Facial Recognition Forces My Coworkers to Do Their Dishes

0:00 People aren't doing their dishes anymore. So, we're going to combine custom hardware and custom software to create a dishwasher. In order for this experiment to work, we have to have nobody know about this. What's going on? It don't honestly I I [Music] Sam, we got a problem. Big problem. Big problem. People aren't doing their dishes anymore. I don't know what happened. People just are leaving all their dishes in the sink. We've mentioned in meetings it's just not sticking that I thought what if we just put a camera there that's like who's going to watch the camera? We have a rule here at Corridor which is we don't secretly record people. If you're going to be on camera it's with their consent and you know people know they're being

0:45 filmed etc. So I got really into using cursor and Meta just dropped this new project a thing that lets you feed it a video file and it lets you ask questions about what happens in that video file. Sam, here's a clip from Honey, I blew up the kids. So, what do we got? We have like a We're in a laboratory. We're seeing like the shrink ray and this guy is creating a very fancy dinner in the laboratory. Let's feed this into the computer and let's ask it a question about that shot. Where is this shot taking place? Write me a poem about what happens in this video. Problem master. Two men sit at a table. One pours a glass of wine. The other answers a phone in a room with

1:30 a mashy. We could technically give this a video stream. We give this a series of videos. We could have this constantly be looking at things and constantly judging and analyzing what's happening in the video. That's crazy. What we're about to introduce here is top secret. So, here's how we're going to catch people leaving dirty dishes in the sink. There's a painting. There's a painting above the sink that's already there. It's about like a centimeter thick. You also got a Raspberry Pi recently. Yes. And that's less than a centimeter thick. Yeah. Behind the painting, we'll have a Raspberry Pi. Yeah. It's going to stream to a computer that will then only record when there's people in the kitchen doing stuff around the sink, which will save the clips to the folder. This computer will be listening to that folder or watching that folder. So when our clip arrives in that folder, this computer is

2:16 then going to make a judgment call at which point if it determines that somebody has left a dish in the sink and it's going to break down what they look like, what facial hair they have, what hair they have, what clothes they're wearing, and it'll make that into a message so we can get the message and go out there like, "All right, who has long hair?" It's a lineup and black hair and it's wearing a red shirt. So I actually took it upon myself to make a Discord server for this project already. Oh, is it # dirty ones? No, it's called Discord. So pretty good. So then it goes to our Discord server at which point we get the message. Oh, just uh just setting up a background here for our filming situation. Yeah. One second. We do have to keep this a secret in the studio. If anyone knows I mean the the experiment slash whatever this is fails. So we're basically

3:01 telling everyone that this video is actually about converting liveaction Disney movies back into cartoons. Why don't we do a screening that's specifically just for all the people who didn't do their dishes before? You guys are like spies. All right, we'll bring him down here and we'll sit down all together. It's like Oh, wait. Bren was Ren was in there the whole time. Was the door open? I I need to I need to make sure. Does he have his AirPods in? Did you have your headphones in? Yeah. Thank god he did. Yeah. What's going on? It's Don't honestly I I look I messed up. Don't worry about it. I'm going to pull. Can I come out? If if you need to come out, just let us know in

3:47 advance. But no particular That went really well. It was really good. He was not suspicious. We're pretty close. The camera's there and it is streaming. I've never done it before and there's no like graphic interface. It's just through text here. So, you know, you change it, you think it's working, you check some things. Okay, it's says it's working. And then the thing is like how do we get this which is streaming video to something to here to this computer? And there's always issues of like how big is the video file? What's the frame rate? What's the codec? Is our network blocking any of these things playing? That's it. It's working. My crystal ball is on. There it is. Oh, look at that. Oh, wow. This is the first time I've seen a feed ever. This is the first time I made my own like wireless camera. See,

4:34 look at this. I can walk around wherever I want. Yeah, look at that. Yeah. Okay, that's pretty sweet. What What's funny is how all these projects often begin with a knife. Uh the last vibe coding video that I was in was uh the crime detector, which I think actually worked really well. We had the same ethics. We We want to anonymize people. We're not going to be looking at the footage. We just want the computer to make the call for us. We're going to be using the perception model to actually give us a lot of the information. Ideally, it can detect some silverware. No, it can't detect the forks and knives. It can do the cup, which is nice. Basically, we need to create an intelligent motion detecting system here. Last time we were simulating crime. This time, the crime is real. That we're not simulating doing dirty dishes. Like, they are actually

5:20 dirty and people are actually not washing them. We don't want false positives and we want to make sure we eliminate any mistakes. Yeah, the I think the V V1 of the system is kind of working right now. Like the bones are working and I'd love to show you. We have a little camera plugged into here. It's going to send over a video signal via Wi-Fi. So then I took that streaming URL and I did some wacky stuff in cursor. We don't want to record all day. We only want to record when people are in the frame. Like how do we only select people that are operating at the sync? We don't just want strangers walking past in the background. We don't care about them. We only want to record people when they access a certain area. So, in order to do that, we now have a feature we can just open the video feed and click and drag. And now there's a little sector here that's being analyzed. It only analyzes that area. So, for instance, the sink, the counter,

6:06 etc. Yeah, that works. All right. Sweet. We got a video clip. Drop it in that folder, Sam. File detected. A file has been detected in the folder that this computer is looking at. Have any dishes been put into sink? No. A man is posing for a photo shoot. You told me to go pose in front of the camera. I did tell you to go. Yeah. Well, he's right. So, now here's the thing. I should be able to change these questions here. Is the man in this picture holding a firearm? So, the conditional should theoretically be a yes. If you save that video of you with a gun, it's file detected. Yes, he is holding a gun. And theoretically, I should get a Discord notification. Wait, what was that? I heard the Discord's on the book. Guys, we got a fresh dish alert here. Wait, I got a fresh dish alert, too. Uh, the person in

6:53 this image is holding a dish. Uh, oh my god, it's hard evidence. It's literally It's so funny. It's such good testimony. Well, that's perfect. So, the system works. I think we need to put this camera in the painting and begin detecting this these these criminals. Um, it's ready. This has been hanging over the sink for years. Today, we'll give it even a newer purpose. The goal is to make the smallest hole possible. And honestly, like something like in here actually kind of work. Relatively inconspicuous, right? You see it? You see it's on the edge. Wow. Oh, that's smart. I didn't notice that at first. It actually kind of works. It actually kind of works. Okay. Yeah,

7:38 we're good. We're good to go. Let's just lay this down this way. I'm nervous. My heart's beating. I mean, we didn't have a whole lot of time to prep on this. This up and everybody finds this and we don't have a video. Ideally, it's as simple as just taking it, walking up to the wall, pressing it, and then walking away. I'm going to go check if the coast is clear. Okay. Um, that's not sick. That had to have been glued. Is that a go? Uh, yes. Somehow I think it just suctioned really well. [Music] The artificial mind is awakening at which you will perceive these videos so that we do not have to. It is under our control. I promise. All right. So, I

8:26 have some dishes. Let's just do a test. I'm just hanging out. I have a time to do other things than worry about the dishes now. I just got a Discord notification. Sam, did you get that Discord notification? Somebody put a dish in the sink and uh it was a green plastic dish and that person is a man with a ponytail wearing a sleeveless shirt with camouflage pattern. He has a beard and he walked into the kitchen, placed a green bowl in the sink and he walked out of the kitchen like aing Are you serious? Wait, it's working. It's working. It's just let it cook. Let it cook. Now we can actually do other things. Aside from a few quirks, the system is working great. Now we sit back and let it run for the rest of the week. Let's see who we catch.

9:11 All right, it's day three. Got to swap the batteries. So, first off, it's great. No one noticed the thing. So far, we we have one perpetrator. Using deduction, it wasn't hard to figure out who the one perpetrator was. Well, I sure hope that nobody saw the camera cuz you could kind of see it sticking through the mesh a little bit. Yeah. Uh that would be really bad if someone noticed this. Oh my god. Hey guys, Christian, did you um see something that you weren't supposed to? Well, it seems like I wasn't supposed to. I was minding my business, being the responsible employee, and washing my dishes. And something just caught my eye. Little mesh square. What is that? And I, no one was around, but I just

9:56 decided to lift it up and I just hear Velcro. I look under it, and there's just a computer chip, a battery, and some wiring. And I kind of just went and put it back. I have zero context outside. Christian, can we go talk to Let's just go. Yeah, we need to do this now. So, I just discovered that the guys actually built a robot to narc. It's a little intense, but hey, around here, we're all about building systems that keep things running smoothly. This video is sponsored by Chime, which is doing for your money what we're trying to do for our kitchen. Make things work a little bit better with a little less stress. I'll be back for this. Chime is an award-winning financial app and secured credit card that helps you manage, save, and spend your money. The product I've been using is their secured Chime credit builder Visa credit card. It helps you build credit by using it on

10:42 stuff you're already buying. It's a super simple way to build credit using everyday purchases. No interest, no annual fees, and no credit check to apply. You just move money into your credit builder account, and that becomes your spending limit. And if you turn on something called safer credit building, it'll automatically pay off your balance on time every month with the money that you added into the account. So, you're just building credit by being a functional adult, you know, buying coffee, groceries, gas, maybe a dishwasher, whatever you need. So, if that sounds like something you can use and you want to start building credit today, check out Chime by hitting that link in the description below. So, while Chime is going to help you stay on top of your credit, let's see if this AI robot's going to keep the crew on top of their dishes. Um, I'll come back to that later. Honestly, I think it was all your imagination. We're done. Yeah, it's always been there. didn't

11:28 mention it to anybody else outside of this room. Here's Christian. Welcome to the inner circle. Nice. I'm going to send you an invite to our Discord. But there's a whole backlog here. What's going on? The big thing is here in order for this experiment to work. We have to have nobody know about this. Circle of trust. This is trust. I'm in. Just remember, we're not we're not watching it. Do anything. Feel comfortable at the sink. Do do change your habits at all. Good. Yeah, exactly. We're good. I'm all on board. [Music] Bad news. We've encountered a mysterious rare week where everyone's actually doing their dishes. Yesterday, we're all like, "Oh, we finally caught someone. We never confronted him." And it doesn't matter because whoever's dish it was

12:14 been cleaned. I'm simultaneously dreading and also hopeful that people don't do their dishes for the sake of this video. Okay. So, as you're walking to the office here, did he notice anything? Yeah, it was perfect. It was spotless all day. I thought like just the aura of the camera being there was making people behave better whether they knew it or not. What's up, Nick? The ghost. Boo. What do you work at? Can you give me a hint? Oh, wedding videos. Wedding videos. Yeah. You already know it's you know. So, this whole backstory or this the secret cover story we're having converting live action Disney movies into animation again. Nick thinks we're working on that. Okay. He's like trying to help us make a video we're not actually making. I think we're going to actually have to give Nick something to work on here to

13:00 get him off our backs. I've heard a lot of giggling coming from behind closed doors. This is the WAN video model WAN 2.1 that's been out. The 14 billion parameter one which barely fits on a GPU. This seems like it's the promised land for turning visuals back into cartoon. And I can just grab video of Will Smith in his performance. I can turn him into a cartoon genie for the first time ever. We measure the metric of visual advancement in AI with Will Smith videos. Okay, here we go. Come on, genie. Oh my gosh. Yes, it's a genie. Oh my gosh. The fact that he even did that is insane. That's like crazy. So, now we have Nick churning out weird animated genie shots, but there's still one more obstacle standing in our way. So, one huge problem we've been having is

13:46 everyone in the studio has been wearing the same clothes because we've only had a few merch drops this year. So, I've went out of my way to make some new merch just so that way this camera system doesn't get confused. And it's on sale right now at corridor digital.store. Look at that. The logo broken down to all the individual shapes. That's genius. Who came up with that? There's kind of a theme. I I came up with it, Matt. No longer do we all have to wear the same shirt, but we can start mixing it up. Yeah, they're only on sale for like a week or two. Go to Digital.store. So, yeah, as I said, to kind of break things up, I'm going to go hand these out to the rest of the studio. Well, but then we're all going to have the same shirt again. Oh. So, with Nick distracted, Christian in the inner circle, and new merch to break up matching outfits, I was confident we'd catch some criminals soon.

14:38 There's uh there's like two dishes in the sink right now. Do you know who did it? I think I have it right here. Did anyone put a dish in the sink? Yes, someone put a dish in the sink. He walks over to the sink and puts the frying pan and spatula in the sink. He then walks back to the table. The person is a male with a beard and a mustache wearing a t-shirt, black shorts, and black socks. Four of us wore the same t-shirt yesterday, which I thought was going to complicate things. And only one of them was wearing black shorts. It was Christian. What? Christian knows we're doing this. Well, we should talk to Christian. Obviously, this is going to be confusing. Uh, but there's also another dish in the sink. The man put a dish in the sink. It's a glass dish. It's colorless. The person is a man with short hair and no beard wearing a t-shirt and pants. Wait, that

15:25 could be Dean. No, Dean was wearing shorts on Monday. He made a whole thing about it. Oh, that's right. He did. I mean, there's only one other person. I mean, it's Ren. Let's Let's Yeah. Show them a little bit about what you've been working on. I've been hard at work trying to learn how I can turn liveaction footage into a cartoon using the power of neural networks. Last week, what do you guys think of these shots? What do you think of this shot? Design, I guess. No. Okay. All right. I hear you loud and clear. Evil, dude. G looks like LeBron in his 20th year. What do you think of this shot? Someone didn't do their dishes. Yeah, that's crazy. So, yeah, we've been trying to figure out who that was. I'm so confused now. So, no, but Ren, so no,

16:12 it says a man put a dish in the sink. It's a glass dish, colorless. He has short hair, no beard. He walks up to the sink, places a glass dish, and uh maybe a fork into the sink. He walks away. It It sounds like it's you. That description could be Ren. Is that clear dish yours? Well, probably. Okay, that's that's that's that's all we need to know for now. Yeah. No, the main thing is Ren. Yeah, I'm so lost. Ren, what about the genie? Why did he look so janky? The truth is, Ren, there is no Disney live action video. That was a cover story for us to work on this other video here. After installing this and building it out and getting it set up, we have it watched in the video feed. Wa. So, all we get is all we get is a text alert on our phone when a dish has been put in

16:58 the sink. I knew something was up. That's all I knew. And only because you guys were not chill about it. All I want to do right now is go up there and inspect. That's what I saw. You noticed that? What? This painting was made in 2017. Joel Lores. Thanks for the painting, Joel. Only one thing left to do. That means one dish left to do. When we implemented this, the studio went on the longest streak of people washing the dishes that we've ever encountered. The first person to leave a dirty dish in the sink since we implemented this watchful eye was seriously

17:44 that it exists. We were all fired up thinking, "Oh yeah, we were going to take down this crime syndicate. A big bust was going to happen." Now that we're kind of actually here in the room together, does it feels like the stakes are a little bit lower? You're finally in this position of, you know, being caught. Do you think this is Do you think this is an effective way to have people uh do their dishes, which I this type of interaction I hate more than anything? No, we're keeping this low stakes, which is why we showed you the funny cartoon thing. I don't know. The reason why I don't feel great about this is because I have taken the mental step of being like, I will be better at this. I will do my dishes, but I'm not perfect. So, here I am looking at the one of the times that I haven't done it. And I'm just like,

18:33 I've done it, too. I've left dirty dishes in the sink before. What if there's a way to like reward people instead of like shaming them? Exactly. We had an idea, but first we had to tell the crew everything. Wait, R. We should see if anybody can find it here. You guys want to come to the kitchen and see if you can spot it? Did you see it? Where is it, guys? Dude, I was like, how did this get off? I noticed this. It's bugged. It's bugged. Ren and Christian helped illuminate us that maybe this whole like trying to catch people and shame them into doing the dishes is not the best approach. What we're going to do is we're going to retool the system and we're going to make it so it awards people for doing the dishes so we can celebrate the clean

19:19 dishes in the clean kitchen. If we're going to do a reward system, a point system, that's so much more work. Now, we have to keep track of the people. It's going to be a camera with facial recognition. So, all we're using here is I've taken four pictures of each of us with the same webcam we're using. We need to line up everyone in the kitchen right now and take pictures of their faces in dishwashing conditions. So, here's how the new system works. We have a camera streaming video over Wi-Fi, which then is run through YOLO V8 to detect people in front of the sink. Those clips are stored and processed through Deep Face to do facial recognition. If someone is recognized in a video clip, we rename the video file using their name. Then it goes out of my PC into a tube and out into Nico's PC,

20:05 at which point it passes through the perception language model, which judges if dishes are washed. Then that passes into a duplicate detector to see if that person was already caught washing that dish earlier within 5 minutes. If no duplicate is detected, it awards one point to that person, which updates a JSON file, and that gets posted to Discord to let people know who's winning. This is crazy. This is absolutely insane. This whole thing kind of works. Now, there's only one thing left to do. Let's run this the whole week and see who the best dishwasher is. I saved this from lunch yesterday until the system was operational. I brought dirty dishes from home. Literally saved this from yesterday. You know, at the very beginning of this video, we were thinking, "Oh yes, we're going to find the guilty people. We're going to take them out." You know, like it's like an FBI raid on the mafia. We

20:50 came into it with all that force over one dish. It was It was definitely awkward. It's I think it's easy to like fall into the mental trap of like thinking that like there's bad guys to catch, right? When in reality it's like 99% of people at all times are trying to do their best and we all forget doing our dishes. Making a way to change your attitude towards it to make it something you want to do is really the best way to approach something like this. And we still got to do a bunch of sick programming which is really why we made this video. Yeah.