DESTROYING Donkey Kong with AI (Deep Reinforcement Learning)

Donkey Kong it's a game let's do that I guess I don't know and because I love you and hate sleep in a social life I'm gonna be testing not one not two but three different AI algorithms I thought it'd be cool to like compare all the algorithms I've used on previous projects on a single project and see who wins I'm talking the best of the best we got genetic algorithm we got neural evolution of augmented topologies and we've got proximal policy optimization going head to head but we're getting ahead of ourselves game first AI later okay let's make Donkey Kong okay player on screen done it's a square very nice creative work nice now I could move around the player using unity's built-in physics system but that will animate the player way too smoothly and in order for it to have that classic arcade feel it's got to have a bit of Jank movement in it so unfortunately I'm gonna have to be a big boy and code the physics myself okay let's move that player can I go right yes can I go left uh close enough cannot jump ah well yeah ah close gravity yes less gravity yes [ __ ] yeah everything is coming together this must be what God felt like dude you're just moving a square around the sideways jumping yes sideways jumping and gravity at the same the same time you stopped ground no ground no tutorial parents true or false depending on weather uh the move was successfully come yes ground no [ __ ] ground no this is getting annoying Grand no that's actually pretty aesthetically placing but not at all what we're looking for around oh no but we're close we're close and I can feel it Grant yes yeah [ __ ] easy bit of a challenge next time okay so the ground's done and it's all looking pretty good now I know it's nothing revolutionary but you know baby steps so problem with Donkey Kong is that we've got to handle these little steps and it's not like a smooth slope either it's a bunch of individual steps you see if I try this now it's just gonna oh oh what the what it already works I I don't understand it never works first try [ __ ] this is never this has never happened before Oh I'm I'm scared oh no it's okay it's okay it's [ __ ] up whoa It's Fox the code's [ __ ] oh it can't step down oh that was close oh false alarm false alarm okay here's a more extreme version of the problem we ain't falling right luckily this is a pretty easy thing to solve just shoot a laser down and see if there's ground beneath us if there is ground beneath us then we don't apply gravity because then we will fall through the ground and that's bad but if there isn't any ground beneath us then we're good to switch gravity back on okay here's a clip of the code working on the first attempt honestly if you fell for that that's on you you have not been paying attention but I did manage to get it to work on the second attempt just because it's literally my entire job to make these games that I would be competent at it that you fool but on the third attempt I did manage to get things working no one didn't I I've been your head now you got no idea what to think no I'm just getting this working now I got it okay so now we've got a little steps working we should be able to create the Donkey Kong stage and we'll have our dude run around and [ __ ] it'll be great okay so let's position the floor blocks so they align with the stage and send it oh you bastard [ __ ] you damn it I thought I had it [ __ ] [Music] okay I think we're good I didn't even know what was wrong with that one something to do with when the dude climbs with like two floors at the same time everything shits the pad I don't know it's probably fixed now honestly who knows okay it's floor placement Montage time foreign [Music] Ty let's go brother easy okay sweet so this does work but it does also look like [ __ ] so let's make the floors a bit prettier now I could get the floor texture from the game and apply it to all the squares make sure all the pixels and [ __ ] line up and then I would have to do it for all the ladders and [ __ ] as well or I could just grab a screenshot of the game edit out of the characters and set the floors to be invisible and boom it's done magic okay now we're looking good and you can kind of tell what game this is supposed to be mainly from all the stolen art assets but that's still a win plagiarism more like a later I'll kill myself soon don't worry [ __ ] me that was terrible okay ignoring that thing that happened uh it's a lot of time ladders are actually pretty easy just check if the player is colliding with a ladder and if it is it can move up and down oh yeah or you can just blip out of existence that works too yeah okay real funny yeah yeah it went down instead of up wow okay now you're just taking the piss all right [ __ ] you hey letters are done okay wasn't as easy as I thought it was gonna be but we got there to make the ladders work out and some things for the players to collide with one at the top of the ladder which tells us when the player should get off at the top and another one at the bottom telling you when you can get off there now we just need to copy and paste the ladder colliders and readjust them to fit and the ladders are done you beauty we're getting there okay starting to look really good now but the main dude is still just a square which is a bit [ __ ] uh so let's work on that one art theft later and I got the Run Sprites but because I'm a good boy and Nintendo has lawyers I'm gonna switch them up a little bit [Music] ah yep perfect also it's my first time doing Unity 2D Sprite stuff so there was a little tutorial watching that was cut out that time lapse but we got it it's happening okay next up it's Barrel time the barrels move by going in a single Direction and then if they hit a wall they bounce off and turn around they can also climb down ladders sometimes you know how barrels love to climb down ladders pretty standard Barrel stuff most of the collision and movement logic is actually pretty similar to The Players so I'm just literally gonna copy paste that [ __ ] and change a couple of details so they can move like barrels and yeah not bad time to make them pretty I tried to add some code bullet branding on the barrels but it just looked like [ __ ] so I reverted to the original design because [ __ ] me for trying to be creative I guess all right very nice I also did all the logic on the barrels climbing down ladders and [ __ ] but honestly who the [ __ ] cares they can climb now how they just can stop asking questions next it's time to add our boy the thick lad himself now in order to understand the algorithm behind doggy Kong's Behavior I found a video of the game and analyzed all the animations the frequency of the barrel throws along with the consistency of the throwing window but then I got distracted and watch beams for like an hour and probably forgot everything I learned but after Emily is it gonna get in the next day I was ready and I'll have you know I [ __ ] nailed it yeah this is way better try getting out the ladders now you fat piece of [ __ ] this is dumb I and I got no idea what these Woods are doing but it's very intimidating right well as fun as that was let's get our boy looking right and let's get him to also just settle the [ __ ] down okay here it is ah you gonna throw another one buddy are you just vibing up there okay you just finding out there that's okay you do you bud fair enough okay maybe a little too chill maybe okay here we go if he's finally doing something with his life good for him that makes one of us [Music] so previously when I coded all the player movement speed and the gravity and the jump height and stuff I just I kind of [ __ ] eyeballed it uh so let's get that more spot on I did this by just playing the game a bunch and seeing if my version feels about right it's a very scientific process and now all we need to do is restart the game when the player hits a barrel and add the [ __ ] and we're done baby yo I'm in pretty easy game it's a Donkey Kong it's not [ __ ] Eldon ring but you know still a win and I know there's gonna be a bunch of people in the comments that are gonna be like oh there's other barrels that light on fire and turn into a fire demon thing that follows a player and there's also a bunch of other levels and moving platforms and conveyor belts to different enemies the Donkey Kong guy moves sometimes and [ __ ] and to those people I say ah can't be [ __ ] uh this shit's close enough plus did you miss the part where I was struggling to get the [ __ ] floor to work for like four hours it wasn't a moving platform it wasn't it was the floor so I just I ain't up for that [ __ ] you can watch all the other YouTube videos where someone tries to remake Donkey Kong from scratcher that had three different AIS to it oh wait you can there aren't any you should call the FDC because this shit's a [ __ ] Monopoly baby oh I can do what the [ __ ] I want you're lucky you got barrels that can [ __ ] move would have made the AI [ __ ] a way easier if they couldn't okay sorry now that I'm finished addressing the like four people in my audience that actually cared about that I'll move on and you know what time it is baby [Music] okay so Bachelor number one is genetic algorithm chat gbt do my job for me what is the genetic algorithm the genetic algorithm oh my God I think it's weird that I thought it would be an evolution which can be used to optimize problems optimize problems representing potential Solutions initial populations of these chromosomes and iteratively cool cool recommodation and mutation to evolve towards better Solutions the algorithm operates on the principles are more likely to be selected yeah yeah of the population over Generations the process continues until a satisfactory solution is found or a stopping criteria is reached thanks mate that was [ __ ] okay let's actually explain this the genetic algorithm is simply just human beings playing God and utilizing the power of evolution to solve [ __ ] that we can't be [ __ ] doing by hand in evolution God spits out a bunch of children with different traits some of them have useful traits like Scissorhands or lizard scales or an extra nipple most of these children get in by lines because Nature's [ __ ] scary but some of them survive and pass down their traits to the Next Generation and the Next Generation the process of building on the traits of the previous generation and so they now have flamethrower hands or dragon scales or two extra nipples did that make sense probably not it's just Evolution have you read a book like ever it's just that so the vague flow shot for this algorithm is this create a bunch of random children test out those children by making them play Donkey Kong give each player a score based on how well they did at Donkey Kong in this case that score will be how high the player got next we need to choose which players will be the parents for the Next Generation this is done somewhat randomly using the scores of the players so play with a scorer 2 will be twice as likely to be chosen as a player with a score of one once we have all the parents make copies of them to fill out the Next Generation now if we just run these guys it's not going to really achieve much because they are just perfect copies of the previous generation so they will do just as well as the previous generation which is not what we want we want them to improve so that's why we have the important step of mutation this is where we make small changes to the player's brains so they'll act slightly differently than their parents some of these mutations will make the players worse but some of them will make them better and that's the whole point of the algorithm okay next generation is done and now we need to test them until we're back into the testing phrase and when we we just repeat the whole process test the players evaluate their Fitness function select the parents which did the best mutate their babies and repeat test evaluate select mutate bam that's an algorithm baby let's go now an important thing to explain is I said that we make small changes to the player's brain but I haven't actually described what the player's brain is in this case our brain is simply a list of instructions which tell the players what to do it's no fancy neural network or anything like that it's simply a list of instructions so when we mutate the brain and yes I know brain is the strong word for what we're talking about but you're looking at a guy who can call a single if statement an AI so what do you expect [ __ ] off anyway when we mutate the brain we are simply just changing one of the instructions from left to jump or from right to do nothing and that's pretty much it there are some intricacies that I left out but this is the longest I've gone without telling a dick joke and this is the simplest of the three algorithms so we gotta move on another important thing to know is that we're going to grow the number of instructions as the training continues this way the players can Master the first 10 moves before they move on to the next 10. okay enough nerd talk let's [ __ ] send it oh yeah right I remember I suck at forgot that for a second it really undermines the whole intelligent explanation thing oh well let me spend probably like four hours bug fixing and then we'll resend it uh yeah that sucked but she's working now so let's let's go all right we got the lads going at the start they don't really have many moves to work with so they can pretty much only just go to the right by generation 10 we've got more moves to play with so they figure out uh ladder is up and up is good so we should ladder at this point before the barrels come down it's not that challenging really at all you can pretty much guess what they're gonna do they're gonna go up and they're gonna take the ladders so uh we're gonna Zoom through these Generations pretty quickly by generation 51 they get their first taste of barrel and they do not like it luckily they're making up the ladder just in time to be able to jump over it I'm gonna let you enjoy the time lapse here because frankly I don't know what to talk about just chill out Vibe enjoy the music and watch the guys jump around you'll figure it out I'm sure your monkey brain can find a way to be entertained actually I'm just gonna put this here yeah there you go that's that's for my gen Z audience enjoy [Music] thank you [Music] foreign [Music] [Music] here's a rerun for everyone who was too distracted by the contract actually pay attention to what was going on with the AI but here it is I like this bit it's just like hold old and hope for the best and then you [ __ ] they definitely shouldn't be able to do that that's definitely not how the game should end oh well I'm not I'm not changing that now wow the genetic algorithm is so impressive I I can't believe how good the AIS [ __ ] mute buttons not working hold on do you see them jumping over all the barrels wow they even climbing the ladders oh I found it okay there we go that AI might look pretty but honestly it's pretty [ __ ] and quite basic now let's not discredit all my previous content that uses this algorithm it has its upsides it's fairly simple to program it's very reliable pretty much guaranteeing a successful player and it looks pretty so if you're a lazy YouTuber then it's perfect but there is a big problem at the core of this algorithm and that's the player's brain this [ __ ] is garbage it has no concept of where the barrels are or what a ladder is or anything it really hasn't learned [ __ ] if we change the conditions even a tiny bit and then it completely breaks because it just does the same actions I had to use a fixed random seed for this game because if the barrels were allowed to randomly choose whether or not they climbed down the ladder then everything breaks if you watch the first Barrel in every generation then you can see that it always makes the exact same decision that's because this algorithm can't handle any level of Randomness because then the jump timings would be off and everything gets [ __ ] and it dies again it's not thinking oh there's a barrel in front of me I should jump it's like oh my 43rd instruction is to jump so I jump also if we start the player just like a little bit to the left then it completely [ __ ] up because it's assuming the ladders are going to be there and the barrel timings are all wrong it's yeah it's just a mess okay you get it that's enough [ __ ] on my own content the generic algorithm I give it six out of ten all right it's time for bachelor number two [Music] foreign [Music] you're up again don't [ __ ] it this time need which stands for neuro evolution of augmenting topologies is a type of machine learning algorithm that uses an evolutionary approach to optimize neural networks the idea behind need is to start with a simple neural network okay okay shut the [ __ ] up this sucks I guess I'll have to do my job again that's like twice in one video tragic anyway Nate is also a genetic algorithm so like the previous technique we're going to be utilizing Evolution yeah let's reuse that graphic Evolution but unlike the previous algorithm this bad boy is way more advanced first of all none of that list of instructions stuff that's some [ __ ] [ __ ] this is a big boy algorithm so we're gonna be using a big boy neural network so the players brains are like actually brains which is nice we're moving up in the world if you don't know what neural networks are they're like little magic spider webs you put the inputs in this end and then some magic math [ __ ] happens in the middle and then spits out an output now the sexiest thing about this algorithm is the augmented topologies without this the algorithm would just be net which is way less marketable anyway augmented topologies means that the structure of the neural network can evolve over time most of the time when you're dealing with neural networks you have to define the size and shape of the neural network before you start training it so you just kind of pull a size out your ass and be like yeah this feels right and then just hope for the best Nate doesn't [ __ ] with that instead it starts with a very small neural network which can perform very basic behaviors then over the generations will evolve into a larger Network resulting in more complex behaviors and strategies plus this [ __ ] just looks really cool so that's always a plus that's like the main difference with this algorithm is just swapping out the instructions for a brain but a heap of smaller things that it does to make this actually closer to the evolution that we find in nature it does [ __ ] like pulling players into different species and instead of cloning one parent this algorithm actually combines the brains of two parents to create a child it's actually a pretty dope algorithm and if I explained everything that this bad boy does that will be here all day and I got Netflix to watch so I ain't doing that [ __ ] before we send it we need to Define two things the inputs and the fitness function think of the inputs as just what the players can see too many inputs and the player will be overwhelmed and not be able to achieve anything too little inputs and the player won't be able to do anything because it's just blind as [ __ ] so the inputs for Donkey Kong are the X and Y position of the player whether or not the player is touching the ground whether or not the player is currently climbing a ladder whether or not the player is currently colliding with a ladder the relative positions of the ladders on this level and finally the relative position and velocity of the closest Barrel after a bunch of testing these 10 seem to be the best inputs but I don't know maybe there's some better ones out there but these are the ones you're getting okay the other thing we need to Define is the fitness function which once again will be the height of the player plus adding some bonus points to the player manages to not get hit by a barrel okay and that's pretty much it it's yeah it's pretty much copy paste the same algorithm but just fancy neural network brain thing okay easy let's send it okay first couple of generations as you expect to kind of pretty ass they do figure out if you go to the right then you go up a bit and that's good I guess but still obviously not great but generation five some of them actually figure out how to climb lighters which is a good start that's you're gonna need that most of them just fall off immediately after climbing the ladder because they just go to the right but this guy look at this guy look at him go what a star holy [ __ ] because the movement of the barrels is random now how will the players do really depends on Lucky sometimes they have fairly good attempts getting to the second last level but sometimes they all die immediately so at this point it's really up to chance and how they do [Music] [Music] thank you all right after sending the training for a while now we finally get to this run watch this clean stuff boys clean easy ah finally oh they're so close no no dude that was so close you [ __ ] okay more training [Music] what I know you're there all right so yeah we had the first successful run at level 67 they figured out that I didn't have any Collision detection on the Donkey Kong they just kind of hung out behind him for a bit before deciding to go up the ladder okay let's see how good these guys can get [Music] and these are the guys we got after 178 generations of training they're still doing the thing where they Vibe behind doggy Kong for kind of no [ __ ] reason but apart from that not bad so that was pretty neat shut up shut up sorry um yeah so that was an AI uh to be honest not the most impressive showing I've ever seen but it did its best at the end we got an AI that can sometimes win which ain't bad but there are huge problems with this this is the brain of the AI if you look closely you might notice that there are no connections to this top right node this is the output that is responsible for the jump controls yep that means this AI can literally never jump which is a bold strategy seeing that that's like a massive component of the game pretty sure the main character in this is called jump man that was like the one thing he had going for him now he's literally just a guy so most of its strategies result nervously running away from barrels and praying that they decide to go down the ladders [ __ ] [ __ ] [ __ ] [ __ ] [ __ ] [ __ ] [ __ ] [ __ ] in this defense this sometimes works it sometimes definitely doesn't but it does sometimes work and it was able to get to the top but you know I feel like it's not too much to ask for that the AIS can [ __ ] jump am I being crazy am I the [ __ ] apart from the jumping and to reiterate that is a big problem another problem is the size of the brains that Nick can create they just they're kind of small if you want to have strategies that are a bit more complex than if Barrel near run from Barrel are we gonna need something a bit bigger so to work with bigger brains we're gonna bring in my personal favorite machine learning algorithm yep it's time for bachelor now oh wait uh I give Nate like 7.5 out of 10. I don't know why I'm rating these they all have their own merits none of them could be said to be objectively better that being said it's time for an algorithm that is objectively better okay it's last but not least time for bachelor number three [Music] uh and yeah I'm gonna be completely honest here I don't really know how this bad boy works I know broadly what it is and what it's used for but as for the specifics of what the [ __ ] is going on behind the scenes I've got nothing I've tried to research it but a lot of the information looks something like this and you know life's just too short for this [ __ ] as soon as some letters are wearing little hats then I know I'm out of my depth okay tattoo pity please at this time I actually need your help so come on do not disappoint me proximal policy optimization is a reinforcement learning algorithm that is used to train agents to perform a variety of tasks oh yeah this one's a new and yeah we're just going to ignore everything else he said I'll save you the time it's all technical [ __ ] that means nothing okay so unlike the other two this is a completely different category of machine learning algorithm instead of emulating evolution this bad boy more emulates how a brain learns things think more trial and error than you know making babies we just have our AI [ __ ] around in the world and whenever it does something that we don't want it to do we punch it in the dick and whenever it tries to do something that we like we tickle its balls or something I don't know and then the PPO algorithm takes these rewards and punishments and modifies the brain so it's more likely to get bold tickles than dick punches how does PPO do this brother your guess is as good as mine something to do with proximal policies I suppose but luckily we don't need to understand that to be able to use it because of a little magic thing called other people's code other people have sacrificed their social lives to be able to understand this and I thank them for their service okay so let's talk Donkey Kong once again we're going to need to define the inputs to the neural network PPO is a much more powerful algorithm than neat so it can handle way more inputs when I use PPO for the walkie boys it had something like 224 inputs which is a lot more than 10 which we had for Nate so first of all let's grab the 10 inputs from Nate now the neat AIS only knew where the closest Barrel was but for bbo we're going to tell it where the closest Five Barrels are and what their velocities are so that's an extra 16 inputs with a total of 26. nothing compared to the 224 for the walkie boys but it's still a few we also have to define a bunch of things called hyper parameters which set things like how long we're going to train for and learning rate and the amount of random actions the player should do and stuff like that so I'll just do that yeah don't worry about it and also we need to define the size of the brain it big okay great fair enough now for the reward functions this is when we're giving them the Bold tickles and the dick punches I'll reward them for whenever they are gaining height jumping doesn't count towards that otherwise the players will just jump forever racking up that mad dopamine we don't want that so the height has to be from climbing or going up the stairs and they'll also get a fat reward if they reach the bridge at the top as for punishments I will punish them whenever the players are hit by a barrel or if they go down at all we don't want that [ __ ] uh and yeah that's pretty much it there's so much more I could be talking about which is why I generally don't like to explain things because it [ __ ] takes ages but that'll do you get the gist okay let's send it all right we got the lights going we got like 50 or so fellas running at the same time now they do all share the same brain think of it like kind of like a hive mind situation so all the lads are out trying to get experiences and then they send those experiences back to the main brain which then updates itself and then controls all the players although you notice some of them are like falling down the ladder um for the life of me I've got no [ __ ] idea what's going on there oh and so instead of actually fixing that problem I just penalized the [ __ ] out of them if they fall and so hopefully they should learn to not do that uh so instead of actually fixing the problem I'll just trade the AI to not do that um yeah absolutely surely they are making progress same as Nate how well the players do is very luck based like if the barrels take as many ladders as they can and just b-liner for the players it pretty much kills them in suddenly uh but they're learning to avoid that and if you look closely they are actually jumping which is fantastic didn't think that'd be something I'd be celebrating so much but here we are foreign [Music] so we're like 25 minutes into the training the lights have been stuck in like the second level for ages this is mainly because you can get to the second level without really worrying about the barrels because it takes them some time to reach that point But as soon as you get to the third level you start to experience the barrels so it's at this point where the players actually have to learn how to jump over the barrels and stuff which is like way harder than learning how to just climb ladders and so for a while they just decided it wasn't worth it it was easy just staying on the second level but now they've kind of started to figure out how to avoid the barrels to some degree so they are actually making it further yeah sorry about the frame rate that's just gonna happen but look at this look at them ready hey okay I'll just one guy went to the right not the play but one guy made it but even though we've got one guy to the top once that doesn't mean that the years learned how to do it well you can see still most of the players are dying sometimes they are reaching the top but it's only like one or two players so hopefully we can get to a point where it's way more consistent but they actually reached the top fairly quickly that was only like 30 minutes of training which normally for PPO it can take [ __ ] ages so that's really good all right let's crank it for like another half an hour and see how good they get crank it was a weird choice of words I I meant like just run it crank it sounds like you jacket it all right let's check it for another half an hour and see what happens [Music] all right we've successfully cranked it um but yeah look at them go there's actually some times where they do all die but most of the time most of them get to the top which is pretty good I tried running it for many more cranks but it didn't really improve the play as much this is about as good as they get now I know that was hard to see because it was sped up as [ __ ] so here it is at Double speed and as you can see they're doing pretty well they're actually doing some pretty dope [ __ ] he's like waiting for battles to pass before climbing ladders and he's successfully jumping over them it's pretty good stuff they seem to [ __ ] up at the last stage a lot I think that's just because the barrel appears out of nowhere so it's just like a jump scare and then they die to be fair I probably should have told them how long until Donkey Kong throws the next battle but is what it is we're past that point now I was actually somewhat disappointed with how well they did this they did fairly well but it took me so long to get this working and I figured if I can fly a plane and walk a human then it should crush Donkey Kong right hey why don't you just change the rewards a little bit maybe that'll help okay okay shut up let's go behind the scenes a little bit and give you a glimpse of what it's like to use this algorithm it probably looks like this was my first attempt at running the algorithm because editing but it's probably closer to my 20th here's some things that I tried to see if they would improve the algorithm bigger neural network smaller neural networks normalizing inputs requesting decisions manually seeing less barrels seeing more barrels Higher Learning rate longer training lower learning rate higher reward for climbing higher penalty for death penalty for just chilling in the corner because they [ __ ] love that [ __ ] for some reason change the learning rate algorithm higher time Horizon smaller buffer size and more players so those are some of the things I tried and each experiment is like a couple of hours so yeah I [ __ ] tried all right there's some more things that I could have tried like feeding it in a picture of the screen raycast inputs coin-based rewards rewards for jumping over barrels or incremental learning by slowly increasing the difficulty all these things I could do but yeah I've already done two whole other algorithms of this video so I'm pretty happy with calling it there and I think this fairly well illustrates a big annoyance with this algorithm and that's just how many things you have to tweak to get it to work it's a [ __ ] and it takes ages so because of that I give PPO an 8 out of 10. it's still a banger when you get it to work but getting it to work is a pain I went into way more of the technical side of things in this video kind of to counteract the last video we don't talk about the last video but if you enjoyed learning a bit more about the technical side but behind the AI then you're going to love the sponsor of this video brilliant.org friend.org is a dope way to learn maths and computer science interactively if you're interested in neural network stuff and you're here so you are then oh boy are you going to be eating well with the introduction to neural networks course just look at some of this Banger content we've got interactive neural networks that show how the connections and neurons work together to predict which shape you're drawing oh baby that's the stuff and they've got a whole lesson walking you through how this bad boy works it's also got lessons in how we use math to replicate the activation of a neuron and it's got lessons of why we have a bunch of hidden layers that's these guys in the middle of the neural network and they've got so much stuff they've got concept for days with literally thousands of lessons to pick from so what are you waiting for I don't know I just wish there was a way we could experience everything brilliant.org has to offer for free for like a full 30 days I have good news for you because you can experience everything brilliant.org has to offer for free for a full 30 days oh wow yeah that's uh yeah that's what I wanted okay cool you just have to go to boone.org code bullet to get the 30-day free trial plus the first 200 people to sign up will get 20 off their annual subscription seriously it's all good stuff so go check it out thanks for watching that was actually a pretty fun video it was a [ __ ] getting all the AIS to work for some reason I didn't think about the fact that three AIS meant three times the amount of work but it all came together well uh yeah oh I should shout out my second Channel I actually should have done that at the start of this video but you know I'm bad at my job I'm doing a challenge to see how long I can do weekly uploads there for and so far I've survived for three weeks which is Against All Odds at the moment most of my videos are chat jbt related although I swear that's just a coincidence I'll do other [ __ ] so if you want to see more code block content then check it out oh yeah that's pretty much it thanks for watching and I'll see you guys next time later

Channel: Code Bullet

Views: 3,218,480

Keywords: artificial, intelligence, coding, processing, java, machine, learning, genetic, algorithm, evolution, computer, science, programming, comedy, educational

Id: ovIykchkW5I

Length: 29min 45sec (1785 seconds)

Published: Sat Feb 25 2023