don’t worry, Albert is coming back in the next video, he’s hard at work right now improving his walk:)
If you're interested in training your own AI like Albert but don't know how, there's now a really easy way to do it! Luda, an AI lab, recently built a web app that allows you to create and train your own AI using deep reinforcement learning (just like Albert) completely for free in your browser! You build your own character (called a Mel) with lego-like building blocks then watch it train in real-time on their website in just a few minutes (really). It's an awesome project, and just like my videos, makes deep reinforcement learning so much more accessible, which is why I love it so much. This section of the comment is sponsored by Luda, but these words are entirely my own, it's an amazing project that I would have been obsessed with had they released it before I built Albert. I've genuinely been looking for a sandbox/game exactly like this since I was a kid. They're still early, but they're giving my audience first access to their closed, pre-alpha build. Make sure you check out their site and create an AI agent for yourself!:D prealpha.mels.ai
Now, back to our agents,
If you want to learn more about how the agents actually work, you can read the rest of this very long comment I wrote explaining exactly how I trained them! (and please let the video play in the background while reading so TH-cam will show the project to more people)
THE BASICS
Although it seems like there are only 5 agents training here, there are actually 40 copies of the video being simulated simultaneously behind the camera in order to speed up the training, so although the video makes it seem as though there are 1638 attempts, there are actually around 65k.
Each agent is controlled entirely by an artificial brain called a neural network. Their brains have 5 layers, the first layer consists of the inputs (the information they’re given before taking action, like their limb positions and velocities), the last layer tells them what actions to take and the middle 3 layers, called hidden layers, are where the calculations are performed to convert the inputs into actions.
Each agent is given quite a lot of information about its body, they’re given everything that Albert was given in the last video (which I explain in great depth in this pinned comment th-cam.com/video/L_4BPjLBF4E/w-d-xo.htmlsi=HHv3vrmgIxUGo54f).
Just like the last videos, the agents are trained using reinforcement learning. For each attempt an agent has, we calculate a score for how 'good' their attempt was and the training algorithm we used (PPO) makes small, calculated adjustments to that agent's brain to try to encourage the behaviors that led to a higher score and avoid those that led to a lower score. For this video there are 6 different ways each agent is rewarded/punished, and I tried to make these reflect our normal movements as much as possible.
REWARD FUNCTION
Movement: Each time the agent takes action we check to see how much closer the agent is to the target and we reward them proportional to that distance. If they move a lot closer to the target, they’re rewarded a lot, if they move away from the target, they’re punished.
Limb Fatigue: This is the heart of the reward function for this video, every time an agent takes an action on a limb, we punish it proportional to the strength of the movement and the current fatigue of the limb (so if the agent moves a limb that’s already really fatigued, the agent is punished severely), then we increase the fatigue level of the limb based on how strong the movement was, and with each frame we slightly lower the fatigue of each limb to simulate the limbs resting. This reward is meant to simulate muscle soreness and encourage the agents to find the movements that are most efficient for their body design, but also make for more interesting gaits, since without this punishment the agents would all likely opt for a safe shuffle and avoid taking large steps.
If you're still reading this, you're probably really smart and want to learn more about Albert, so make sure to join my discord server I just made where we can talk more about the details of Albert's AI! discord.gg/jM2WkNuBnG :)
Limb Hit: I wanted to punish the agents for falling over, so any time a limb that isn’t a foot hits something it’s not supposed to hit (the ground, other agents, etc.), we slightly punish the agent, and we also slightly increase the fatigue on that limb.
Abrupt movement: Each time the agent takes action we calculate the average velocity of their
body and compare it to the average velocity of their body when they last took action, the greater the difference in these two values the more we punish the agent, since a great difference implies abrupt movement was made, something that generally is bad for our bodies. For anyone looking to make something similar to this, this reward is really important for smoothing out the final gait!
Chest up: We give the agents a small reward whenever their chest/head is in the upright position, this helps the learning converge easier, without this reward the agents might never learn to stand up and instead just learn to crawl to the target.
OTHER
I only allowed the agents to make a decision every 5 game ticks, which made the movement look a bit more jagged than if I allowed them to make a decision every tick. I found if I allow them to make a decision every game tick it’s too difficult for them to commit to any proper movements, they end up just making very small movements like slightly shuffling forward instead of taking a full step. The 5 game tick decision time forces them to commit to their decision for at least 5 game ticks so they end up being able to take the less safe (but cooler to watch) large steps.
Though you only see one version of these agents, there were actually 40 copies (so 200 agents) training simultaneously behind the camera in order to speed up the training process. Despite this, it still took my computer (threadripper 3960x, rtx 4090, 128gb ram) over 3 days to train/record!
We're looking to hire people to help make these videos! If you're a talented Unity game developer you can apply for a full time position here forms.gle/ko54z1LQmZNUT9Vp8 And if you're a talented AI developer (ML-Agents), you can apply here! forms.gle/Uou1Vwb5Q9VccaAY7 We're looking for full time employees, but part time works too, what we're really looking for are skilled and passionate people, so feel free to apply if you're interested! :D
Thank you so much for watching, and please, if you enjoyed the video or learned something, share it with someone you think will also enjoy it! :)
Your animations are great
Fr
So long not see you
How are you
Pog
I’ve been waiting so long, thinking you stopped making videos, thank you for your dedication!
red throwing a tantrum in the middle of the track so now noone else can pass either was hilarious😭
Mood
6:51 😂
also on 8:45 💀💀
Typical bipedal behaviour if you ask me 😅
AI don't have emotions so your comment isn't very funny.
Purple was robbed! Constantly getting tackled by others, and starting on the disadvantageous side track with less space to manoeuvre... Where is the competitive integrity?! Red deserves a DQ for that awful, childish behaviour on run 912.
and on run 1410
They did my boy dirty 😢
ay dont insult my boy red
YEAH #DQRED
Exactly. red would always take it out on purple
This is amazing. I love Red's enthusiasm.
"Yeah I love the high jump."
"This is a race."
"HIGH JUMP!"
Front flip
3:13 FREE BJ
I have never felt so disappointed to see a video end. I hadn’t been watching the time stamp, and I was so invested in seeing them (especially purple) reach the point that they were truly racing.
i'm glad to see albert's knowledge of walking is being used to help teach
balbert, gralbert, ralbert, yalbert, and palbert how to walk too!
Okey now its they canon name
in case i was the only one who saw it, the names are based on the colors of each one, for example yalbert is yellow and ralbert is red
@@tulliuscicero852 noice :D
@@tulliuscicero852no way
@@tulliuscicero852 Yes, you were the only one who saw it, for you have special eyes. So look, look with your special eyes and spread your wisdom upon us, the unwashed masses! /s
10:02 one must imagine purple happy
*intense Syphius music plays
*Syphius picture fades in and out*
PURPLE WON GODDAMNED
its sisyphus not syphius
@@mcdonaldswi-fi2720 bro switched account to recreate a meme
If these were developed into characters.
Purple: Wild Card, Optimistic, Sometimes Lazy
Yellow: Straightforward, Patient
Red: Entitled, Whiney and Immature, Show-Off, But Very Determined
Green: Quirky, Meek
Blue: Clumsy, Curious
-Purple often gets screwed over by others but stays determined
-Red is slow to learn, behaves badly, and suffers from karma a lot
Alright PacMan relax lol
good personification！
albert is albert, he is the eldest of them all and the most learned
This... this is literally the color gang from alan becker
10:30 when you try to run from a monster in a dream
Bruh💀💀
@@bobblabbruh 🎉🎉
I can’t wait for all of these AI’s to get their own characters and lore. I can just imagine a cinematic universe for this channel
Edit: how the HELL did this comment blow up
Lol
Next video: ai fight club
This remembered me The Amazing Digital Circus.
We should make this just make a random backstory made by Chat gpt sounds good by my opinion
I guess, but I’m afraid content farms might yoink his idea and use these characters in the worst of ways :/
Red was the definition of character development
Also purple being the Pixar lamp was funny tough
Starting positions should be randomised to give each ai a fair shot at learning. Loved the captions music choices.
Now I just want to see these five AI’s learn how to work together…Similar to Albert’s puzzles!
Again, it was very interesting.
I felt that when Red fell, it dragged everyone down and negatively affected the learning of those around him, so if the focus was on running, I felt it would be preferable to have him run the race alone and then composite everyone's movements in later editing, etc.
I was rooting for Purple's run because it was so careful and beautiful... I still wonder if the longer length of one step is more advantageous?
From a Japanese fan
Translated by Deepl
Purple was the only who learned how to walk properly with one leg.
Exactly! After red started to fall consistently over purple track, purple unlearned how to hop.
@@arthurfraco2970 purple lost brain cells looking at red
100% I feel they could have just changed one variable to make them not collide with each other and it'd be even better
Having collision between the different agents adds a level of randomness that seems like it would severely hamper the learning progress.
but falling is funny
Purple's idea to piggy back on yellow was genius, I wish they kept developing that pattern.
It’s amazing how well the AI learns, even if it takes a while
It learns faster than us
@@RubyPiec Did it take you more than 1638 attempts to learn to walk? Most people manage in a little less than that
@@9nikolai what counts as an attempt?
@@9nikolai the ai learnt to walk in 3 days, most people take about a year.
@@omphya6229 The ai doesn't need to eat, sleep, or anything else than walk.
you're actually revolutionizing the AI genre on youtube
you might be interested in carykh's evolution series
yea
Lol he hearted your comment. What a delusional loser. You really shouldn't hype up content like this. It's nothing special.
This made my day, thanks 😂 i’ve been searching awhile for something that could cheer me up
LOVE this Content. Its different to all this junk on TH-cam. This is ACTUALLY Entertaining!
Can we appreciate the fact that at 10:18, yellow turned around and walked backwards but still went super quick?
yo he moonwalked
699 likes...
guess I'll be one to ruin it
700 likes...
I guess I saw it happen
bro is twerking
*HE HE*
It was interesting to think about how some AI probably got steered in a less efficient direction because they were trying things and getting stuck on the other models. I wonder how differently this would have worked out if they couldn't bump into one another.
yea sure is interesting
That's the fundamental limiter on all this deep learning stuff. The data set in reality is always messy and incomplete, which quickly leads "AI" down bad paths that living beings tend to suss out easily.
@@travisjohnson6703 AI always faces this issue, it frequently randomises to local better locations that are actually globally worse. This is easily fixed using general annealing algorithms etc. that tend to be used in most complex AI systems.
Funniest thing I've watched all year by far
Really enjoyed this work
I just appreciate that the victory was done with a john cleese silly walk
They are not identical, they are each special in their own way 😁
i am special
@@doob.yes
@@doob.yes
@@doob.yes
@doob. yes
I love that red is walking around like a extremely drunk person and how he randomly keep bullying the others like purple or green. Truly a drunk Florida man
I love how he colors some of the words red if the AI does something bad , Yellow if its okay And green if its excellent
LOL, that was so funny. My wife also enjoyed watching it. Thx for sharing.
Do you think it would've run differently if they were encouraged more to stay in their own lanes?
Yeah
prob for purple
Red’s movements resemble those of the character from that QWOP game ungainly and spectacular spills. The way he sabotages the rest of the athletes inadvertently or otherwise in the process of tumbling is outstanding .
5:01 Red: I call this the QWOP shuffle
Fun fact: that hop Red does isn't too dissimilar from the way astronauts bounce around on the Moon. It's also essentially Purple's locomotion, for that matter.
5:20 purple _deliberately_ throws itself on yellow in the hope to piggyback along 😊😊
11:09 you lied about the cake, now he's crying 😭
IEIEJEHED THE CAKE IS A LIE THE CAKE IS A LIE THE CAKE IS A LIE THE CAKE IS A LIE THE CAKE IS A LIE THE CAKE IS A
He really looks like he is hysterical
The cake is a lie 🎂
too little people catching that obvious portal reference lmao@@delta1234s
I love how you treat each AI as if they were your child. Your channel is just really wholesome.
He casually mentioned the fact that they made it in a way which they're in great pain when they fall and you're calling it wholesome😂😂
@@mjvafadar2526 true. I forgot about that.
@@mjvafadar2526 Spare the rod, spoil the child.
(Disclaimer: Do not actually follow this.)
Your billboards and advertisements are hilarious, so bad 😂😂😂❤
I honestly felt so surprided purple performed so well. I though it want going to be able even to stand up. Amazing video as always!!!
purple had the advantage of less parts, and less learning and tweaking
There's a reason worms and fishes evolved first.
9:21 That's a pretty solid frontflip. Maybe you could do some challenge in that direction too?
Lol stupid red AI
Love this ❤
purple: hop hop
yellow: slow but steady
red: "watch me do a frontflip"
green: yellow
blue: "hello where is the finish"
Gosh I absolutely adore these tiny AI buddies. I can almost see their personalities. Watching them go from confused wobbly wormies to successful walkers and jumpers is extremely entertaining! Your commentary is, as always, brilliant. Just like the joke in the end :) I do wonder though what happened to the rest of the team who was not able to make it to the finish line. Guess they're on the AI vacation where they're rewarded for simply lying around 😂
Also a huge thank you for the thorough explanation of your work, it's really interesting to read! Good luck in your further work, I'll look forward to the new video!
No theyre in a butterfly farm upstate
7:13 purple just “resting” 😭😭😭😭😭😭😭😭😭
what
lol
he's had enough
he wasnt resting he was doing the deed to himself
No way youtube geniuinely recommended a good youtuber for once. Can't wait to see Albert learn more and more (and maybe make some friends along the way?) :)
Wow, these AI videos are great when the video finished i just saw that there more about AI its so cool
can we appreciate the effort albert puts into these videos
Noooo wallibear became an ai
Didn’t expect to see you here!
Minecraft youtubers annoy the hell out of me ngl
Can we appreciate the lack of effort this waste of space puts into the garbage slop he calls content?
Also Albert is the channel mascot, not the channel owner. Moron 🤣
If you ever do another one, here's my suggestions:
1. Make the AI not collide with eachother, this will avoid dirtying the training set.
2. You could try using the old Albert agent/model (or other old agents/models) as a comparison
3. And in terms of ideas for other models, you could try a 6 or 8 legged model, alongside a spring-esc/jellyfish model that I've seen in old Framsticks simulations
Those are all cool ideas
Exept for the first one cuz funni
@@pitori. It reduces comedy, sure, but it's more scientific. Besides, in final races the collisions could be turned back on.
@@enderjed2523 how would they be able to adapt? they should be able to learn with collisions, and need to adapt to the other contestants
its funny watching them crash@@womp47
@@womp47 depends on how they set it up. If they put in inputs that tells the AI that they collided with other AI then they might be able to handle it. However if they didn't the AI will have no clue that they were being interfered with and just believed what they were doing was wrong, even if what they were doing would have got them further, making them unlearn their improvements.
4:26 had me 💀
This video taught me that by tackling others you can slow down their development and win over them
To be honest, the test rooms always made me think about the game portals. That cake reference was amazing!! Love your humor in your videos. This one was amazing!! Keep going!
The green one 💀
The purple one..
Lol
@@Spamtinglecommitting public indecency with red
Albert your working so hard and becoming wiser by the seconds. I want to find out what you're all good at or like doing and i want to help your dreams kids❤❤❤
I wonder if, when AI finally enslave us, they too will give us difficult tasks to perform in exchange of pastries, just to laugh at us. And when we finally succeed, against all odds, they will not give us cake. You're setting a dangerous record, my friend.
I noticed that since the separate AI models can collide with eachother and start each run with relatively the same behavior as the previous, an AI could use another's strategy to create an advantage for themselves. I noticed red started to lean on yellow around 2:50 to get a boost.
yeah if they would have trained seperately it might have been different.
Yeah just like how purple tried to ride yellow in 4:18 😂
green knew exactly what he was doing 3:12
TOO DEEP AI LEARNING 😂
Kinda sus
"THE CAKE IS A LIE" written on the wall of the second video finally makes sens !!! Nice easter egg haha !!!
Red's antics had me laughing all throughout the video
As a parent of a toddler i felt this.
4:18 Purple learns horse riding
Red definitely figured out how to maximize his reward by diving forward the moment he loses balance.
Would be really interesting to see how competitive this would get if the ai are all allowed to fully mature into sprinting masters. Assuming red would still end up winning with those long legs, but the others might put up some impressive competitive performances.
Red also learnt to stay on top by hindering the other agents' movements, causing their learning to be sabotaged and set back.
Smart but dirty.
@@redthered279 I doubt it got rewarded for that at all. I do not remember any indication that the ai were getting rewarded for overall placement, just for personal achievement.
I've always wanted to experiment with AI like this, what program or how do you even accomplish something like this?🤔
8:35 bruh this scene out of context xD
“You’re kind of flopping around like a worm” What do you expect? You gave him the body of a worm!
@aiwarehouse Regarding the 40 copies of the simulation running behind the scenes, when updating the policy through PPO, do you aggregate and consider the experiences from all 40 to inform the policy adjustments, or is the update based predominantly on the best-performing instance among them?
More like this please!
Red trying his hardest to be the favorite child
Thank you so much!!:D Red may have tried his hardest but nothing compares to Purple
incredible. I fully predicted the others would learn this much sooner than red simply because it's by far the most complicated body. But I guess the size of its leap, once it can finally leap, simply makes up for all the difficulty of learning to leap!
And by the end it's even a *sort of* natural motion. Like, not really, but at least it's imaginable that somebody would intentionally walk in an incredibly silly ways.
it's like a horse's gallop with springy foot joints
He might even get invited to the ministry.
Personally I am thrilled with your videos. Could you tell us, if you haven't already done so in some comments, which tools you used for these simulations? That is, what software and/or programming language and library? Thank you very much and congratulations!
i love these videos i would love to actually watch albert on my pc running around doing stuff but thats probably really hard to do
red was so smart, messed up the others so that their good habits weren't rewarded as they wouldn't get far due to the red's sabotage, meanwhile red couldn't be sabotaged and could learn without major problems
I smell someone's boutta make an amogus joke
@@its_Hazerin 2024? I sure hope not 😭
Its because red learned the way albert learned
Red sabotaged the others?? amogsus reference??
@@redthered279hazer was right all along
THE KING IS BACK!!!
I love these! Can't wait for the next one...
red just falling on his face every time always got me cracking up 😂😂
The 100 Meters AI Race - A Summary of the Competitors (From 1 to 5) (Contains Spoilers)
*Purple*
Purple is the one-legged fellow with one eye. They have a major, major problem with balancing, one that prevents them from actually being able to race most of the time. When the stars align and Purple is actually able to begin racing, they use swift short hops that are remarkably consistent... so long as he doesn't bump into anything.
*Yellow*
The quadruped. Yellow was the first AI to figure out how to properly race, and has proven to have the most stable gait. Once Yellow figured out how to walk, there was nothing that could make him fall over that I could recall. Unfortunately, Yellow is extremely slow, and unless we're talking about a competition between tortoises and hares, slow and steady *does not* win races. Yellow also has this strange obsession with walking along the fence.
*Red*
The one who's form mimics that of Albert, the Most Heavenly and Holy Strider. Red has precisely none of Albert's grace and coordination, having a false-start rate that's as bad as or perhaps even *worse* than Purple's. When Red does manage to figure out how to use the holy form they were blessed with, they use either some sort of strange tip-toeing walk or big, lunging gallops.
*Green*
The tripodal unit. Green is just behind Yellow when it comes to stability, and just ahead of Yellow when it comes to speed. Green's most notable accomplishments include being the second one to figure out how to stand stably and that one time they got stuck doing a headstand.
*Cyan*
There's a fifth racer? What're you talking about-- Oh, that one! Yeah, Cyan is shaped sort of like a Goomba, essentially a head and two legs. Cyan... look, Cyan might as well not even be there. They have one notable trait, and that's that they tend to walk down off of the track assuming that they go on for long enough.
*And the Winner is...*
RED! W-Wait, Red? The one who can barely even stay upright even though they've got two legs, two arms, and the inherent holiness of Albert's form? That guy? Huh, okay... I can only assume that it was the aforementioned holiness that allowed Red to win... or maybe it's because they had the longest stride? It's one of the two...
Wow
Red's also a big crybaby who loves taking his anger out on others, especially Purple
I was upset too because Purple was stopped when he was just about to win :(
4:38 blue crying on the floor
Green seems worried about red.
didnt notice that😂
blue is down pounding his head against the kitchen floor
This is hilarious, congratulations!
I love how red just chills on the ground in the end :D
1:48 Red wants to push up XD
|8 |
5:00 Red found his inner QWOP.
11:00 The cake is a lie.
Small detail. The cake is shiftted.
NOOOOOOOO
I remember seeing this channel when it was super small, crazy man, awesome vids ❤
Best channel
At 8:10 I love that green came up to the skill of breakdancing 😂
Green prob like "screw running i wanna dance"
0:22 ALBERT
real!!!11
OMG ALBERT STEAL CAKE!!
*ALBERT*
Well this was very entertaining!
Red the funniest lil dude, every sinle time he fell over the barrier I bust out laughing.
6:55 i like how green looks like as if he is really concerned for purple
This was an absolutely delightful video! It was extremely entertaining how each body took a unique personality - I found myself rooting quite a lot for Purple as they really put the effort in! I can't wait to see Albert's return, and maybe the return of our newfound friends here.
i mean albert DID escape.
Another fantastic example of deep reinforcement learning. Great job! Keep it up. I hope you are able to share your information and methods one day.
Red doing gymnastics is something I didn’t think I’d laugh so hard at
This was so worth the wait, thanks for showing us this masterpiece!
Before I watch this to the end, I am rooting for purple. I want to see the monopod win.
(After watching it, second place ain’t so bad. They would have gotten it if they had an extra second or two.)
This was really cool. What amazed me is how, by the end, Red basically almost solved walking. Of course it was wonky, awkward, and unstable, but the fact that it was closing in on the right solution for bipedal walking is fascinating.
The look of determination on red's face is astonishing!
Yay finally another video, it's just unfortunate that they take so long to make
I also am trying to make my own walking ai and i also am planing to (hopefully) make it working phisical body, so these videos always are a great help and inspiration for me, keep it up!
3:10
Aint no way green did that in front of the camera 💀
He wanted a DIFFERENT cake
He wanted to put the frosting.
@@penutbutterandjelly8523WTF?
Red's character development is absolutely beautiful
This was a lot more interesting and intriguing then I thought it would be. I couldn't stop watching 😂 I voted for Purple because the poor thing only had 1 leg - I always go underdog! And laughed when Red just trips him in one scene 😂
I was rooting for red from the start and honestly halfway through I thought I made the wrong decision, but when that man came jeeping and juking thru the other competitors on good pace and crossed that 100m mark. I cried a tear of joy
no bro im sorry he removed from the video the part where purple won :(
gaslighting is real
@eugenioreale7588 no, Purple didn't reach the end in time. Red won.
@@eugenioreale7588purple didnt win it ran out of time 💀
@@eugenioreale7588 purple didnt win bro, he ran out of time
7:21 I like how hitting 60m is literally hitting 60m
Oh my god it's been a while since i've laugh so hard, the red one left me in tears
Blue just being a stocky head on legs brought out the Battletech fan in me; I can only imagine this was how getting the very first Industrialmechs to walk went.
So interesting! Thanks for entertaining me with this silly and fun video