[
  {
    "startMs": 1309,
    "text": "[music]"
  },
  {
    "startMs": 13040,
    "text": "Hi everybody."
  },
  {
    "startMs": 15200,
    "text": "Welcome."
  },
  {
    "startMs": 16960,
    "text": "For those that came to actually see me,"
  },
  {
    "startMs": 19840,
    "text": "thank you. For those that are just"
  },
  {
    "startMs": 21760,
    "text": "hanging out here, also thank you. I'm"
  },
  {
    "startMs": 23680,
    "text": "going to try to entertain you guys,"
  },
  {
    "startMs": 25600,
    "text": "teach you something new."
  },
  {
    "startMs": 28000,
    "text": "Um again my name is Rafael. I work at"
  },
  {
    "startMs": 31279,
    "text": "Bright Data. I've been with Brighta for"
  },
  {
    "startMs": 33040,
    "text": "over eight years. And in Bright Data we"
  },
  {
    "startMs": 35440,
    "text": "are collecting data and uh we're trying"
  },
  {
    "startMs": 37760,
    "text": "to innovate. So today I want to talk"
  },
  {
    "startMs": 40000,
    "text": "about video discovery for agentic world"
  },
  {
    "startMs": 42879,
    "text": "models training. And uh we're going to"
  },
  {
    "startMs": 45440,
    "text": "explain what that is in a little bit."
  },
  {
    "startMs": 48160,
    "text": "So"
  },
  {
    "startMs": 49680,
    "text": "what does it take to take to build a"
  },
  {
    "startMs": 52160,
    "text": "system that can see understand and act"
  },
  {
    "startMs": 56480,
    "text": "right? I mean, we figured out what an AI"
  },
  {
    "startMs": 60239,
    "text": "is. We're constantly improving it. Like,"
  },
  {
    "startMs": 62480,
    "text": "you know, that's kind of handled. But"
  },
  {
    "startMs": 64640,
    "text": "now, the hard part is what data do you"
  },
  {
    "startMs": 67439,
    "text": "provide to it? Because everybody knows"
  },
  {
    "startMs": 70159,
    "text": "AI without data is just a box, right?"
  },
  {
    "startMs": 75680,
    "text": "So, let me take you a little bit back in"
  },
  {
    "startMs": 77759,
    "text": "history, right? In 2022,"
  },
  {
    "startMs": 80799,
    "text": "Google taught a robot by images, real"
  },
  {
    "startMs": 85600,
    "text": "world actions, right? Then going"
  },
  {
    "startMs": 88400,
    "text": "forward, there was a leap uh in 2023. Uh"
  },
  {
    "startMs": 92320,
    "text": "we had uh an AI actually control"
  },
  {
    "startMs": 95280,
    "text": "multiple robots."
  },
  {
    "startMs": 97360,
    "text": "In 2024, the open source came out. Now,"
  },
  {
    "startMs": 100240,
    "text": "this where the game changed. Once we had"
  },
  {
    "startMs": 102560,
    "text": "an open source, all the labs had access"
  },
  {
    "startMs": 105600,
    "text": "to AI and they started already putting"
  },
  {
    "startMs": 108960,
    "text": "AI into robots, right? And so now these"
  },
  {
    "startMs": 112799,
    "text": "models are already driving humanoid"
  },
  {
    "startMs": 115840,
    "text": "robots. I'm you seeing some of course"
  },
  {
    "startMs": 118640,
    "text": "most here are remote controlled, but I"
  },
  {
    "startMs": 121759,
    "text": "mean in San Francisco you got weo"
  },
  {
    "startMs": 123600,
    "text": "driving without any drivers. How cool is"
  },
  {
    "startMs": 126560,
    "text": "that? First time I got in the car, I"
  },
  {
    "startMs": 129280,
    "text": "thought I was going to die, but it was"
  },
  {
    "startMs": 130879,
    "text": "actually very nice. Right now, how do"
  },
  {
    "startMs": 134400,
    "text": "you think it learned how to drive? By"
  },
  {
    "startMs": 137680,
    "text": "actually teaching itself. By by learning"
  },
  {
    "startMs": 140000,
    "text": "on dashboard cameras, right? Every car"
  },
  {
    "startMs": 143040,
    "text": "has a dashboard. So, if you provide that"
  },
  {
    "startMs": 145920,
    "text": "to an AI, it can figure out how to"
  },
  {
    "startMs": 148239,
    "text": "actually operate a machine."
  },
  {
    "startMs": 151840,
    "text": "So, as I said, the AI is no longer the"
  },
  {
    "startMs": 153680,
    "text": "hard part. The data is. And this is what"
  },
  {
    "startMs": 156800,
    "text": "I want to talk about. Right? So for chat"
  },
  {
    "startMs": 162080,
    "text": "LLMs, there's trillions of words and"
  },
  {
    "startMs": 164720,
    "text": "texts, right? So we have so much text"
  },
  {
    "startMs": 167440,
    "text": "data to train the LLMs. For image"
  },
  {
    "startMs": 170480,
    "text": "generation, we got billions of images,"
  },
  {
    "startMs": 173040,
    "text": "labeled images. So that's also a huge"
  },
  {
    "startMs": 175599,
    "text": "database."
  },
  {
    "startMs": 177120,
    "text": "But for robotics, there's only about a"
  },
  {
    "startMs": 179440,
    "text": "million videos. It's a very small"
  },
  {
    "startMs": 182720,
    "text": "limited data sets of robots doing"
  },
  {
    "startMs": 185760,
    "text": "things, right?"
  },
  {
    "startMs": 189760,
    "text": "And so the idea here is that of course"
  },
  {
    "startMs": 194959,
    "text": "you a lot of companies out there what"
  },
  {
    "startMs": 196800,
    "text": "they do is they pay people to record"
  },
  {
    "startMs": 199440,
    "text": "actions, right? So for example, hey"
  },
  {
    "startMs": 202239,
    "text": "record me how you open a door or record"
  },
  {
    "startMs": 205360,
    "text": "to me how you're sitting on the chair."
  },
  {
    "startMs": 208400,
    "text": "But then the problem becomes is that"
  },
  {
    "startMs": 210560,
    "text": "when somebody is told to do something,"
  },
  {
    "startMs": 213040,
    "text": "they do not do it naturally. So I call"
  },
  {
    "startMs": 215519,
    "text": "it instructed, right? If you're told to"
  },
  {
    "startMs": 218080,
    "text": "record how you open the door and you're"
  },
  {
    "startMs": 220000,
    "text": "doing it for somebody, it's not going to"
  },
  {
    "startMs": 221519,
    "text": "be the same as if you just walk into the"
  },
  {
    "startMs": 223360,
    "text": "house. The movement is totally"
  },
  {
    "startMs": 225040,
    "text": "different. So is that data good for"
  },
  {
    "startMs": 228000,
    "text": "robotic training? In my assumption, no."
  },
  {
    "startMs": 230799,
    "text": "It's biased data and it doesn't deliver"
  },
  {
    "startMs": 232879,
    "text": "the same results. So obviously the robot"
  },
  {
    "startMs": 235680,
    "text": "is not going to be as effective as if"
  },
  {
    "startMs": 237760,
    "text": "it's intuitive. Right? So I just threw"
  },
  {
    "startMs": 240879,
    "text": "in some examples on the slide, right? So"
  },
  {
    "startMs": 243920,
    "text": "uh"
  },
  {
    "startMs": 246640,
    "text": "the built by hand data is not the same"
  },
  {
    "startMs": 248640,
    "text": "as the data that is kind of on the site,"
  },
  {
    "startMs": 251360,
    "text": "right? So and also the fact is that"
  },
  {
    "startMs": 254319,
    "text": "people behave very differently when they"
  },
  {
    "startMs": 256079,
    "text": "on the camera. Some of you know some"
  },
  {
    "startMs": 257919,
    "text": "people are camera shy. So you know as"
  },
  {
    "startMs": 260479,
    "text": "soon as you put a camera on them you"
  },
  {
    "startMs": 262160,
    "text": "they change but in the natural world"
  },
  {
    "startMs": 265680,
    "text": "it's a totally different thing. So this"
  },
  {
    "startMs": 267919,
    "text": "is what we are trying to solve in bright"
  },
  {
    "startMs": 269600,
    "text": "data right"
  },
  {
    "startMs": 272320,
    "text": "um"
  },
  {
    "startMs": 275280,
    "text": "again so why is the usual data sources"
  },
  {
    "startMs": 277120,
    "text": "are not enough right simulations"
  },
  {
    "startMs": 279600,
    "text": "uh virtual it's it's cheap but it's um"
  },
  {
    "startMs": 282880,
    "text": "you know everybody's playing video games"
  },
  {
    "startMs": 284639,
    "text": "the physics in video games are almost"
  },
  {
    "startMs": 286400,
    "text": "there but they're not good enough where"
  },
  {
    "startMs": 288639,
    "text": "to train a robot on it right a hand"
  },
  {
    "startMs": 291120,
    "text": "control the person controlling a robot"
  },
  {
    "startMs": 293520,
    "text": "it's doable but how many hours a day can"
  },
  {
    "startMs": 296320,
    "text": "you record the robot, how many people"
  },
  {
    "startMs": 298400,
    "text": "you need to actually get huge scale"
  },
  {
    "startMs": 300479,
    "text": "data, right? I mean, you can record"
  },
  {
    "startMs": 302479,
    "text": "maybe eight hours a day every day. How"
  },
  {
    "startMs": 305040,
    "text": "many hours are you going to get in a"
  },
  {
    "startMs": 306320,
    "text": "year? It's not really scalable."
  },
  {
    "startMs": 309840,
    "text": "And of course, there's already pre-built"
  },
  {
    "startMs": 311440,
    "text": "data sets, but as I said before, they're"
  },
  {
    "startMs": 313360,
    "text": "small. There's only about a million"
  },
  {
    "startMs": 314880,
    "text": "videos at this point. So, what is the"
  },
  {
    "startMs": 316720,
    "text": "alternative? The alternative is the web."
  },
  {
    "startMs": 320560,
    "text": "Let's just think about YouTube. How many"
  },
  {
    "startMs": 322720,
    "text": "videos are there on YouTube? I don't"
  },
  {
    "startMs": 325680,
    "text": "know, five billion videos. How many"
  },
  {
    "startMs": 328880,
    "text": "actions are in those videos that a robot"
  },
  {
    "startMs": 332720,
    "text": "can replicate?"
  },
  {
    "startMs": 334720,
    "text": "I mean, let's think about it. How many"
  },
  {
    "startMs": 336080,
    "text": "do you think videos are there of a"
  },
  {
    "startMs": 337919,
    "text": "person opening the door firsterson view?"
  },
  {
    "startMs": 342560,
    "text": "Millions of hours. You'll be surprised."
  },
  {
    "startMs": 346160,
    "text": "So,"
  },
  {
    "startMs": 350080,
    "text": "every day the web video shows gravity"
  },
  {
    "startMs": 352800,
    "text": "motions, right? So um cause and effect"
  },
  {
    "startMs": 356800,
    "text": "obviously accidents big the biggest"
  },
  {
    "startMs": 359199,
    "text": "cause and effect right how handle how"
  },
  {
    "startMs": 361360,
    "text": "people handle objects and billions of"
  },
  {
    "startMs": 363840,
    "text": "hours great training material for robots"
  },
  {
    "startMs": 367360,
    "text": "but what is the problem the problem is"
  },
  {
    "startMs": 369440,
    "text": "that there's a lot of noise right oh and"
  },
  {
    "startMs": 373440,
    "text": "for example right so one of the AI"
  },
  {
    "startMs": 375520,
    "text": "models that meta trained they gave it"
  },
  {
    "startMs": 377759,
    "text": "about a million hours of real world"
  },
  {
    "startMs": 380080,
    "text": "videos"
  },
  {
    "startMs": 381600,
    "text": "and then all it took is 62 hours of real"
  },
  {
    "startMs": 384400,
    "text": "robotics data to actually control to"
  },
  {
    "startMs": 387600,
    "text": "control a real robot. So if you think"
  },
  {
    "startMs": 390080,
    "text": "about it,"
  },
  {
    "startMs": 391759,
    "text": "the data was collected publicly, right?"
  },
  {
    "startMs": 393840,
    "text": "So it trained the the the robot on just"
  },
  {
    "startMs": 397039,
    "text": "random things and then in 62 robotic"
  },
  {
    "startMs": 400000,
    "text": "hours, it was already moving. So there"
  },
  {
    "startMs": 402960,
    "text": "no simulations needed, autonomous"
  },
  {
    "startMs": 405520,
    "text": "robotics delivered fast."
  },
  {
    "startMs": 412880,
    "text": "But the videos don't show how the robots"
  },
  {
    "startMs": 414720,
    "text": "move, right? So you might think, well,"
  },
  {
    "startMs": 416160,
    "text": "listen, how useful is it a person"
  },
  {
    "startMs": 418400,
    "text": "pouring water into a cup for a robot. So"
  },
  {
    "startMs": 422080,
    "text": "there is methods actually out there how"
  },
  {
    "startMs": 424880,
    "text": "an AI can distinguish and actually learn"
  },
  {
    "startMs": 427440,
    "text": "from the actions that are in a video. If"
  },
  {
    "startMs": 429919,
    "text": "you take two frames frame by frame and"
  },
  {
    "startMs": 432960,
    "text": "the AI measures the movement this"
  },
  {
    "startMs": 434960,
    "text": "difference, right? So you have an image"
  },
  {
    "startMs": 436560,
    "text": "A and image two and there's a small"
  },
  {
    "startMs": 438560,
    "text": "movement that is changing between the"
  },
  {
    "startMs": 440560,
    "text": "two images and AI can actually measure"
  },
  {
    "startMs": 443440,
    "text": "that change. So if you keep doing that"
  },
  {
    "startMs": 446160,
    "text": "through the whole video and AI can"
  },
  {
    "startMs": 448240,
    "text": "actually figure out angles, distance and"
  },
  {
    "startMs": 451440,
    "text": "everything that it needs to train and"
  },
  {
    "startMs": 453199,
    "text": "robot right. So we don't necessarily"
  },
  {
    "startMs": 456240,
    "text": "need the sensor's data but of course the"
  },
  {
    "startMs": 459919,
    "text": "video itself that you download from"
  },
  {
    "startMs": 461440,
    "text": "video from YouTube or the video that uh"
  },
  {
    "startMs": 464800,
    "text": "you extract the data from doesn't"
  },
  {
    "startMs": 466400,
    "text": "necessarily get you 100% there. You do"
  },
  {
    "startMs": 468560,
    "text": "need to have some processing on top of"
  },
  {
    "startMs": 470080,
    "text": "that but"
  },
  {
    "startMs": 472160,
    "text": "it's available. It's out there and all"
  },
  {
    "startMs": 475039,
    "text": "you need to do is process it."
  },
  {
    "startMs": 478560,
    "text": "So a few more examples right? So for"
  },
  {
    "startMs": 480400,
    "text": "example, Nvidia when they train in the"
  },
  {
    "startMs": 483039,
    "text": "Cosmos robot, they're throwing out about"
  },
  {
    "startMs": 485599,
    "text": "96% of the video. So what does that"
  },
  {
    "startMs": 488720,
    "text": "mean? They download a million hours of"
  },
  {
    "startMs": 491120,
    "text": "video and the only part is 4% is"
  },
  {
    "startMs": 494479,
    "text": "actually useful. Everything else gets"
  },
  {
    "startMs": 496400,
    "text": "thrown out. That's wasted compute."
  },
  {
    "startMs": 498400,
    "text": "That's wasted bandwidth. That's wasted"
  },
  {
    "startMs": 500479,
    "text": "storage. That's just a lot of wasted"
  },
  {
    "startMs": 502479,
    "text": "money. Okay. Stable video diffusion is"
  },
  {
    "startMs": 505199,
    "text": "throwing out 74% of the videos that"
  },
  {
    "startMs": 508080,
    "text": "they're downloading."
  },
  {
    "startMs": 510240,
    "text": "So in bright data we are trying to solve"
  },
  {
    "startMs": 513599,
    "text": "that problem and we're trying to take it"
  },
  {
    "startMs": 516240,
    "text": "a step forward. So what we propose is"
  },
  {
    "startMs": 519279,
    "text": "search first collect second. What we do"
  },
  {
    "startMs": 522240,
    "text": "is we are actually indexing videos and"
  },
  {
    "startMs": 525920,
    "text": "we are allowing you to search for"
  },
  {
    "startMs": 528399,
    "text": "specific actions. Since we're talking"
  },
  {
    "startMs": 531040,
    "text": "about a person opening a door, let's"
  },
  {
    "startMs": 532880,
    "text": "keep using that uh example. What you can"
  },
  {
    "startMs": 536399,
    "text": "do on our platform is actually input"
  },
  {
    "startMs": 539920,
    "text": "detailed information, right? A query,"
  },
  {
    "startMs": 542399,
    "text": "person washing dishes, person folding a"
  },
  {
    "startMs": 544800,
    "text": "t-shirt and so on. And we will provide"
  },
  {
    "startMs": 548320,
    "text": "the snippets from the videos of these"
  },
  {
    "startMs": 551920,
    "text": "actions."
  },
  {
    "startMs": 554000,
    "text": "So what does that mean?"
  },
  {
    "startMs": 556640,
    "text": "That means that"
  },
  {
    "startMs": 559279,
    "text": "there's less waste of data collection."
  },
  {
    "startMs": 562720,
    "text": "there's less waste of storage and"
  },
  {
    "startMs": 564720,
    "text": "bandwidth, right? So, a little more on"
  },
  {
    "startMs": 566320,
    "text": "how it works. Obviously, you define what"
  },
  {
    "startMs": 568880,
    "text": "you're looking for."
  },
  {
    "startMs": 572320,
    "text": "We search through billions and billions"
  },
  {
    "startMs": 574480,
    "text": "of pre-indexed videos, not by keywords,"
  },
  {
    "startMs": 578000,
    "text": "but by actions, and we provide you with"
  },
  {
    "startMs": 581680,
    "text": "ready to use clips."
  },
  {
    "startMs": 584080,
    "text": "And obviously, they're already trimmed,"
  },
  {
    "startMs": 586080,
    "text": "prepared for your training. So all you"
  },
  {
    "startMs": 587839,
    "text": "got to do is process them a little bit"
  },
  {
    "startMs": 589519,
    "text": "for maybe motion distance sensors and so"
  },
  {
    "startMs": 592320,
    "text": "on and it's ready to be ingested into a"
  },
  {
    "startMs": 595360,
    "text": "robot."
  },
  {
    "startMs": 597839,
    "text": "I have a small example of a video of how"
  },
  {
    "startMs": 600080,
    "text": "it actually works on our platform. So"
  },
  {
    "startMs": 602800,
    "text": "let me just hit play here if I can find"
  },
  {
    "startMs": 605120,
    "text": "it."
  },
  {
    "startMs": 607360,
    "text": "So here we see person washing dishes by"
  },
  {
    "startMs": 609519,
    "text": "hand in sink removing grease from the"
  },
  {
    "startMs": 611360,
    "text": "plate using sponge and soap including"
  },
  {
    "startMs": 614240,
    "text": "rinse and placing dishes into drying"
  },
  {
    "startMs": 616000,
    "text": "rack. close-up hand interactions."
  },
  {
    "startMs": 619519,
    "text": "So, right now what is happening is it's"
  },
  {
    "startMs": 621200,
    "text": "searching through"
  },
  {
    "startMs": 623360,
    "text": "billions of videos. Now, on this video,"
  },
  {
    "startMs": 625760,
    "text": "it's a bit old. We only got about 100"
  },
  {
    "startMs": 627519,
    "text": "million index there, but now we're up to"
  },
  {
    "startMs": 629360,
    "text": "1.1 billion videos. And it will"
  },
  {
    "startMs": 631839,
    "text": "literally like this is a little sped up,"
  },
  {
    "startMs": 633440,
    "text": "so I didn't want to waste your guys"
  },
  {
    "startMs": 634880,
    "text": "time, but it will provide to you clips"
  },
  {
    "startMs": 638000,
    "text": "from videos of people washing dishes."
  },
  {
    "startMs": 641440,
    "text": "And you can see right here all the"
  },
  {
    "startMs": 643920,
    "text": "videos in"
  },
  {
    "startMs": 646320,
    "text": "That would be fine. Now, some of these"
  },
  {
    "startMs": 648000,
    "text": "videos might not have to do anything"
  },
  {
    "startMs": 649600,
    "text": "with washing dishes. They might have"
  },
  {
    "startMs": 651279,
    "text": "cleaning kitchen. They might have to do"
  },
  {
    "startMs": 653360,
    "text": "something else. And here's another"
  },
  {
    "startMs": 655440,
    "text": "example. For example, human folding"
  },
  {
    "startMs": 657360,
    "text": "different types of clothes,"
  },
  {
    "startMs": 659600,
    "text": "right? So, a little more description."
  },
  {
    "startMs": 661839,
    "text": "The more description you give, the more"
  },
  {
    "startMs": 663760,
    "text": "you actually get back."
  },
  {
    "startMs": 666959,
    "text": "So, again, a little speed up. Let's see."
  },
  {
    "startMs": 674480,
    "text": "I want you to understand it could be"
  },
  {
    "startMs": 675760,
    "text": "anything. A person putting on makeup."
  },
  {
    "startMs": 677680,
    "text": "Let's say you have a brand and you want"
  },
  {
    "startMs": 678959,
    "text": "to find videos where your brand is being"
  },
  {
    "startMs": 680800,
    "text": "found. All that can be also provided. So"
  },
  {
    "startMs": 684240,
    "text": "you see people folding clothes. So now"
  },
  {
    "startMs": 687200,
    "text": "you can train a robot on how to fold"
  },
  {
    "startMs": 689200,
    "text": "clothes."
  },
  {
    "startMs": 694560,
    "text": "Of course everything is available via"
  },
  {
    "startMs": 696160,
    "text": "API. We don't expect people to actually"
  },
  {
    "startMs": 697839,
    "text": "do anything manually, right? So you can"
  },
  {
    "startMs": 700079,
    "text": "trigger it by API. you get a returned"
  },
  {
    "startMs": 703040,
    "text": "snippet of the video, the URL. We give"
  },
  {
    "startMs": 706880,
    "text": "you the link to the original video if"
  },
  {
    "startMs": 708640,
    "text": "you want to watch it. But the main thing"
  },
  {
    "startMs": 710240,
    "text": "here is that"
  },
  {
    "startMs": 712800,
    "text": "we can give you the snippets of the"
  },
  {
    "startMs": 714880,
    "text": "video to train your robotics. Now, I'm"
  },
  {
    "startMs": 718240,
    "text": "not sure if any of you training any"
  },
  {
    "startMs": 719839,
    "text": "robots, so I'm going to give you another"
  },
  {
    "startMs": 722560,
    "text": "example of how useful it could be. Let's"
  },
  {
    "startMs": 725360,
    "text": "say you have a brand and you want to"
  },
  {
    "startMs": 727440,
    "text": "know where your brand is being demoed on"
  },
  {
    "startMs": 731680,
    "text": "videos."
  },
  {
    "startMs": 733839,
    "text": "The video titles doesn't matter because"
  },
  {
    "startMs": 735519,
    "text": "it doesn't actually consume. It's not"
  },
  {
    "startMs": 737839,
    "text": "about your product. It's just a person"
  },
  {
    "startMs": 740480,
    "text": "doing some podcast recording something."
  },
  {
    "startMs": 743040,
    "text": "But you see a woman putting on makeup"
  },
  {
    "startMs": 745680,
    "text": "and she you see on the table that this"
  },
  {
    "startMs": 747839,
    "text": "is your brand of a makeup. So now all of"
  },
  {
    "startMs": 750720,
    "text": "a sudden you can find all videos for"
  },
  {
    "startMs": 752560,
    "text": "your brand where it's being used"
  },
  {
    "startMs": 754560,
    "text": "whatever it is right by searching name"
  },
  {
    "startMs": 757600,
    "text": "of the video you won't find it by by"
  },
  {
    "startMs": 760240,
    "text": "video indexing you can actually find"
  },
  {
    "startMs": 762560,
    "text": "specific things that you are interested"
  },
  {
    "startMs": 764880,
    "text": "in apple falling from the tree and so"
  },
  {
    "startMs": 768160,
    "text": "on. So what comes back? We provide to"
  },
  {
    "startMs": 771279,
    "text": "you this time stamp. We provide to you"
  },
  {
    "startMs": 773120,
    "text": "the score matching, how close it is to"
  },
  {
    "startMs": 774959,
    "text": "your query, and of course the frame"
  },
  {
    "startMs": 777440,
    "text": "count. How many frames are about what"
  },
  {
    "startMs": 780880,
    "text": "you're looking for?"
  },
  {
    "startMs": 784079,
    "text": "So what does this change?"
  },
  {
    "startMs": 786399,
    "text": "This changes a lot. Okay, less waste."
  },
  {
    "startMs": 789760,
    "text": "You don't need to download million of"
  },
  {
    "startMs": 791760,
    "text": "hours of videos. you can actually get"
  },
  {
    "startMs": 793920,
    "text": "exactly the the specific actions that"
  },
  {
    "startMs": 797519,
    "text": "you are interested in."
  },
  {
    "startMs": 801040,
    "text": "Again, it's very crucial for robotic"
  },
  {
    "startMs": 803920,
    "text": "training because noise creates problems,"
  },
  {
    "startMs": 807440,
    "text": "hallucinations, right? This removes the"
  },
  {
    "startMs": 810880,
    "text": "noise."
  },
  {
    "startMs": 814160,
    "text": "What is it useful for? Not only robots,"
  },
  {
    "startMs": 816720,
    "text": "but obviously self-driving for example,"
  },
  {
    "startMs": 818880,
    "text": "Wimo, right? trained on dash cam video"
  },
  {
    "startMs": 822720,
    "text": "cameras and there's millions, hundreds"
  },
  {
    "startMs": 825120,
    "text": "of millions of hours on YouTube of dash"
  },
  {
    "startMs": 827600,
    "text": "cams. How many of you like to watch"
  },
  {
    "startMs": 829519,
    "text": "accidents? Dash cam c accidents. I'm"
  },
  {
    "startMs": 832399,
    "text": "sure everybody watched them. Crazy"
  },
  {
    "startMs": 834399,
    "text": "driving in Russia, right? So, this is"
  },
  {
    "startMs": 837839,
    "text": "what we're talking about, right? The"
  },
  {
    "startMs": 839120,
    "text": "videos that are out there, okay? A"
  },
  {
    "startMs": 841920,
    "text": "person running a red light, person"
  },
  {
    "startMs": 843279,
    "text": "stopping on red light, making a left"
  },
  {
    "startMs": 845279,
    "text": "turn, making a right turn, and so on."
  },
  {
    "startMs": 847120,
    "text": "All of that can be a very useful data"
  },
  {
    "startMs": 850160,
    "text": "for training and any other world models."
  },
  {
    "startMs": 853760,
    "text": "Physics, right? Robots needs to"
  },
  {
    "startMs": 855680,
    "text": "understand physics. What is the gravity?"
  },
  {
    "startMs": 858480,
    "text": "How to sit? How to walk? How to move?"
  },
  {
    "startMs": 861839,
    "text": "Again, of course, you can pre-record it."
  },
  {
    "startMs": 864880,
    "text": "Um, a lot of times right now I speak to"
  },
  {
    "startMs": 867040,
    "text": "some of the people and they have"
  },
  {
    "startMs": 868399,
    "text": "millions of people recording videos for"
  },
  {
    "startMs": 870320,
    "text": "them. Hey, can you guys record me how"
  },
  {
    "startMs": 872160,
    "text": "you open doors? That's crazy. Why do you"
  },
  {
    "startMs": 875440,
    "text": "need a million people doing it when"
  },
  {
    "startMs": 877199,
    "text": "everything is available online? Online"
  },
  {
    "startMs": 880399,
    "text": "is one of the hugest databases out"
  },
  {
    "startMs": 882160,
    "text": "there. It's just people don't really"
  },
  {
    "startMs": 884560,
    "text": "like it's it's not that easy to use it,"
  },
  {
    "startMs": 886480,
    "text": "right? I mean, right now the only search"
  },
  {
    "startMs": 888560,
    "text": "available to us is by the keyword of the"
  },
  {
    "startMs": 891600,
    "text": "name of the video."
  },
  {
    "startMs": 893920,
    "text": "And of course, one source, the whole"
  },
  {
    "startMs": 896320,
    "text": "public video web in one place. I mean,"
  },
  {
    "startMs": 900800,
    "text": "we got YouTube, we got Vimeo, we got so"
  },
  {
    "startMs": 902959,
    "text": "many different uh video providers out"
  },
  {
    "startMs": 905519,
    "text": "there, billions and billions of hours"
  },
  {
    "startMs": 909440,
    "text": "of training data just there waiting for"
  },
  {
    "startMs": 912000,
    "text": "you to grab it, collect it and process"
  },
  {
    "startMs": 914639,
    "text": "it. So basically this is a new product"
  },
  {
    "startMs": 916399,
    "text": "that we're doing a bright data right"
  },
  {
    "startMs": 918320,
    "text": "indexing videos so that you guys can"
  },
  {
    "startMs": 921040,
    "text": "find specific actions anything you can"
  },
  {
    "startMs": 924639,
    "text": "think about you know again it could be"
  },
  {
    "startMs": 926639,
    "text": "useful for brands uh as well it's not"
  },
  {
    "startMs": 929680,
    "text": "only for training data anything you can"
  },
  {
    "startMs": 932079,
    "text": "think of could be useful"
  },
  {
    "startMs": 935839,
    "text": "maybe you are a gamer and you want to"
  },
  {
    "startMs": 938720,
    "text": "know how do you beat this level but"
  },
  {
    "startMs": 941199,
    "text": "there's no exactly video like that so"
  },
  {
    "startMs": 943120,
    "text": "you can search okay people beating the"
  },
  {
    "startMs": 945360,
    "text": "level on the video game, right?"
  },
  {
    "startMs": 947760,
    "text": "Anything."
  },
  {
    "startMs": 949279,
    "text": "The world is yours."
  },
  {
    "startMs": 952240,
    "text": "So,"
  },
  {
    "startMs": 953759,
    "text": "I'm going to wrap it up. I don't know if"
  },
  {
    "startMs": 955120,
    "text": "you have any questions, but uh if you"
  },
  {
    "startMs": 957040,
    "text": "want to connect with me, talk about it"
  },
  {
    "startMs": 958560,
    "text": "more, feel free to to add me on"
  },
  {
    "startMs": 960800,
    "text": "LinkedIn. And um yeah, I'm going to"
  },
  {
    "startMs": 964560,
    "text": "thank you guys for your attention. And"
  },
  {
    "startMs": 967279,
    "text": "uh I'm at a bright data booth if you"
  },
  {
    "startMs": 969120,
    "text": "want to talk more about it. And thanks"
  },
  {
    "startMs": 972639,
    "text": "for coming."
  }
]