Blum masterfully bridges the gap between real-time infrastructure and neural decision-making, proving that a robust data pipeline is just as critical as the model itself. It’s a pragmatic blueprint for turning raw telemetry into intelligent action through the lens of reinforcement learning.
Install our extension to search inside any video instantly.
Teaching AI to Play Game: Part 2
Added:How's it going everyone in? How's it going? Happy Wednesday. We're going to be continuing with our re reinforcement learning. That's what we're doing. We're doing some reinforcement learning. We're going to figure it out. I really want to figure out how we combine the reward with the loss. I don't know if we'll get to that today. We do have a few like, you know, homework items that we have to take care of before we can actually get [music] the system working here. So, let's zoom on this so we get the right size there. Does it look good right about there? Yeah, I think that's pretty good. All right, let's center the windows. Okay, so what we need to do is we need to extract the state of the game so that way we can capture it in our Python process. Let's open up the game really quick so that way we can see it running. Open index html. Double check that I've got this here.
See real quick. Open. There we go. Okay, so we got the game. It's running. It's doing its own thing. Now I need to export the data. We got to export the data.
Let's see if we can make that happen. I think we can. So this is the game speed.
And then we need to see what like the the jump. We need to do a couple other things. We need the jump. We need uh obstacles.
We need to see things like uh are these the only thing that we really need? Is there something else we need? Like we need to know where we are in the air. We need to know basic Oh, if the game is over, right? game over. Something like that, right? I think that'll be something that we need to take care of. Uh and then cuz like if you lose, that needs to be a penalty, right? Not a reward, but a punishment. Hey, there we go. A punishment. That's what we're looking for. Uh game over. Gets like a -10 score. If you jump, you get like a plus one. And if you make it for the time, right? the time like how many steps we like to sequence you get a a plus 0.1 or something like that so that way you get an increased score as time goes on right so those are the kinds of things that we need so let's see if we can export that data I would like to capture it over here in the terminal so let's open up this here hello how's it going there the KVC welcome on in good to have you here. His python. Yep, we're doing little Python.
Yep, we're going to do it. You remember?
Good to have you here. Welcome on in.
Thank for the hundred. Thank you very much. Appreciate it. Thank for the hearts.
We're going to be building a reinforcement learning today is what we're doing. We're doing reinforcement learning. Thank you so much for the hearts. Appreciate it. [laughter] All right. Good to have you here. Uh let's see. So, what we need to do is we need a bash script next, right? We get some bash scripting going here. So let's do that vi what do you want to call this um like receiver or game monitor monitor.sh.
All right. So let's do bin bin zsh.
And we need to be able to capture the data from the subscription. Let me see here. Where is this pub knob?
Okay. So we need to grab that data. I'd like it if we did the stream API. I think that would be a lot of fun. We grab the stream data immediately and then we can just loop on that.
I think that's a good idea. So, I have that code somewhere. Oh, in fact, I kind of have it already. I have it right here. So, let's see if we can open that up. PubNub. Here we go. Stream. So, this is the answer right here. Right there is the answer. Yo, how's it going, the Los Pros? Yo, how's it good to see you?
Welcome on in. Do we need a [laughter] paleontologist?
Ross from Friends, right? I remember that show, Ross. Rachel.
Uh, the paleontologist. Yes. Hey, good idea. I never even thought of that. I see a dinosaur going nuts. Yeah. This is that game, you know, the the um the game that you have from Chrome when Chrome is like out of uh the um it's like this is hard mode. I'm I'm forcing it to go further here. So, the game is Oh, hey.
How's it going there, Nina? Ola. What are plans for today? May I ask? We're going to extract the state of the game so that way we can start putting that into Python. We're going to do game extraction.
That's what we're going to do. The big brother. Hey, how's it going there, Tubercio? Good to see you. Welcome on in. Happy Wednesday. So, if I grab this code here and I bring it down here, we've got our URL.
And what is the channel name? I think the channel name is this right here.
Right. Let's grab that really quick.
Here we go. Whoop. Here we go. Paste that in here. All right. So, game channel. Here we go. Let's put this in all caps. Game channel equals T-Rex.
There we go. Looking good. Looking good.
In one line. What are we doing today?
Reinforcement learning. Making the AI learn how to play the video game. That's what we're doing today. Hey, how's it going there, Bungie? Welcome on in. Come on back. You know what that means? It means Bungie is here. That's what it means. [laughter] It means it's Bungie. Good to have you here. Oh, a normal question to people as I sit down with AI now. Oh, yes. We're Yep. This is going to be reinforcement learning. We're doing reinforcement learning, you guys. I'm so excited about it. I mean, I can start coding. Oh, that's what it means. Oh, you can start coding. All right. Good to hear it. Good to hear it. Looking good. Looking good.
Hey, Gloomy. Good to see you. Welcome on in. Happy Wednesday. Finally understand Rust. Yo, it all only took 37 compiler errors and complete personality rewrite.
Isn't Rust a little bit extra? It is.
You're like, "Wait a second. Hold on.
What is this Rust business?" Oh, it's something completely new. It's completely new. Hey, how's it going there? Mia Rose love. Thank you. Good to see you as well. Happy Wednesday. Hey, how's it going there Jaden? Jaden Stalber Vlogs, good to see you. Welcome on in. Happy Thursday. Wednesday. Good to have you here. Mia Rose, hey, thank you for the hearts. Appreciate it. All right, so let's see. Our URI is going to equal, you know, we're just going to say channel. Well, you know, we could say game channel. It's fine. All right. URI HPS origin. We need to grab the origin.
And the origin is going to be around here somewhere. Where's that origin at?
Around here. Origin is here.
Where is here? There's a default origin around here somewhere. There it is.
Let's grab that. Here we go. Origin.
Origin. Origin.
Delete this. We're going to capitalize that.
Organize this. There we go. Perfect.
Excellent.
There we go. There we go. All right.
Hey, how's it going there, Zippy? Hey, Zippy. What is this? We are building a an a reinforcement learning AI that's going to be able to play the game. We're going to build it. Hey, how's it going there, Schubam? Good to see you. You're here. Welcome on in. You love the challenge yourself. Rust is great, isn't it? It's a really good challenge. Rust is a good challenge. Mia Rose, are you single? Yes. Hey, KVC. Hello. Hello.
Hello, guys. All right. Good to have you guys here. KVC, thank you saying no personal questions. [laughter] Hey. Hey. Hey. Uh uh. Hey. Are there any social security numbers? Exactly. Any social security numbers here? Can we have pull? Uh pull for what? Yeah. Yes.
Yes. Rust or Python? Uh Rust. Usually Python today. Today we're doing Python and Bash and JavaScript. Yo, Bonzupi, good to see you. Yo, Squirrel Sniper. Hey, I saw that you uh you contacted us.
Squirrel Sniper. Uh, are you in the UK?
Squirrel sniper. Squirrel sniper in UK.
>> Hey, there we go. Got some bash going.
You got it. You got it. You got it. Bash going there.
>> There we go. Next song. Next song. Hey, Gunheis. Yo, good to see you. Someone flirting with you on YouTube. I know, right? Gloomy traffic. I know.
[laughter] Yeah.
Yes. Almost 4 p.m. here. Hey. Yeah, I saw I saw you did a thing. You did a thing. Now, can we do let's see here. We need to do double quotes on this. All right. Origin stream subkey. And we need the sub key is going to be Here we go.
Grab that. Subkey equals demo. All right. Almost there. Almost there. Let's see. Oh, hey. Is this JavaScript? Yes, we got JavaScript here on the left and then we've got Bash over here on the right. a little bit of bash or bash.
It's both. It's both, you guys. Uh, JavaScript commands or anything, but it looks like fun once you learn it. Hey, Jaden. It is pretty good. It's pretty good. Hey, Rebellion Ant. Good to see you as well. Happy Wednesday. Good to have you here. Yep. You got some Python.
We We'll be adding in Python here in a bit. So, we're doing three. Auto clicker 11. Yeah, it's almost autot clickery.
It's almost right. Almost almost.
[music] You too. Nice. Thank you. Hey, Cardinal. Good to see you. Er, welcome on in. Happy Wednesday. I'm single, too.
Do you want to be [laughter] I see I see what you said there, Mia. I see what you said. Hey, 11. How's it going? Hey, by the way. Excellent. What's the goal? Oh, so the goal is to train the AI to learn how to play the game really well. That's what we're going to do today. Hey Vicki, how are you been? I've been doing good.
Thank you for asking. Good to guys have you here. This is great. Thank you for saying hi. Yeah, we're going to be building some stuff here. Yo, loose juice. Hey, loose juice. You remind me of Spruce Goose. Loose Juice. Spruce Goose from Adventure Time. From Adventure Time. Now, is this game channel? We want to call this game channel. We called it game channel, right? All right. So, this be game channel here. And then we'll need to grab the time token. Let's see here.
Grab the time token. Set this equal to zero to start. Okay. And then we're going to have to loop it. while true do echo. Do we want to echo? Yeah, I suppose we now let's just do e echo.
Yeah, we need to set a URI equal to a thing. So, we could set that here. And we say uh done.
Grab the done there. Okay. All right.
All right. Uh I have a question two.
Will you be my personal mentor? I'm happy to help. Just uh hit me up on Discord. You're worried you're going to be a victim of Nigerian love scam. Hey, don't worry. We're good. We're good.
Hey, Gloomy. We'll be fine. We'll be fine. How's it going there, Winston?
Hey, Winston. Good to have you here.
[music] Welcome on in. I see what you said there. People who say hi. Uh, wait.
People who say ah are toddlers. Oh, wait. Ah, Kirby. Hey, good Kirby Fox.
How's it going? Good to see you, Kirby.
Interesting. Machine learning. Yes, exactly. It's exactly what we're doing.
You got it. Machine learning. AI is playing like a toddler right now. Yeah, it is. It's not very good at this point because it doesn't have any It doesn't know what's going on. It doesn't know.
We're going to make it better. It's going to take us a few days. It's going to take us a few days, you guys. It'll take a little while. A little while. My Jaden, my day has been good. Thank you for asking. The jumping all the time right now must be something gets fluff fulfilled. Yes. Waiting for the AI to learn. Right now, we don't. It's not really learning. It's just we're we're capturing data so we can do some data [music] research, right? We want to see what it looks like. How much have you earned so far on this platform? Uh, you know, an amount, [laughter] you know, some some amount there. A good amount.
Gaming with GL. Hey, how's it going?
Good to have you here. Do you play Call of Duty? I have played it before. I [music] have. Cood. Cood. You guys sing.
You want to be my boyfriend? Yes or no?
Mia Rose. Hey, Mia. I mean, I don't know. I have no idea. I can't answer that question. I don't know. Days might sub then. Yeah, it's going to take a while. It's going to take us a while to do. Yeah. Yes or no? Hey, I don't know.
I don't know what the answer is. I can't answer yes or no. Okay, you played too.
Nice. Excellent. Excellent. Good to hear. You sent a friend request. Oh, nice. Okay, cool. Excellent. Shuab, thank you. Thank you. Thank you. Thank you, Mr. Blockies. Hey, good to see you.
Welcome on in. Happy Wednesday.
You Oh, this is so cool, you guys. Mia seems to be giving AI prompt vibes. AI prompt vibes. A little bit of Mia. I think Mia is a real person. I think Mia is real. I'm pretty sure Mia is real.
Getting dates while I'm fixing bugs, [laughter] you guys. Hey, Mac attack. Good to see you. What kind of setup do you have?
This is a Mac. We're We've got a the game at the top of the screen. Yo, Ryan, cool. Hey, Ryan. Cool. Thank you for subscribing and joining the right channel for software engineering. We're currently building a reinforcement learning AI model that's going to learn how to play this game. It's not very good right now because it's not really learning anything, but it will. It will.
Is it difficult to code? Yeah, it is a little bit difficult. Yes, just a smidgen difficult. So, we need to grab this. We need to do a curl. We'll do a curl on that. And then we need we we're just we just want to grab the date is what we want to do. So, you know what?
We don't even need to worry about the time token. We could just say like 1,00 yo. Hey, Loose Juice with the silver star. Thank you very much. My gratitude to you. Good to have you here. Oh, you guys are great. [music] Where do you learn this? Uh, we're learning it right now. Yo, Rio. Hey, Rio. Thank you for subscribing as well. All right. Good day, you guys. Good day, bro. You test a real person. You [music] testing a real person. Uh, well, I don't know. What do you mean? Yo, and a laughing disco ball.
All right. Thank you very much from 20 23 Coach Martin. Yo 23 Coach Martin. Yo, thank you very much. My gratitudes to you. Thank you for the laughing disco.
You guys are great. You guys are great.
Love you, man. Ray, Ricky, Roland, thank you. Thank you very much. Appreciate it.
All right. You told sub I get interested. All right. Good to hear it.
Thank you, Rio.
Rio, thank you. All right, you guys.
Man, cactus. What do you mean? Uh, does it take experiments to learn? Yes, it does. It takes trial and error. That's the best way to learn. That's the best way to learn. You got it. Do you have any pets? I used to have cats. I used to have dogs. I used to be the thing. I don't have those anymore. I don't currently have them. Does it take Yeah.
So, we're we're good to go. We are good to go, you guys. Mhm. H. Yep. We're going to do some Python. You got it. A little bit of Python here. All right.
So, let's get back into it over here.
So, I want to get the output from this.
I wonder if this is working. Is this part working here? Do you think it is?
I'm pretty sure it is. I think it is.
Argentina lost the World Cup. I heard.
Hey, Ankit. I know. I heard. That's crazy, right? I mean, I heard the sports were over. I heard they're over. Hey, John. John. Hey. I see what you said there. I see what you said. Good to have you here, John. Thank you.
What's your highest score on the Gino game? Uh, I Well, my own personal score has been pretty high, right? I can I can I can slightly control it.
This is it's it's like hard mode right now cuz the AI is jumping crazy and so in order for me to like manipulate it, I have to like do extra Oh, John. John, you didn't answer. Uh, I did. I know I didn't answer. [laughter] No, on purpose. On purpose. I didn't answer.
Uh, I see what you said though. I see what you said. I see your question.
Please ask Python questions. [laughter] Yeah, right. Exactly. Ask Python questions. All right. So, let's see if we can get the data to stream over to us now. Yeah. All right. Curl game token. I do believe I do believe this probably will work here. Here. See here. Q. All right. Let's try Let's try it here. All right. So get add get status get commit added the monitor added monitor monitor there we go push to main branch there we go have a spine yeah I know right John John I see what you say my name is or mine is 6867 wait what do you mean what do you mean thank for the party poppers you guys appreciate it so yeah Winston my high score is probably like a few thousand maybe at the at the best. We're all a religion of rust. There we go. Hey, get some rust in there. Yeah, I see what you're doing.
Why is it constantly jumping and not strategically jumping? Uh because it's not very smart right now. It's not very smart. So, what is it? Uh set set interval right here. We're having it jump. We're having it jump here. So, let's see if we can Let's see if we can jump it more frequently. See? Is it going to jump even more frequently?
Yeah, there it goes. It's jumping, jumping, jumping. All right. Is it going to work better or worse? It seems to be slightly better. Oh, it's jumping better. It made it further that time.
[music] Learn Spanish, by the way. Do you think it's best for traveling after English or should I learn some other language? I think Spanish is pretty popular. It's a It's a It's a world language, right?
Your personal score is 6867.
Oh, got it. Okay. So, what we could do is we could hide this, right? We could turn this off here and then it won't be jumping automatically anymore. And so, now I have to jump it automatically.
Hey, Peter. Good to see you. Welcome on in. How long have you been coding? For over 25 years. Uh, what are you doing now? Uh, we are going to be playing this game and we're going to train the AI to learn to play the game better than a human can. Right now, I'm trying to play while I'm reading your chat. So, everyone should like the video. Hey, Winston. Thank you, Peter. Hello. Oh, I jumped on into the pterodactyl. Into the pterodactyl. Is it hard to code? Yes, it is though. You get used to it after a while. You get used to it. Answer me.
Oh. Uh, Vyx. Vyrex. What did you say?
Vyrex, where did you say that? I don't see your message. Uh, it might not have shown up. Sometimes it doesn't show up.
Vyx, you have to say it again. If you post a link, the link won't show up either. The link won't show up. It won't show up if you do that. Sent a line adds delay based on how it seems. Yes, detects right as far. So that is that is just to get the game to you know start giving us data. So what we need to do right now is we need to capture all the data that's in the game and then we need to pass it into Python script. Live subtitles. Yeah. Yes. Live subtitles.
You got it right there. These are called open captions. That means that I control them and they're burned into the video so you can't turn them off.
They're they're they're fixed. How will you manage the progressive acceleration of the game? We're going to include the speed. We're going to include the speed.
Do we have that? Hey. Hi. I have a calico. Good to see you. Welcome on back. Good to have you here. I have a calico. Thank you for the silver star.
Welcome on in. It's working. I mean, it was because that I'm controlling it.
This is me. This is me controlling it right now. How do you have such an accurate speech detection model? Oh. Oh.
So if you want I'll show you there is a GitHub repository here if you want. This is code that I wr written many years ago you guys. This many years ago eight years ago. Yo Kurt. Hey welcome on in Kurt. Good to have you here.
So accurate speech. So if you want it's on our discord link if you want to join on discord over here. Over here on our Discord, we've got a link shared channel where you can find all the code that you need to [music] have your live subtitles. Here you go. Right here.
Right there you guys. Hey, QuickFix.
Good to see you. Welcome on in. Good to have you here. Happy Wednesday.
What's my high score? I So, my high score is, you know, maybe maybe like I don't know, maybe a couple hundred.
Really? Probably a couple hundred.
That's what it is. Some hearts for you.
Hey, it's good KVC. Thank you very much.
I'm no IT guy. Go. What? What's going on? Curious. Hey. All right. So, we're building a reinforcement learning AI model that's going to learn how to play this game. It's going to learn how to play the game and it's going to learn how to better than we are doing it.
Thank for the hearts. I was just thinking of donating when some streamers answer me, but nobody passed. Oh, CD.
Really? CD Kraken. CD Kraken. They don't answer you. That's terrible. They should be like, that's what that's why you guys are here. That's what you're here for.
You're here for that right now. So, yeah. If you want the subtitles here, you think I had 1,000? Nice. Yeah, we've Our current high score is 259.
That's just today. 259 just today. So, what we need to do is we need to extract the data from this game so that way we can capture it in our Python AI model.
So, basically making things different in the whole game like making the dinosaur fly. Well, maybe. I mean, we're going to find out what the AI is capable of. Is it going to make it fly? I don't know.
It might figure something out that we've never been able to see ourselves.
Whoops. Just checking on you, Stephen.
How's it going? It's going good. See, thank you for asking. Something I just realized the game gets progressively faster. Hopefully, the AI will yet'll be able to work with that. Yes, we're going to include that as the training feature data. So, it'll be a feature. Batman or Superman? Well, I mean, uh, they're [music] both good stories. Superman is kind of a lot of fun, right? Cuz you kind of like the invincibility, right?
other than kryptonite. Batman uh has to struggle and is in pain all the time, right? Dark Knight, right? So, there's a fun story with that. Fun story.
Can it inject JS? We don't need it to.
Kurt said, "Hi." Thank you, Kurt. I've done a few thousand sometimes. School.
You play until you got a board. All right. You can detect the words kryptonite.
You got it. It's pretty good. Yeah.
accurately classifying and responding to different obstacle types. Yes. Distinguishing them and flying and projects requiring ducking. Yep. There's going to be a ducking as well. So, we're going to have to do ducking, right? As well. Ducking as well. Yep.
Yep. So, there's a lot of stuff. Your channel sounds cool. Hey, thank you very much. Does it also improve timing when it jumps? Yes. Uh, do you like tennis?
Oh, right. The sport tennis. Yeah. Yeah.
I've seen it. I've seen it. I've used to play tennis a little bit. a smidgen. Add speed and state as well. Yep. So, we're going to add the speed and state here.
Let me go where is this at here. So, let's get the game to auto jump again for us. We'll have it auto jump there.
Okay. Uh and then let's do forward slashpublish.
And then we need to add game speed. Here we go. Game speed. Excellent. What's the data set? So, we're capturing Oh, we already have the game speed. We already That's like the one thing that we do have currently right now. That's one of the That's one of the features. All right. So, game speed, current time step, jump is Are we jumping right now?
Obstacles, and then game over is what we need if we lose, right? If we lose. So, that's what we have to take care of the data set. There you go. There you go. All right, you guys. Thank you for good questions. Preventing catastrophic forgetting or sub-optimal policy converg. So I I Hey, squirrel sniper, if you have a good idea, let me know what the best model is going to be for this for this model to learn. Whoa, this auto this jumping is doing really good now.
It's just jumping. Look how far it got.
Oh, it just loaded. Hey, how's it going there, Astro Sky? Good to see you. You just joined. What's happening? We're building an AI model that's going to learn how to play the T-Rex dino game here. The T-Rex running game. So, we're going to capture all the data. And so we don't know what that looks like yet, but we need to capture it here. So let's see. Uh chmod plus x monitor monitor.
Let's see. All right. So we got a problem there. Euro rejected port. Okay.
Oh, because do we have the Okay. Vi monitor. All right. So here's the origin. All right. Looks good there.
Channel sub key time token. That should act. Oh. Oh, I see. URI curl uh money URI. There we go. Let's see if we can get this going.
Monitor. There we go. Monitor.
Excellent. All right. Hey, look at that.
Look at that. We got the speed. We got the speed. Check it out. See how the speed's slightly increasing? We got the data. There we go, you guys. We got the data right here. So, now we can start training. Hey, there we go. We can flood the channel. How will you figure out the acceleration? Uh, it's going to be a feature, right? It's a feature. So, we need to get the max speed and then we need to divide it so that way we have like, you know, a scale factor. Hey, Tugra. Good to see you. Welcome on in.
Happy Wednesday.
Welcome. Welcome. Let's see. Smart people. Hey, how's it going there, Chimp? How's it going there, Chimpalooa?
Chimpalooa. Chimpu Limpu. Chimpu Limpu.
There we go. I like your name, Chimpu.
Let's see. Uh, no matter how much money we have, can you cannot buy enter into the contract? Wait, Rudra, what are you saying?
DQN. Okay. With experience replay. Okay.
DQN is what you're saying. All right.
So, we need to grab the DQ in here.
Let's do that. All right. Read me. All right. So, we're going to do a [music] DQN with experience replay. All right.
That's the answer right there. There we go. There we go, you guys.
Mia, thank you. Mia, appreciate you. Is that true? My friend Nal Nalicken Radre Radri, Radre. [music] What are you saying? No matter how much money you have, you cannot buy or enter into that contract. What's the contract? That's my question. Uh oh, wait, wait, wait. Uh uh. Oh, see at Vyrex.
Hold on, let me see. So, uh Vy, what is the question? What is the question? Oh, Ricky. Uh, would it be easier if it ran?
Uh, it it's running as fast as it can go. How long do you plan to stream because I have stuff I need to do and I want to come back? Oh, Chef, will we be here for about an hour? We're here for another hour. Another hour. Mia Rose, thank you for all the reaction the the emojis there.
What do you do for channel? This is a software engineering channel where we code live on screen and we do all the things that we need to do in order to achieve our objective and we're today we're trying to train the AI to learn how to play the game really well.
Favorite animal? Cat. I would say cat.
That's an easy one. Cat. There you go.
No one. Hey, no one. Thank you for subscribing. Your name. Your name. Good to have you here. You joined the right channel for software engineering. Thank you very much. I'm glad to have you here. Would you rather gain muscle or gain eagers? What's What are eagers? I don't know. I think muscle. What are eagers? H. Yep. We're going to do some Python. You got it. Only auto jump. What about auto run and auto duck? Yes, we have to do that as well. So, we've got a lot of things that we need to take care of. So, let's grab the current time the kind time [music] stuff. Yeah. Yo, Sergio is the coolest coder.
>> Hey, thank you. Thank you very much for the dono. Oh, I appreciate it very much.
Thank you, Sergio. Thank you.
Exclamation point. Party poppers. Oh, Sergio. Thank you. Oh, yep. Didn't need didn't mean to do that. There we go.
Thank you, Sergio. [laughter] All right.
Hey, Brad. Say it fast. Uh, which which say it fast. Where is that at? Uh, would you rather gain muscle or gain eagers? I don't I don't I don't see what it means.
I don't see it. I think I understand it, though. I think you're trying to get me to say something. I see you trying to get me to say something fast.
Don't say it. All right. Yeah, sis me. I know. [laughter] Don't say it. It's uh Yeah, I figured. I figured I figured. Thank you. Saying really fast. Yeah, Brad, I see what you're trying to do there. I see it now.
I figured it out. You're trying to get me to trip. You're trying to get me to trip there. Yo, how's it going there, Brad? Hey, Bradbury. Thank you for subscribing. Good to have you here. You joined the right channel for software engineering. You did. This is where we build code. We're going to do a lot of AI. So good. So good. All right, let's get the next song going, you guys. Here we go. Yeah, there we go. This one. I like this song. This is a This is a bopper. It's a good one.
See? Better electronics or computer science? Which is better? Electronics or computer science? Let's ask the chat, you guys. Let's ask the chat. Thank you.
For those of you who are new here, my name is Steven Blum. I am the CTO at PubNub. Let me see if I can grab some uh details for you. My name is Steven Blum, CTO at PubNub. We are an a communications API provider where [music] you have in-app communication.
We raised 134 million series B. We have over a billion devices on my network and we are the largest real-time infrastructure provider for multiplayer games, for tele medicine. Uh we are number one in the world. We're the biggest. We are we're number one, you guys.
What does it look like? Cactus. Oh, cactus. I get it now. Oh, it took me a while to figure that out. Lined up perfect for every jump in the dinosaur doll. I know, right? It's got to be perfect. It's got to be perfect. Oh, don't worry. Don't worry. We're going to train it to learn a lot better. It's currently just basically RNG at this point. AI fully developed artificial intelligence supposedly legitimate mindsets claim to accuse the worldwide poisoning of a threat of to security of the system. Wait, in India and the UK.
Well, I mean, there's a smidgen of that, right? We do have to be careful. We do have to be careful. All right, you guys.
All right, so that dino must have strong thigh muscles. It's Yeah, by now it does. It sure does by now. Hire me. The the KBC, we do have open rolls available, you guys. If we scroll over to here, where is it at? Where is our Here we go. All right, here. So, you can do that. Let me see. If you scroll down to careers, where is that? Careers. Here we go. There it is. Click on that. Let me get our jobs. Here we go. Here's our available jobs currently listed, you guys. If you are interested, cactus.
Hey, there we go. The cactus. You can also press space longer to jump farther, right? Can I?
Let's find out. Let's find out. Let's find out. All right. So, set interval.
Uh, where is that at? Here. Is this it?
No. Right here. Here we go. Right. Yes.
Let me double check. Yeah. Okay. See?
So, reload. All right. So, you say jump.
Uh, it seems like Oh, yes, you can. You can. I didn't even notice that. Whoa.
And it can jump longer. Yeah, you're right. You're right. I didn't even notice that. Okay. So, that is something that we're going to have to learn as well. We're going to have to learn that as well.
Uh, I asked you earlier how much money you earned. I didn't expect it to be that much. [laughter] I know that's just that's just venture capital. That does not count as as my salary.
It doesn't count as my salary. You guys, you guys are funny. You guys are funny.
All right. Global warming. No. Hey. Hey, Rodri. It's actually misinformation.
Global warming. If you eat meat, you are 100,000 times more pollutive. and global warming contributor. If you eat meat, cattle farming and and and sort of like in general like farming with animals is about 10 to 20% of the world's pollution. Data centers, not AI data centers, not just AI data centers, all data centers is only.1% of the planet's resources. So, eat me. If you want to pollute the planet, you guys, that's how that works. Uh, let's see. I asked uh how much is your typing speed best? Uh I like 60 to 70 words per minute. Uh or no words? No, not yeah 60 to 70 word. Maybe 60 70. We'll have to try it up again.
What about fish? Fish is better because fish isn't >> [music] >> um very pollutive to the planet.
Definitely subscribing. Hey. All right.
Good to hear it. Thank you. The AI did the evaluation for you. Uh, so the AI itself, you need to compare it with the other things that are pollutive on the planet to get a comparison, right? Thank you for the hearts, you guys. If you can see, if you can't see my message, then turn on live chat. You subbed. Hey. All right. All right. Thank you for subscribing, Vyrex. Thank you. Why is eating meat pollutive to the planet? Uh, because farming process, the farming process uh is wasteful. It has takes a lot of water. a lot of water and also there's a lot of methane produced and also carbon for all of the shipping and processing of the of the animals and also you have to feed the animals, right? You have to feed the animals grain. 90% of the calories are lost in the conversion to the animal meat. So if you just eat the grain instead, you would save 90% on that. So that's that's kind of how that works. What's the channel about? Uh you know, we are a software engineering channel. We are currently building a game. Well, building an AI that can learn how to uh play the game basically. You're being summoned. Catch you upload smiling. All right, Gloomy. Thank you very much.
Thank you. Thank you. Thank you. Are we considering immediate objects height?
Yes. Yes, we are. How is it's how people lived? Uh Tyler. Uh yeah, I know. It's up to you guys. It's up to you and the people. How much you want to pollute the planet is up to you. And if you choose to eat meat, you are the biggest polluters. And that's okay. That's that's up to you. I'm not I'm not judging you for it. I'm just saying like if you you're saying like data centers are the problem. That's not really the problem. The problem is eating meat. If you think about it, look at your eyes though, man. Oh, I know. Wait, what do you mean about what what do you mean?
What are you talking What are you talking about? What do you mean?
Uh who did you root for? Wait one second. One second. Who did you root for in the World Cup? Um, who was that? Who was that I wanted? Uh, there was there was there was a country I rooted for.
There was a country. I'll have to remember it here in a second. You hunt.
You see, hunting is better. Hey, Tyler, hunting is better because you're not cattle farming. You're not polluting.
Animals uh are are if you hunt, that's better. All right, we are getting distracted. Let's focus back. I know the KBC. We are very distracted. We are very distracted. All right. So now we've got the data. I would like to export this.
All right. So curl. Let's grab Well, we're fine. We're fine. Okay. So we got the speed. Let's get more. Let's get more data. Let's get more data. So we want to do set in no set timer. No interval. Right. Here we go. All right.
This good here. Looking good. Okay. This one here. Okay. Let's grab the current time. So, let's say do do we have like the current frame? Do we have the frame? I don't know if we have the there's like a frame ID. All right. Frames. Current speed time. There we go. All right. You have two questions. What do you got? Do you eat meat? I don't. Uh, don't buy stuff. That also works, right? Who uh let's see here. Software, but it's super Oh, hey.
Uh, Inelin. All right. You have zero zero knowledge about software, but this is good. All right. Good to hear it. If you go to the store and buy a pound of steak and eat, that is really going to make a difference than if I don't. So somewhat somewhat, right? It's it's part of the uh if you're just part of the problem, right? You're not you're not the ultimate problem. You're just part of the problem.
Hunting fish. Nice. Squirrel sniper. All right. Hey, how's it going there, AFXM?
Good to have you here. Happy happy Wednesday. grew up in homestead where raised animals at least when more humane than the industrial farm.
See, now that's better. That's fine.
Like it's the big scale that's the problem. Only programming question, the KBC.
Can you show me how the ad live subtitles in OBS? Yeah, absolutely. So, let's jump back over that here real quick. So, is that around here? It is.
Yeah. So, go to this go to this page here, right? And then follow the instructions. It's actually pretty straightforward. You just click on a link and then you and then you paste the link into your OBS like this right here.
Real easy. It's real easy, you guys.
It's three step. It's a three-step process. Super simple. Super simple. Oh, real straightforward.
So, you basically can click this link here and paste that link into your OBS.
You click it. You got to have your browser open. And then you're good to go. And then you're good to go, you guys. All right. And if you want, you can click this link right here. This link will give you the one word at a time. So you can do it like this. So this is the link that does that for you right there.
All right, you guys. Let's get back to coding. Let's get back to coding. Okay.
Uh what is this here?
Speed. Let's see. Current speed. Yes.
All right. So we got our speed. I like to get a time so we know what our current step is. And I'm going to I'm going to add this dot current step equals zero. And then for each interval, we'll increment this step here. So we'll grab this increment it here. We'll do like a plus+, right? A little bit of plus action. And then we'll grab the current step and then include that in the time. Maybe we should maybe we should call this time step. See, time step. So we know better what exactly the deal is with that. Okay, looks good.
Looks good. All right, Sergio, thank you. Appreciate it. Are you studying the time between spawn of cactus? We the AI is going to learn the current. So you see all the different parameters that are being posted over here. Here, let me see here. So the AI is going to learn these things.
So it says speed. Oh, wait. Oh, we got to reload the page here. Let's reload the page. There we go. Speed and time.
So this is this is the this is the time delta here. Speed and time. So we need to know what the frame is, what the current game speed is, and then we need to know more. We need to see if we're jumping and we also have obstacles. So let's see what we get with the obstacles. Yeah. Uh let's see here. In your opinion, is classical machine learning needed as the industry moved into deep learning engine AI? Well, yeah. Uh I think deep learning is still very valuable. We are going to be using deep learning for this. This is a perfect use case for deep learning. The AI is going to see all the features of the game state and then we're going to have it try to predict what moves it needs to make. Is it jumping? Is it doing nothing? Or is it ducking, right?
Cuz it cuz it can duck. You can duck here. You can jump and then you could do nothing basically. So that's what we're trying to teach it.
Hey Heritic. Yo. Hi. Good to see you.
Welcome on in. Happy Wednesday. All right, so let's see if we can get the obstacles. This uh see crashed. I think this is a thing, right? Crashed. See, crashed. CR run. Crash. Crashed.
Crashed. This crashed. Okay, so this is a thing here. So, let's grab that. This crashed is equal to crashed.
Crashed, which is game over. That's game over, right? That's game over essentially right there. Okay. And let's see. Let's do this. And then reload that. All right. Do we have crashed is false? And we might h let's see. See, the thing is that I'm not sending enough updates per second. It needs to be more than the the speed of the jumping. So the current interval for this. So the jumping is we'll do 150. And for the interval here, I think we'll do 500 is not enough. So, we have to do a We'll do every 50. Let's do every 50. There we go. All right. Let's try that. Reload. Okay. So, that is our training data right there. So, you can see that. That's a lot better right there. That is our training data, you guys. Right here. Looking good. Looking good. Okay. It's a lot though. It is quite a bit. It's a lot.
What are the similarities between software engineering and camper computer science? [laughter] Uh so you usually you've got your app developer who's going to plug in plug in components. That's going to be your uh software engineering or developer, right? That's your developer, your software engineer. You're putting pieces together. Here's a database. Here is the front end. These are the ReactJS or you've got your Cotlin. You've got your Swift for mobile apps. you build apps basically. And in computer science, you're going to be learning more of the infrastructure architecture, how everything fits together, data structures and algorithms, like the really important tough parts, right? The really important tough parts because you're going to be building architecture of the world. This is going to require systems level architecture and design usually with C, C++, and Rust for systems level languages.
So that's about it. What about computer engineering?
merge them together. I mean, it's all it's all fuzzy. It's all it's all fuzzy, right? It's all fuzzy. Add a machine lock state to my Python packer protector. Okay, good to hear it. Do you have any suggestions for tips for ReactJS YouTube video for entry level? I actually don't know. I don't know. I All I can tell you is that React is a pretty decent framework. It actually is pretty good. I much prefer spelt.
Spelt.
There you go. Spelt. That's much better.
It's a lot It's a lot newer.
Your promise Your promise remains the same. Wait, what do you mean? Uh, you dislike entering long conversation pages on Google because constantly rewinding and supple. The original message add to the question and can lead inaccurate information. Yes, if you take it too long, if it goes on too long, definitely it can cause problems.
If you make it detect one point, uh, wait. Awesome, bro. Awesome, bro. Can you describe more details? Let me know more details on that. Awesome, bro.
Swift/ Roland or Ronald. Ronald or Messi and why? Uh, you're talking about sports. [laughter] You're talking about the the football soccer. I know you are. I know that's what you're talking about and that is as far as I can go on that. I know I know those are both of the stars in the the game, right? They're they're like world known world known.
One second. I spilled water on my keyboard.
Whoopsie, dude. Whoops. Uh, hey. Hey, I was going there, Chitter J. Bro, she loved me. Hey. All right. Good to hear it, Chitter. Good to have you here.
Happy Wednesday. Congratulations.
That sounds good to me. All right. Can we get our keyboard back? Poor fabric.
Did I put it back? Okay, we're back.
We're back. Uh, you have a script running so much to add less clutter.
Make the script detect the state of the dinosaur crash. True or false? So, is it working on the true or false crashed?
Let's see. Let's take a look over here.
Did it crash? Yeah. So, we see it crashed. True. Okay. So, that that's at least working. We know at least when it crashes, right? All right. Good to hear it. Good to hear. That's awesome, man.
Malcolm, thank you. All right. [music] All right, you guys. Let's see if we can keep going on this. Let's see. So, we need to do more. Let's do this here. All right. So, we've got crash. We've got time. Got speed. Now, we need to add in the obstacles. This is game over. Oh, we need to know jump. We need to know if jumping. Jump. I don't know what to do with this. So, let's see if we can find it. See? Start. There's a start jump, jump count. Oh, interesting. There's a jump count. I didn't know that was a thing. All right, jump. Add AI command jump. So, this is where we can add the AI to actually influence the game right here.
Don't forget to like. Hey, quick fix.
Thank you. How's it going there, Craig Pauls? Hey, good to see you. Welcome on in. Happy Wednesday. Let's see we what can be suggested as tips for an entry-level job interviewer. Okay, heretic. My recommendation is you have to be a really good communicator. Really good communicator and collaborator. That is probably the most important part. Be a good collaborator during the interview. It's very likely that you'll be put into a situation to solve a problem. Make sure you work together on the problem with your interviewer. Don't just sort of like say all the things. If you kind of know what the answer is, like, "Hey, check this out. Look, it's pretty neat." Then, you know, try to figure out a way to bring them into the conversation. Get them to talk. You have to get them to talk. Get them to try to sell you on the role as well. But don't do that too much because it won't feel like you're selling yourself enough. How much memory needs for basic games like this for day-to-days? Uh, I not a lot.
Not a lot. A few megabytes, I think. Going to the win. a few megabytes. Hey Pashan, thanks for saying hi. Okay, so let's see if we can get the jump going here. See, where's this at here? Jump. How do we get the jump?
Let's go. Let's see what all the different data that we have is. All right, so images, audio, context, sound effects, audio buffer, play count, invert, invert timer. What is this? What is invert timer?
Timer equals zero. Night mode. Hey, there we go. Yo, Pasham, thanks for the gold coin. My gratitudes to you. Thank you very much. I appreciate it.
Thank you. All right, Pashan.
Wow. Wow. That's all I can say. Thank you. Uh, will I have to make the project live in the resume? What do you mean?
What do you mean make it live? What do you mean make it live? What is this?
What? Let's Let's just figure out what this is. We'll see what the invert timer is. Delta time is like the current amount of time spent in the game. Okay.
So, let's do here. Let's grab this invert timer. Uh, timer. I don't know what that is. Let's grab this here. And then we're [music] going to see what this looks like. See what this is here.
All right. Reload. All right. It says the timer is zero. It's just zero. It's just straight up zero. We can't do anything with that. That's boring.
Something thing. We don't know what that is. Okay, let's close this so we can see more of the screen. There we go. Okay.
All right. Yo, how's it going there?
Thank you for saying hi. Good to have you here. Happy Wednesday. I can't say what your name is because it's a bunch of kanji or symbols that I'm not familiar with. It looks like uh actually it could be I don't know. I don't know what that is. What is that? Is that Mandarin? It's an Asian language.
Okay, let's see here. timer inverted crashed playing activated.
So, obstacles. Yes, current speed. We have MS per frame. I'm actually kind of curious about this. Let's see what we got here. And then I want to grab the obstacles as well. So, we're going to grab obstacles in here. We go. So, we're going to grab these guys. And then we have time. I'm curious about that. Let's see what time is. I think this will be fun to get. Yo, Leo, thank you for subscribing. Good to have you here. You joined the right channel for software engineering. We're currently training an AI to learn how to play this game, the T-Rex running game that, you know, the one where you don't have internet and the internet fails and then all of a sudden, uh, you have to play this game because Chrome is like, "Oh, there's no internet. Play this fun dinosaur run game [laughter] until your internet comes back." We're going to train an AI to learn how to play this. We've got a plan. Our plan is right here. Let's see here. Under readme, we're going to do a DQN with experience replay. And I'm not sure exactly what this part is, but a DQN is a deep learning model. It's an MLP with activations. Pretty standard.
And then we have a reward-based system.
And so that's what we're going to do.
Live project means to do a live demo in a live project on GitHub. Oh yeah, sometimes there will be a coding exercise. Yes, that can happen. There might be a coding exercise. You're the most useful life happening in YouTube shorts. Yo, going to t going to win.
Hey, thank you. Going to the win. I appreciate it. That's really nice for to hear. Make it multiplayer. Oo, Se forum.
We could We could make this multiplayer.
We could have so many dinosaurs on the screen. Okay, let's do that. Let's add Let's We can make the game multiplayer.
Go to stream ideas and then we're going to say a new post. All right, T-Rex running game multiplayer. Let's see.
Make the game multiplayer.
All right, and then we'll post it. All right, the idea is added. Hey, how's it going there, Rock Manth?
All right, I subbed and it didn't do the thing. Oh, it didn't show up on the screen. Here's the deal. On YouTube, it seems to throttle it. It seems to throttle it because there's a lot. There tends to be quite a few subscribers and so it only shows like every every 10 or every 20th. It sometimes it just won't show up. I know, Rock. I'm so sorry.
Well, you joined the right channel for software engineering. Good to have you here. Today we're building an AI that can learn how to play the T-Rex running game. That's what we are currently doing here. And we are going to make it work.
We're going to make it work. All right.
So, we just need to see what these things are. We can also get high score.
Let's do high score. Well, we don't want the highest score. We want the current score. We want the current score.
Distance ran. Ooh, that's a good one.
Let's grab that. Okay, let's grab the distance ran. That'll be a feature input. Uh, we could do distance meter.
Let's see what this means. I don't know what this is. We're going to find out though. We're doing data exploration right now.
Config, dimensions, details, button, outer. All right, I think that's [music] everything. This is all the stuff that we have. I don't think there's anything else we could do here. So, let's pull all of this into our here. We'll go one.
Let's see. No, we'll we'll we'll we'll make this simple. We'll make this simple. Here we go. Let's indent everything. Five.
Okay. MS per frame.
There we go.
Uh, and then we put at the end. Here we go. Put a comma at the end of all these things. All right. Perfect. Perfect. New here. What's the stream about? Hey, how's it going there, Drowsy? Good to see you. We are a software engineering channel. We are currently building a reinforcement learning model that's going to learn how to play the dinosaur game, right? The T-Rex running game. We are currently doing data capture. We're doing data capture. So, when you Hey, Wolves. Good to see you. All right.
Sephora, automate it. We're going to do it. jump 30 pixels before the obstacle.
Yes. Yes. So, you could basically do that. You could you could basically you you could you could program it to auto. We could make it work really good if we just programmed it ourselves.
You're right. We could program it so it's perfect. Uh and we could we could absolutely do that. We certainly could.
Can I also join the Discord server? And yes, uh Sooie, yes, Sooie, absolutely.
Here, let me say a link for you under general. Let's copy that. All right, Suie. Here we go. Sui. There you go.
Right there. The Discord link just for you. Just for you. All right. All right, you guys. All right. Looking forwards to it. See you there. See you there. Okay.
Are we changing the game code or the AI code? Hey, Rock. Rock, we are going to be changing the game code to extract the state of the game. So, we're going to pass that into an AI that can learn what to make a decision on, and we're going to give it a reward system, right? We're going to give it a word reward system.
All right, I'm looking forward to it.
Your elbow touched grass yesterday. Oh, Bonzupi. Yeah, now it's itchy. That's how that happens. Yeah, you know it. You know it. Hey, Torva. Good to see you.
We're doing your idea. Instead of multiplayer, you can make it so everyone who has its open trains the AI. Ooh, best version gets transmitted via PubNub so we don't have to have redundant learning. There you go, Torva. I see what you're doing there. Yo, how's it going there, Vickers? Hey, Vikars, thank you very much for the sunshine. I need that right now. Thank you very much. I'm looking outside and the sunshine feels great.
Thank you very much. My gratitudes to you. What if every player trains their own AI? They all race against each other in the same game. That reward function you are using, the AI learns to survive longer instead of just maxing score.
Yes. Survive longer. Survive longer.
That's what we're trying to do. Got to survive longer. You know it. Hey, good job, N. Yes, Nina. Exactly. All right.
So, let's see what our our stuff is here. I don't know if we can get obstacles. I don't know if we can do that. Maybe we can get time. So, we need to get the last obstacle is what we need to do. Distance ran. Let's grab this here.
Okay. Distance ran. Excellent. Okay.
Frame. And then let's do obstacle.
Obstacle. Right. Did I spell that right?
I think so. And we'll do just the last one. Just the last obstacle. Okay. Let's see if that works. Yeah.
Let's see if that works. We're going to make it happen, you guys. Make machine a slate. That's basically what we're doing. Yep. Quick fix. Exactly. How does it feel? Hey, Drowsy.
Future of software development. As a software dev, I feel slightly worried.
510 announced by AI. Yes, you're going to be replaced by AI. No, really. AI, the current AI tech is just a tool, right? It's a tool. You need to get used to the tool. And it's going to be common standard practice to have the AI. I do that a lot. I have the AI do that for me. I have it [music] code for me all day long. Get your replay. Thanks a lot, bro. It was good. All right, Wolves.
Thank you. All right. the game being played by an AI. Sam, technically, yes.
And it's not a good AI, but we're gonna make it good. Yeah, there. We're going to make it really good. Hey, Uncle Dream, thank you for the silver star. My gratitudes to you, Uncle Dream 7. Thank you very much. Thank you. Good to have you here. All right. Is the game being uh Stephen, happy Wednesday. Hey, Muhammad, happy Wednesday. Have you considered screenshots of the game current state rather than inspecting the game state?
Yes. So we could do a convolutional neural network or a vision transformer to capture all the image output data and then we can train the model on the vision part of it. Absolutely we could.
And that's going to take more compute.
And so I figured what we're we could do that version two. We could absolutely do that version two. I just decided that I want to do it this way. We're doing good, Uncle Dream. Thank you. How are you? We're doing good. Do you still review AI code nowadays?
Mostly what we do is we have uh people review it lightly, but then we have AI review the code is what we do.
Yo, Learning with Musa. Hey, how's it going there? Learning with Musa, welcome on in. Good to have you here. Happy Wednesday.
All right, let's ask a poll, you guys.
Actually, I'm kind of curious. All right, start a poll. All right. Um, do you have the AI review code written by AI? And the answer is yes is the correct answer. Uh, yes. AI reviews. AI reviews. Uh, IE WS AI written code. AI written code. Uh, or no. No, I read it. Okay. Do you read the code or does the AI read the code? What do you guys do? Okay. So, let me see here. I think we've got everything I need. Let me let me pull in the new data that we just captured. All right. Let's see here. Uh, config. Oh. Oh, we got something weird going on there. Hold on.
What is that? We got some weird stuff here. Wait, wait. Weird stuff. What's this? Speed time crashed thing zero distance meter. Okay, so we can't have distance meter because that's a bunch of weird stuff. Oh, there's a sprite position.
Oh, what is this? Current distance.
Wait, so distance ran. Okay, I'm trying to see see what data we've got available here.
Uh, there should be a third option.
Both. Oh, Bonupi. I didn't even think of that. Bonzupi. Ah, I should have waited.
I should have waited. Current distance, max score, high score, container. Let's see here. Flash timer. See if there's anything else interesting. Oh, we got some limits here. Coefficients. [snorts] Okay.
Max or MS per frame. Oh, number of milliseconds per frame. Okay. Distance ran. All right. So, the one thing is that probably don't need that. So, let's let's hide that. Although maybe that was the obstacle.
What? No. Bring that back. Let's reload this. Okay. Uh obstacle. Where's that at? Here. Do we see it? Obstacle. Obst.
Obstacle. It's not there. Where'd it go?
Here. Load. Save. Save. Save. Make sure it's saved. Yep. Okay. Reload. All right. Do you see obstacle time speed crashed thing? MS per frame, which we don't need to worry about. distance ran.
Where's the obstacle? This Where's this at? This should be there.
Oh, because we're saying zero. No, it should be there, though. It should be there. Let's see here. Obstacle.
Spell that right. Spell it the correctly way. Spell the correctly way there. Uh, fos boss. It's not there. It should be there. Hey, thank you for saying hi.
Good to have you here. Uh, looks like that's a Korean. Is that some Korean? I don't know. Hello. I'm new to coding.
Give me some tips. Learning with Musa.
My favorite tip is to start with Python.
Start with Python. It is a good programming language that's really simple and it gets you going. So, you can do this, right? Python. You can say uh print. And then you put some quotes.
Hello. Say print hello. Hello. Uh learning with Musa.
And then boom. Printed. There we go.
Right there. AI views. All right. Okay, game over. Hey, right, game over. I like that one. Hey, Sigma Edits, good to see you. Welcome on in. Happy Wednesday.
Good to have you here. All right, so why is my thing not working? I want this to work, you guys. Here we go. Reload. All right, space jump. Okay, and then we've got our data. I want my obstacles. Time crashed thing. MS per frame distance ran. And then obstacles is just not there. It's just missing.
Maybe because I need to do more. I might need I need to do something else here.
So, obstacle uh I'm going to say I'm just going to put this in a big string.
There we go. Obstacle, throw this as all caps. Then we reload the game and I should see see. Okay, so there it is. It's there now. It's just missing. Uh how's it going there, Nervin? Uh what are we trying to do today? We are building a reinforcement learning AI model. We're doing data exploration right now. We're extracting the state of the game and you can see that kind of flickering on the screen on the side of the screen over there.
That's a bunch of game state. We're going to add that as features into our AI model as the input data. Hey, how's going going to the win for the for awesome? We're doing it for awesome. How can the AI know where the obstacle is?
Does it have the a way to know? Yes, it does. We are currently trying to capture that right now. So, see right here on the screen the line that I'm currently highlighting and unhighlighting. It's a snake. Hey, Malcolm, it's a snake. We are We need to grab the obstacles array and we need to do something with it. So, right now the obstacle array is obstacles. Let's see if we can bring this back here. See if I can grab this.
See what's in it. Maybe obstacles is not what I thought it was. Horizon obstacles. So, that might not even be a thing. H. All right. Let's take a look for obstacles that go at a different speed from the horizon. Okay. And obstacles. Obstacles. Yeah. See, we've got some obstacles right here. Function horizon. Okay. No obstacles. We need We need to find the actual obstacles. Oh, slice. Oh, zero. Okay. Update obstacles.
All right. So, here this is all the obstacle data right here. So, we need to capture this. Yeah, we got to capture this here. Absolutely. Wrong method. Oh, is it? You think so? I think I think we're doing okay. Joke. It's a programming language. Oh, it's a a snake. That's right. Python. Python. I didn't not I I missed the joke. I missed the joke.
We could do several other approaches. We could also capture screenshots of the game itself, right? We could capture screenshots of the game itself and then use that as into a vision transformer or convolutional neural network. We could do that. What's the stream about?
Learning uh we are a software engineering channel. We build and write all of our code with hands on our keyboard here. Right there on our keyboard. That's what the stream is about. And today we are building a reinforcement learning AI model that will be very good at playing the game.
That's one way. Yes, that's one way.
We're going to try We're going to try this way. So, I've never done this approach. Reinforcement learning is a new adventure for me.
Easy to code, but cuz I can just take the obstacle size and stuff. Yep, that's what we're doing. That's the idea. What coding are we doing? Oh, this is JavaScript. Uh, JavaScript, Python, and Bash. JavaScript, Python, Bash today.
Vist the obstacle maybe. Oh, yeah. Yeah, maybe. Yeah, actually. Cool. Sigma edge.
Thank you. Thank you. Super cool. Hey, Vladivavivar. All right, Vladivar. Thank you. All right. See, jump statically, but you want to figure it out on its own. Yes. Yes. Chamom. Chamom 80085.
Yes. Yes. We want it. We want to train it. Right now, we just need to figure out how best to capture the data and then we're going to import it into the model. Right now, it says jumping cuz I want to get the data. So, we need to grab if I'm jumping, I need that data. I need to see what my current my my sprite, this dinosaur here, the dinosaur, I need to see if it's if it's jumping or not. I got to see if it is jumping. Let me see here. One second.
How do we do that? Let me click this.
Um, how do we do this? One second, you guys. Boom. Reply pin. No. How do we do this? What is this? Ah, perfect. Okay.
Yes. No. Okay. H. All right. How about that? Let's see here. Let's see. Warn.
Okay.
One second.
Okay. All right. Okay. Sorry about that.
All right. So, hey, Snot. Good to see you. Welcome on in. Happy Wednesday.
All right. Let's see. Yeah, Stephen. The lives are phenomenal. Yo, user. Thank you. Appreciate it. Uh, I thank you very much, bro. The second time seeing you.
Love your enthusiasm. Hey, thank you.
All right. Good to hear it. 3D artists, thank you very much. Good to have you here. Hi, buddy. Hi. Good to have you here. Ola me amigo. Maybe you are wrong or it's not obstacles. Oh, right, right, right, right, right, right. Yeah. Um, let me see here. Let's find out. Let's find out, you guys. Let's see here.
Obstacles. Yeah, you're right. It's not.
It's just empty. It's just straight up empty. Hey, thank you for the silver star. Colton Colton Welch, appreciate it. Also, Sarav, thank you for subscribing. Good to have you here. You joined the right channel for software engineering. We're currently building a reinforcement learning AI model that's going to learn how to play the game. And right now, the it's just jumping like crazy. It's just jumping. So, obviously, it doesn't know anything because we haven't really taught it how to see or how to respond. So, we need to get we need to make a reward system. Right now, I'm capturing the data from the model or from the game itself. So, you're right.
Hey, Sigma Edits, you were right. It's uh might be this horizon. You're right.
So, let's see here. Obstacles.
Horizon. Right. All right. Right there.
A little bit of Horizon. Yo, Lupe for the flower. Thank you very much. Thank you very much. My gratitudes to you.
Good to have you here. How's it going there, Chris? Chris Ice and Jared. Hey, Jared. Brothers. Hi. Welcome on in. Oh, and another flower. Lupe, thank you for two flowers in a row. Oh, my gratitudes to you. Thank you so much. Uh, it's stuck on ISO 3 is 3. Uh, we're going to find out. We're going to find out here.
We're going to see. We're going to see this. Horizon is a new horizon which has a bunch of sprites in it. So, this horizon update update invert horizon.
I'm trying to see. There's a reset.
Interesting. What else? Horizon we got?
Yo, Lupe, another flower. Hey. All right. Thank you so much. Uh, how can you capture the data from the game? Are you running it locally? Yes. Uh, and second question, what environment are you using? This is a web browser, right?
It's a web browser here. And I've got my terminal window running T-Max, Ghosty, and Vim. So, this is Vim text editor.
Uh, over here is my bash script. So, you can see I highlight highlighting my bash script here. That's capturing the data from the game. And then in the bottom corner of the screen here over here, this is all the game data. So, we're trying to capture the obstacles. And then if we have all that data, we'll be able to input that into an AI model and it can start learning how to play the game, you guys. And that's the plan.
That's the plan for today. All right, let's see if we can make it work. Yo.
Okay. Uh, let's see here. Obstacles horizon. Let's see if we can get that.
Delete this. All right. Let's see if we can make that look good here. Reload.
Go. Do we have anything there? Obstacles is still empty. It's still empty. We need to find where the cactuses are. All right. Where are the cactuses, you guys?
Where are they at? Where are the cactuses? We got to figure out where those cactuses are. All right. One second. All right. One second here. All right. Clicking button. Clicking button.
All right.
All right. Perfect.
All right.
All right. One second, you guys. Getting back into it. So, we need to figure out how to do this. Uh, where is it at?
Where are the obstacles? Update.
Horizon. Update. Update. See, there's obstacles right here. Look. Check for collision. Horizon. Obstacles. Zero. All right, let's try Let's try zero. Let's see.
All right, let's see here. Uh oh. Obs.
We'll call them obs.
See if that works. All right, reload.
All right, do we have any obs? Do we have any obs there? So, obs.
We have regular obstacles, but we don't have those kind of obstacles. Okay. So, it's still empty. Let's look at the source and open this in a window here.
So, let's grab this. All right. Uh, this.obstacles.
All right. Console.log.
Let's see if we can get this. Let's see if we can debug this here. Little bit of debugging. Yeah, just a smidgen of debugging. Did you finish YouTube? If not, when are you going to uh continue?
Oh, hey, going going to the win. Going to the win. Uh we basically finished the MVP. We finished the MVP. It it you can upload videos, you can watch videos, and you can stream the video, you can skip ahead and skip behind the video. We basically added the MVP. We got it working. And it's at a point where if I kept going, then I'd have to release it.
And if I released it, then I don't know.
What are you guys going to put on it?
[laughter] What are you going to do? I don't know. I'd have to like moderate it. Uh, it's stuck on IL3. Let's see.
Sorry for the last question, but I'm curious. How can the two programs, the dino game and the AI you're coding communicate? Oh, good question. All right. So, it is using Oh. Oh, look.
Look, it's there. The obstacles. Oh, it's there. It was there the whole time.
Hey, look at that. It's just when there's no obstacles on the screen, it just doesn't show any. But when there are obstacles, it does. Oh, you can see it. You can see it here. Let me close this so you can see a little bit better.
Take a look at this right here. Here we go. Watch this. Watch this. See, look at that. Oh, it's so much better. All right. So, now we need to learn how to parse that. We need to grab like the current like the closest obstacle. We need to grab the closest obstacle and then we need to see like where its position is. Hey, Justice, thank you for saying hi. So, it's done. So, how does it communicate? So you can see here we're using this function called pubnub.publish, right? Pubnum.publish. This is extracting the state of the game that allows us to capture it. And you can see it being printed out on the right side of the screen. So this pubnum.publish is going to extract the state of the game. That allows us to capture that game anywhere else in the world. You could you could see it right now if you wanted to. You could see my game state because it's transmitted globally. It's a global transmission.
So it's it's pretty straightforward. All right.
All right. Perfect. Perfect. Okay. So if that makes sense and if you want me to go further the order to receive data so we can we can hide this here. So watch this. Yo. Hey give me my channel back.
Just subscribed. All right. Thank you very much. Good to have you here. You join the right channel for software engineering. We are currently building an AI that can learn how to play a video game and it's called reinforcement learning. You might build a game someday with PubNub. Hey. All right. Good to hear. Yeah, you can build multiplayer games, which is my favorite approach, right? You can build multiplayer games.
So, vi monitor. So, as you can see here, we're subscribing. So, I'll set wrap so you can see it a little better. Wrap.
So, this line right here is responsible for capturing the the data. It can capture the data. It can receive the data. So, this little monitor, this little 10 10line file monitor script allows us to see the data, right? Pretty easy. It's pretty straightforward. See how small it is? It's super clean, super easy. This is the game channel, and we're publishing data to the game channel. This is the game channel here.
And our game channel is called T-Rex Dino Game State.
And so, you can see that these two channels match here and here.
So these match here. So those things match and that allows me to capture that data and now we can bring it into our AI model. Yeah. So what I need to do is I need to grab and I need to normalize the cactus data. So the the collisions I need to be able to make sure that there are are obstacles present. And then I need to always make sure to say how far out like the nearest obstacle is.
Basically giving it sort of like a vision. It needs to give it vision.
The music's way too dramatic. Frog talk.
Oh, well, how about now? [laughter] It So, it's like an op a public API.
Yes, tell me. Yes, exactly. Public API.
You got it. When you become the software engineer, which one is more fun, playing or making the game? [music] Making the game? Making the game. Yay. You know it.
All right. The music is now no longer dramatic.
Okay, here we go. So, uh, we Let's see if we can grab this. All right, so was it OBS? Hold on. Let me see here. I want to I want to find out if this was OBS or not. Here it is. Sp Look at all that stuff here. Okay, so it was OBS. Was it also obstacles? Obst.
It is. It's both. It's both. Does it have Let me see here. So, where do we put that? Obstacles first and then OBST second. All right, let me see here.
We're scrolling. I'm I'm doing I'm reading the data. I need to be able to see what's going on here. Distance run.
Okay, so it is only obstacle. No OBS.
No OBS. Okay. For/OBS.
It's there, too. It's there, too. Is it both? Are they both the same? It looks like they're both the same. H. Okay. One second. Let me see. OBS obstacle zero versus Okay. So, they are okay. All right. So, I'm going to say obstacle.
We're going to grab the last one here.
And then we need to we need to extract the data from the obstacle directly. So, here OBS grab this. Okay. Uh let obstacle const obstacle.
There we go. Making a game is way more fun. Exactly. You know it. Uh, see, do you know to reset tomograph tomog? I don't know what that means. Hey, loose juice, thank you for the star.
Appreciate it. All right. My gratitudes to you. Now you want to learn now. Now use that for learning programming. Hey go. Yeah, there we go. Go. You know what's going on. You know it. You know it. Here we go. So, I want to be able to confirm. Let's see here. Uh, let's do console log on our obstacles. And I need to see this a little bit better here.
Yeah. See if we can see it a little bit better. Okay. So, let's console log obstacles if you don't mind.
All right. Hey, Tova. Thank you for the moderation. Hey, Flawless Revelations.
Good to see you. Celestia, also good to have you here. Celestia Abstract, thank you for saying hi. All right. So, let me open up the terminal window here a little bit. All right. Reload.
terminal. Here we go. Obstacle.
Okay. So, there are obstacles. It is there.
It's there. Here, let's just open up one real quick. So, let's look on this.
There's a gap. There's a size.
Uh, and there's sprite position, which I think is what we really want. We want sprite position, right? And then maybe X pose and Y pose. Those things are what we're going to want, right?
So, these are the things that we're going to want on our object here.
These are we we we need to grab we need to grab these things here. A little bit of grab and just a smidgen.
All right, let's try that again.
Undefined. Okay, so when it's not there, it's undefined.
This makes sense. This makes sense.
Lean. Why? Why? Why exactly? Lean. Why?
Buddy, hey, loose juice. Lean. Okay, sounds good. All right. I want to figure out here. Okay. So, where where do we want to grab this? Here. Let me go here.
Close that. And then here, I've got an idea. Split.
Close VI monitor.
Monitor. There we go. Okay, now I can see it better. All right. So, what do we got? We got dimensions. That's a big dimension. Uh, current frame gap coefficient.
Uh, that's weird. Oh, maybe that's the size of it. So, there's like a little bit of extra in there. Like a smidgen extra. VMT. Do you know how to reset a bit buddy? I do not know what that means. I don't know what that means.
All right. One second. I need to confirm my bit. So, I do have a meeting coming up here. So, let me let me get into it real quick. Real quick so that way I can be ready to join it when it's ready.
Here we go.
Uh, and my main goal is to get the obstacles. If we can get the obstacles, we're good for today. All right. One second. Uh, all right. Looking good. All right.
Let's see here.
Can we do One second. One second, you guys. I'm trying to figure out how I want to do this. Yeah, I suppose we could do that. There we go. Okay, perfect. All right. Uh, please, Stephen, can I have a mod? Oh, hey, Libron. Uh, ask a mod. Yes. So, uh, you have to ask a mod and if they say yes, you get it.
Especially on Discord. So, at um Oh, Tova. Tova. Tova is here. So, Torva will say yes or no. By the way, Steven, do you really not know what a Tamagotchi is? Oh, yeah. Yeah. I remember the digital pets from the 2000. Yes, of course. Of course I know. I remember wanting one. All right. See, Torvvis said no. Tarts Liberin, you have to be here for a little bit longer. Okay. So, Tommy, you did a project in school. Minecraft mod connected to a Python AI that was able to tell if the newspaper article was fake news or real. Hey, would you write in a book the article and I would tell you, hey, tell me that is a really great project. That is fantastic. Fantastic project. Hey, Winston. Oh, welcome back.
Good to have you here. All right. Uh, don't worry. We've got time left, you guys. We got time there. No one's on the meeting. See, it's just me. It's just me. See, that's you guys. See, that's you guys right there. That's you right there. That's you if you didn't know. Okay.
All right. Uh Snot. Oh, Snot. Hey, what about what about Snot, Tova? Snot is a regular here. Okay. Uh I need to grab the data. And from the obstacle, I need to check if the obstacle is uh null. So, if obstacles, wait, just obstacle. Just I grab the obstacle. Grab one of the obstacles. And I need to be able to grab the data from the obstacle is what I need to do.
Uh how is the AI doing? It is not existent yet. We are still doing data research. We're doing data research which means that it's a lot of it's a lot of work to do the data research.
Okay. So if obstacle then we need to do some extra logic here. Uh then I need to be able to grab the the the nearest obstacle, right?
nearest or maybe expose obstacle x equals const although how do we want to do this? How do we want to do this? I suppose we could set this up here. Yo, hey, thank you for subscribing. Oh, welcome on in. Good to have you here.
Happy Wednesday. You joined the right channel for software engineering. I would say what your name is. Oh, like Liban Liban Jas. Hey, Liban Jas. Thank you very much for subscribing. Good to have you guys here. you guys joined the right channel for software engineering.
We are currently building a reinforcement learning AI model that will learn how to play the dino jump game, the dino run jump game. So, we are trying to figure it out. So, we've got this X pose and Y pose. And I'm going to see if I can capture it. Let me let me explore it a little bit better here. Let me explore it. So, we see we've got a bunch of options here. I'm going to grab and copy this so I can really really look into it. I want to see what's going on. All right. So, is that the full obstacle? Hold on. Let me check here.
Obstacles. All right. We'll grab that.
Copy it. And then we're going to explore it here. Let's explore it. V. Paste.
Here we go. Set. Wrap. All right. So, let's explore this really quick. Yeah.
Let's do a little bit of exploration here. Here we go. All right. What does that [music] look like? All right. So, do we have a bracket? Oh, there is a bracket. Collision boxes. Is that what we want right there? We want some collision boxes. Correct. Congrats on never getting mod. [laughter] I'll make a note in mod chat just for you. Thank you, Torva.
Okay. Uh, are you going to make an AI play other games in the future? Yeah.
So, we're doing reinforcement learning right now. So, we're trying to do the data capture part of it. So, sprite position. This might be what I'm looking for. I might want the sprite position.
[clears throat] Type config custom.
don't care. I also want the width and height of the sprite. So, we could grab that as well. There's also a Y pose and then a min gap collision boxes.
Let's see here. What else we got? We got a width and height. Got some coefficients on the gap, which I assume is something. Let's get the next song started. There we go. Okay. All right.
Can we hurry this up? Oh, uh, it's going to take a while. Don't worry. It's going to take a while. It's going to take a while. Coefficient dimensions. Remove false expose.
Collision boxes. H. I'm trying to figure out what I want. Right. What do I specifically want? So, we've got our collision boxes here. And I don't think I need that. I think I basically just need to see where the current position of it is. So, we've got double collision boxes, which is weird. Uh, let me see here. Is did I get two? Let's see here.
Uh, sprite. So, let's see. Let's see if this is what we want. Maybe this is what we want right here.
Do you know how to reset a Tamagotchi? I don't. Do you have to press like multiple buttons or is there like a little pinhole? You made AI play Minecraft a long ago. Neon Java. That sounds like a a difficult challenge.
That sounds like a difficult challenge.
All right. So, I think I'm going to try grabbing the sprite position. Let's try that. Here we go. Uh, all right. Sprite position.
All right. Obstacle const obstacle. Grab the obstacle if obstacles X. So obstacle dot was it sprite pose. Sprite pose.
Let's see here.
No, let's do let's do let's see if we can grab this. [music] Uh let we we need we need to define it out of uh the scope here. Let's grab let's grab some scope definition.
Uh we say one. We need to have a default. We [music] need a default. So a th00and and then obstacle y.
Let's grab this here. And then [music] we're going to update these like this.
All right. One second.
Okay. Oh, no, no, no. I want that back.
Here we go. Uh, then we're going to paste these over here. And I need to grab the What do I want? Uh, sprite pose X and Y. Okay. X [music] and Y.
Excellent.
All right. Looking good now. Looking good.
Hey, bro. What are you doing? Hey, Rex Xviper. Rex. Rex Viper. Thank you for asking. We are currently extracting data from the T-Rex game so we can train an AI model with reinforcement learning. Do lethal company next. There were a lot of ideas there. Hey, that sounds like a more a bigger challenge for sure. That sounds like a bigger challenge. We can do that, too. We certainly could. We certainly could. All right.
Do you know uh let's see. Are you going to make AI play Call of Duty online? Yo, that is a a bigger challenge. We could though. We could I think we would be cheating, right? Does that count as cheating? That counts as cheating, doesn't it? I think that's cheating. One second, you guys.
A little bit of cheating there. Join server break the wood match yaw pitch and stuff stuck in the cracked item by itself. Uh, that time I was not much into AI. You never tried it again. That was it. Yo, Pro Folk, good to see you.
Wait, Pro Folk. Yeah, I think that's against the toss. Yeah, exactly right.
Against the terms of service. Having an AI is essentially like an aimbot at that point, right? It's an aimbot against the CODs to terms of service, you guys. Yo, Blockot Gaming, thank you for the silver star. My gratitudes to you. Good to have you here. Thank you very much. Now, do you think the code works or doesn't work? Let's try this. Reload. [music] All right. See here. All right. Do we have it? Do we have it? We didn't add it. [music] We didn't add it. Okay. So, it's causing an error. We can fix that though. All right. If obstacle then uh so we need to grab this over here.
Obstacle X obstacle Y. So what we're going to do here and see OBS X and Y.
There we go.
Okay. All right. So I think that does the trick. See if this works yet.
Reload. Uh obstacle X is not defined.
Oh. Oh. Is it not? It It's right here.
Oh, I'm spelling it wrong. One second.
Let's Let's fix this. Thank you for the reactions, you guys. Oh, my meeting's getting started. All right, let's see if I can make this work here real quick.
All right, obstacle. Obstacle. Uh, yo, game reload. Okay, obstacle X not defined. Save here. Let's just delete this really quick. Uh, delete. Delete.
Reload. Okay. Undo. Undo.
Unreo, redo, redo, undo. Okay. Uh, let's see here. Obstacle X, obstacle Y.
There we go. And then we need to spell this correctly. Obstacle sprite X. Okay.
Looking good. Looking good. Good day.
How's it going there, channel? Yuyu's you. Good to have you here. This is very interesting, Rex. Thank you. Appreciate it. Everybody should sub and like the video. Hey, Winston people or Winston.
Winston, thank you. Penelopey. Winston.
Penelopey. Thank you very much for saying that. That's very nice of you to say. We're all gathering shortly. Should be on. Okay. Sounds good. All right. One second you guys. I do have to get going to a meeting, but I would really like to get the obstacles working real quick.
Nice. Nice.
On meeting ready. Okay.
Send. Perfect. Okay. All right. One second.
Okay. All right. All right. Let's get back to it, you guys. One second. Uh 77 viewers. Crazy. Yo, no way. Hey, that's amazing. We're at a new play. We're at a new uh a new record. Um all right, so you guys, so take a look at this. Uh I My name is Steven Blum, CTO at PubNub.
Good to have you here. Uh I do have a meeting coming up that I have to get to.
I had a lot of fun hanging out with you guys today. Uh we're going to be building more AI reinforcement learning tomorrow. And this is uh we're using my communication technology PubNub to extract the game state data and then we're going to put it into an Python AI model.
Competitors, hey, there you go. Hey, we're we're we're almost. We're almost there yet. We're not going yet. We'll be leaving in a moment though. In a moment.
Not fully gone yet, but we will be leaving shortly. Uh I do have a meeting.
So, we probably have like good 5 minutes or 10 minutes. You see? You see? Nice.
So, let's see if I can get this to work here. I want the obstacle X and the obstacle Y. And I want to make sure I do that correctly. Obstacle. Obstacle. Does this look correct? That looks correct to me. All right. All right. Looks good.
Okay. So, let's try to reload this. Hey, Rocks. Hi. Good to see you. Welcome on in. Happy Wednesday. Good to have you here. Stephen Recker is 210. Oh, yeah.
Hey, Torva. You got a good memory. Are you using Windows or Linux? This is a win a Linux variant. So you see here I will uh split the screen. You name you name you and name it is Darwin which means Mac OS. It means Mac OS meeting.
Happy Wednesday. Yeah, we're we're going to head on out here in a little bit.
We're not done yet though. We're not done yet. Uh I want to make sure that we are grabbing our obstacle data. So far so good. Do we have it? Hey, look at that. Okay, so we do hit an error when the obstacles appear. Uh, cannot read properties of undefined reading X. Okay, so our current bug is this sprite. I spelled it wrong. See if that reloads.
Okay, see if we get good now. Good now.
All right, we're going to find out.
We're going to find out. Hey, there we go. That's what I'm looking for. Okay, perfect. Perfect. Perfect. All right, obstacle X is I need to see I need to see this a little bit better. All right, when when it gets close, I'm going to pause this. I'm going to pause it. What?
Okay, let's pause it. All right, so we got our obstacle Y and our obstacle X.
So it was 460. [music] All right, so let me find this here. So I'll do this search. You know what we can do is to make it even easier. We can just hide all this other stuff cuz I just want to see I just want to see the obstacles. the obstacle data. Here we go. Distance.
Here we go. Hide that. Hide that. Okay, let's see if we can see this. All right, let's reload the game. Go. Okay, much easier to see. Okay, so does the obstacle X and Y change. It does, but it doesn't see. I need to see when it gets closer. So, we need to see when the cactuses get closer to the dinosaur. So, that must just be the width and height of the obstacle [music] of the sprite.
So, we need to see more data. So that happens over here. So where is that? We see our width and height and our pos y position. And maybe we have an X position. X pose. Do we have that? Yes, [music] we do. So let's see if we can get the X and Y pose instead.
So let's do the X and Y pose. X pose and [music] Y pose.
There we go. Close that. Looking good.
Okay, let's try reloading. All right.
Reload. Go. All right.
Let's see if it worked you guys. It should red it reduced. Yeah, there it is. There it is. That's what I'm looking for. Hey. All right. Hey. Okay, look at that. And also it goes in a negative, which is important information. That's important information. Mac OS is also officially licensed variant of Unix.
Hey, there we go. There we go. Uh, I believe Darwin kernel was the originally based on the FreeBSD kernel. All right.
Good to know. Good to know.
It was. It was the forts. Hey, the forts. How's it going? Yo, welcome on in. BSD is not Linux variant. Yeah. Did Chrome ever change the coding of this game or is it still the same from back then? It looks the same. It looks the same from way back when it looks the same. Okay. So, now we've got the X and Y coordinate here. So, now the A and Do we have a Y position as well? We do. So, if we get the Y position so you can see.
Oh, that looks so cool. Okay, so now we've got all of our data. We've got all of our training data. So we don't need this, I don't think. [music] Uh we just need the crash, the time, and we need to reset the time step. We need to reset to-do reset time step uh [music] on crash. We have to do that on crash. Reset the time step, which I suppose we could do. Say if uh this crashed um then we say time step here we set time step to zero this timeep equals zero and then we else we will increment [music] it. So we'll do something like this. Yeah. Uh perfect. Okay. if it's crashed at zero.
If not, and we can also optimize it, I think maybe by setting a not operator here. So, we always run this one first.
Yeah, whatever. Doesn't matter. Okay, so set frame for Hey, Food. Yo, Food. Hey, Stephen.
Looks like a lot of fun. Yeah, we're doing some fun things over here. You got it. We are currently getting our uh AI model. all the codeex vibe coding went.
Oh, rocks. Um, good question. Right now, I don't know. I don't know when. [music] What we're going to do though is we're going to capture all of the data from our game and we're going to make it so that way we can pull this into a Python script. The Python script is going to have a reinforcement learning model that we're going to train on all of this data from the game state. Your very first Apple was just approved just now. Yo, Malcolm, that's great to hear.
Congratulations.
All right. So, let me just close some of this here. All right. Let's bring back everything that we need. Do we need the time? Is this something that worked? I don't know if it was or not. Do we need the distance ran? I think we should.
Let's get the time. We need to see when we're jumping as well. Okay. To do when player is jumping. We need that information.
We need to see what that is.
Uh what is the current jump URLs are automatically deleted. Yes, they are. Yes. Nice game, Krishna. Ah, it's a good one, right? It's a good one.
All right. So, let me see here. So, I need I need this item. I need this item here. Do we need the distance? Do we What is this? Do we We don't need that.
Uh, we did this speed, time, crashed, obstacle x, obstacle y, the current time of the game, which I'm not sure about, distance ran, which may be interesting, and we need to see if the player is currently in the air. So, we need to see if it's the player the player's current position in the sky. I wonder why my past URL kept disappearing. Yeah. Yep. That's why. That's why. Yep.
This is what I need to watch. Very interesting. I like it. Hey, Rex. All right. Good to hear. Good to hear it.
Thank you. All right. Let's close this, reload, and then get the game going here. There we go.
All right. Looking good. Okay. Perfect.
Perfect. All right.
Okay. Hey, how's it going there? Uh, Chen Chan, welcome on in, Chen. Good to see you. Happy happy Wednesday.
Okay. All right. Uh, crashing fine. All right, let's reload this. Oh, we got everything we need there. So, we see a big obstacle. I need to divide it by a,000. Although, what is the what is the biggest value that we see? Uh, on the X obstacle X. Let's bring that back really quick. I need to see that cuz we need to divide it by the max of the max distance. So, I need is it X? I think it's X is the answer. Yeah, let's grab this here. All right. So, a th00and I need to grab. It's going to be it looked like around 600 there. Is that about right? Like around 600. Do you guys see that? I need to see this here real quick. Uh, okay. It was roughly 617. Okay. So, what we can do is divide by 700. We'll do a 700 division.
So, divide by 700.
Uh, and then divide by I guess 100.
So, that way we have to scale it, right?
We got to scale it. See here, time, distance ran. Do we need this? Uh, no.
We don't need that. Distance crashed time [music] and crash is going to be zero one or zero, right? So, it's going to be it's a really straightforward option there. Uh, Steven, I need your expertise. Okay, another matter. What do I do? This elegant and this comically large head of cabbage I got from the farmers market. You throw it away, right? at five pounds of assorted mushrooms. You keep the mushrooms. You Oh, wait. The cabbage, though. Oh, wait.
The cabbage. I thought this was iceberg lettuce, right? Because there was an issue with iceberg lettuce recently. Um, what I do is I throw those things into a ceramic glass container. I wrap it so that way it's covered. You cover it and then you steam it. That's what I do. I steam it. I steam it in the oven. I steam it in the oven.
Okay, let's see here. All right, so that looks good to me. That looks like all the data I need. Now, what about all the other things? Maybe we can scale the other things. So, let's see here. I'd like to see what crashed. We're going to do data validation right now, you guys.
Data validation. Thank you for the hearts. Appreciate it. Let's see here.
All right. Data validation. I should see a when you run it into the cactus, it should say one for crashed.
One for crashed. There it goes. There it goes. All right. Good. Good. Good. Okay, let's see what time does. Let's see what time looks like. Here we go. All right, time is just a forever increasing time step.
And it should start over when you crash.
All right, good. Okay, let's see what speed is.
See what speed is. Load that. Speed is increasing ever so little, like a little bit at a time. All right.
Uh, I'm making my leave. Oh, sounds good. going to the win. Uh, see you in the next one. Don't be late or your meeting. Oh, hey, we're good. We're on the meeting right now and uh they're late. They're late. So, we're good.
We're good. We're plenty good. We're plenty good.
All right. Uh, speed, time, crash. All right. So, this is good. We know that's good. Crash is good. What about distance meter? What is distance meter? I don't know what that is. So, let's see if we even need that.
Whoa, look at this. Okay, what is this? Uh, this is very interesting. I have no idea what this is. Yo, Leo, thank you for the gold coin. My gratitudes to you.
Appreciate it and glad you guys are here today. I'm so glad you guys are here.
Payroll deduction for being late. Right.
That's how it works. Exactly. That's how it works. Please. Debbiannet hidden. Oh, you're trying to I see what you're trying to do, Snot. You're trying to give a URL. You're trying to do a URL there. I [snorts] see it. Uh, you can just put it in Discord. Just put it in Discord. That's a lot easier. That's a way easier than trying to do all the other things. LeBron James. Hey, I'm back. Still doing it. Yep. We're making our way. So, we are looking at all the different data points here. Now, let me close this really quick so I can see what's going on. All right. [music] Canvas. So, there's nothing in here.
There's nothing in here that I need. I don't see anything in here I need other than the digits. I have no idea what the digits are. Uh looks like maybe a score of some sort. Oh, it's the score. All right, that's what that is. Okay. Yeah, nothing in here. I need I don't need the distance meter. So, we can delete that.
Uh local lettuce is highly unlikely to be contaminated. Oh, right. Because that was from Mexico. I remember that.
Farming from 10 acres lettuce. Yeah.
Okay. Bonsupi. Okay. Understood.
Understood. All right. So, let's make sure what our time is. I want to see what this is. Let's see what that looks like. Yeah. All right. Uh, it's a very weird. Uh, it's a very big number and I don't really need it. Okay. Cuz I already have my time here. So, let's delete that and then let's see what our distance ran is. All right. Let's load that here. Zero. Okay. So, that is our current progress in the game, which is quite a big number. Do I need it? I'm going to hide it for now. I don't think I need it.
I don't think I need it. Okay. So, now we need to get when the player is jumping. We need to see when the player is jumping. Okay. All right. So, let's grab the player jumping there. Hey. All right. Wait. What' you do? What' you do?
Oh, for the sparkles. Hey, Rachel Goodman. Hey, Rachel Goodman. Thank you very much for the sparkles. I appreciate it. Good to have you here. My gratitude's to you. Thank you very much.
All right. So, now we've got our sparkles. Thank you, Rachel. And then I need to see when I'm jumping. All right.
So, jump back slash C. All right. So, we need to see when the player is jumping.
Initial jump velocity, min jump height, jumping keystrokes, and then we have we don't need to worry about that. Jump, jump, jump. Let's see.
coefficients jump velocity set jump velocity value. All right. So we might be at a place where we can see this. [music] So if we if we look at the jump velocity function then we might be able to see what we need to do here.
Copy. Copy. Copy.
All right. Next. Next. Here we go. Good thing I rarely eat lettuce. Right.
Danger zone. A little bit of danger zone. You love to watch more but I have to sleep. All right. Rex Xviper, thank you for hanging out with us. Good to have you here.
Okay, so let's see here. All right, set interval. Start jump. That's our thing.
We don't need to worry about that. All right, update jump. What we could do is see what start jump looks like, right?
So, start jump here. Oh, there we go.
All right. One second. Oh, >> different view.
>> Hey. Hey guys. Uh, I'll jump on here a second.
All right.
You have the wrong camera.
>> No, it's it's the right camera.
Once >> I'll be I'll be right back. Yeah, I'm streaming. I'm streaming. I'm going to end up stream here.
>> All right. There we go. [laughter] >> All right.
I know.
>> Yeah.
>> All right.
>> Cool.
>> All right. Hey guys, thank you so much for joining. I'm headed on now.
We'll be back tomorrow.
>> All right, one second. Let me let me m mute that volume here a second. Okay.
Uh, thank you so much for joining. Had a lot of fun today. We'll be back tomorrow. We'll be doing more reinforcement learning with our AI model and the dino jump game. And there we go.
Good to have you here. Have fun everybody. Have a good rest of your day.
See you guys tomorrow. I got to go to my meeting. I got to jump to the meeting.
See you guys. Thank you so much for joining. back tomorrow with more AI, more game. Thank you guys. Can I have mod at least? Oh, ask on Discord. Ask on Discord. You should be able to ask for on Discord. Adam, hi Adam. You joined right at the last second. We're wrapping up now. All right. Bye everybody. Have a good rest of your day. See you guys tomorrow. Bye everybody. Thank you. Uh Twitch, we need to also do a Oh, wait one second. You said um it's been a pleasure everyone. Have a great rest of your day. See you guys tomorrow. Thank you. I don't have it. All right. Bye everybody. See you guys tomorrow.
All
Related Videos
12:08
Artificial Intelligence
Expanding Stikbot thumbnails
leopoldshorts
2K views•2023-09-24
32:31
Artificial Intelligence
Digital Discrimination: Cognitive Bias in Machine Learning
redmonktechevents2974
4K views•2019-12-18
1:22:05
Artificial Intelligence
Evolutionary Approach to Clustering by Ujjwal Maulik
ICTStalks
279 views•2019-06-26
55:30
Artificial Intelligence
Rose Yu "Learning from Large-Scale Spatiotemporal Data"
networkscienceinstitute
2K views•2019-03-04
28:24
Artificial Intelligence
Stanford Seminar - Generalization through Task Representations with Foundation Models
stanfordonline
4K views•2025-07-14
8:29
Artificial Intelligence
Satellite-Based Wheat Yield Forecasting using GEE & Transformer Neural Network
gisrsinstitute
634 views•2025-06-15
32:06
Artificial Intelligence
Paradigm Shifts in Data Processing for the Generative AI Era: Robert Nishihara of Anyscale & Ray.io
GradientFlow
2K views•2025-01-02
39:48
Artificial Intelligence
How to Build Your Own GenAI-Based Knowledge Management System
2150GmbH
360 views•2025-06-03
Trending
13:38
Biology
Gremlin Arrives… While Dorothy May Takes Another Step Forward
The-moons
10K views•2026-07-23
13:25
Business
Playstation NO DISC/NO BUY Fight Is Over...
DavidJaffeGames
4K views•2026-07-23
16:52
Anthropology
Americans Confused in Australia for 17 Minutes Straight
IWrocker
17K views•2026-07-23
15:34
Political Science
FURIOUS Raskin CORNERS DOJ over Trump DARK PAST!!!!
MeidasTouch
237K views•2026-07-23