Will AI Take Over Humanity ? | Ben Relles

Ben Relles built a digital twin of Reid Hoffman using AI video and voice tools that can answer questions by drawing on years of his podcasts and posts. He shows how AI can democratize video production and revive scripted storytelling, but warns against AI slop and deepfakes. Relles lists dozens of tools creators can use now, from GPT to Runway to Midjourney.
Chapters
When you think about video, you think about how do I get so many views? How do I make the money? How do I optimize my thumbnails? I actually think AI will enable everybody to be more creative, not just creators, and we're all creative people. Oh wow, it's so good to see you guys in Dubai. Thank you very much. Um, I am very excited to be here. We are talking about something fascinating, the role of AI and the impact on creators. And in a way, this is the most transformational, exciting moment I've had in my career as a creator and in the creator economy. This could completely transform the way people can be creative. And then on the other hand, it's a little terrifying. The creator economy is built on social connection and real people. So what does it mean when you can have digital twins and create people out of nothing? So with that, I am going to cover a little bit about what Read AI is, a little bit about my theories about where things are going, how we made Read AI, and then we'll talk about some tips and tricks for creators, all in 20 minutes or so. All right, so let's get started. Read AI. Am I on slide 1? Are we in the— okay, well, let's go back a slide. Yep, here we go. We're going just fine. Read AI is something that I created with Reid and a couple other people about 8 months ago. And the idea was essentially to experiment with a digital twin of Reid Hoffman, the founder of LinkedIn, where everything he says is based on 20 years of podcasts, books, interviews. The video is all AI generated. The voice is AI generated. So with that, let's have Reid AI introduce himself to Dubai.
Thanks, Ben, and hello everyone. I'm so excited to be here today at the One Billion Followers Summit. And since we have such a global audience here today, I thought I'd try something different and introduce myself in a few languages. Let's start with French, Chinese, and then Hindi. Bonjour, ici Red AI. Je suis le jumeau numérique de Red Hoffman. W de xiǎngfālái zǐ y jiū réder xīnián dē xūjì, fāngtān yǎnjiang, bōkē hēnèi róngdì dīng jījīpētī.
Okay, so that is a quick intro from Reid. Again, for those of you that don't know, Reid is the founder of LinkedIn, but since then, he has gone on to be very involved in artificial intelligence, including a book he has coming out in 2 weeks called Superagency, all about AI and what could possibly go right with our AI future. As we think about video and AI, like I said, the just past 6 weeks has been pretty remarkable. All of this is AI-generated video. I'm sure many of you have seen it. This is all from, uh, v02 from Google. Uh, in addition to all of that, uh, video that are just like initially in the hands of creators as demos, we're already seeing what short films could look like in the hands of creators. Uh, I took the audio off of all of these. You can go look them up afterwards. I'll— get all your emails, send you all the links, but essentially this is a film, 12 minutes long, 100% created using VO2 in a matter of days instead of a matter of months or years like it would take film studios. Not completely there yet in terms of the realism, but still wild how far this has come from a year ago. All right, so let's first put on our optimistic hat. This could be incredible. AI, just like YouTube, democratized distribution, AI comes in and democratizes production value. And all of the early promise of YouTube, and I love YouTube, I worked there for 10 years, I was a creator for 5 years, but early YouTube, it felt like, wow, we're gonna have Oscar winners and TV shows on YouTube, and there were all these big creators doing sketches and short films and sci-fi stuff like Mystery Guitar Man and Freddie W. But then actually over time, And this is getting in the weeds, but YouTube made a switch where the algorithm favored watch time and some other things. It actually got really hard to make narrative scripted content, and most of the biggest creators on YouTube have a very social, vlogging, talking-to-the-camera kind of personality. Nothing against that. I love the YouTube community that's come together around that, but I'm curious what happens when these tools are in the hands of filmmakers and people that want to tell narrative stories. So best-case scenario, new community emerges of AI filmmakers who know how to use these tools and do things that used to require the archaic Hollywood system of pitching projects and taking notes and taking years. And we got Don up there in the upper right corner, he's in the crowd as well, but other brilliant filmmakers that are using these tools. Now, there's also the negative scenario. All of this AI filmmaking leads to AI slop, the internet is— clogged with so much content and so much nonsense that we can't even find good stuff anymore. I only want to say that because a lot of times when you hear speakers talk about AI, it's a little bit too optimistic. It's just a tool. It's going to make us all more creative. And I'll say the same about digital twins. As excited as I am about the category, I also, of course, think there's some real issues. So let's talk about digital twins. This is VO2. This is all AI-generated people. This isn't based on real people. This is 100% AI-generated. Already looks essentially like the kinds of vloggers you would see on YouTube. Now, there are really exciting things about this. Vloggers being able to maybe replicate themselves, being able to do content that, again, you couldn't do before. And you can learn things in ways that you couldn't learn without AI. But of course, there's the risk here where People felt like some of the early attempts at this were a little bit creepy, unnecessary. And what does this mean for the AI, you know, integration into YouTube and creators when the lifeblood of YouTube is real people that have real connections with their audiences? And very briefly, getting into my own story, I've been in this whole YouTube game for about 19 years. YouTube turns 20 next month. I started in 2006 making YouTube videos and immediately loved the small community that grew out of anybody being able to find an audience when you couldn't do that before. I got kind of inspired by lonelygirl15. I created a video called "I Got a Crush on Obama." It was a girl singing about Barack Obama. I called her Obama Girl, and it was one of the first viral videos of 2007 where it started with a single video, and then she ended up on Saturday Night Live, and there was political cartoons about her, and I basically quit my job to make YouTube videos. And for that alone, I have a lot of love for the YouTube community. So then I started this other channel, well, it's kind of a spin-off called The Key of Awesome. I got to build a creative writing team, and essentially, like, the cool part of YouTube, of course, is you, the people who make the videos. The tubes are cool too, but it's the you that makes YouTube special. So I guess then the question is, why, with this world of content creators where they get to have a direct relationship with your fan base, would we want to possibly experiment with digital, robotic-looking versions of these creators? Um, I think it's a very complicated question, but I will say in my case, part of it is because the cat's out of the bag. This is going to happen. And I really feel like there are some brilliant ways that we can utilize AI digital twins and AI filmmaking in general to not only enhance the creative process, but also to improve our lives. The original idea behind Read AI was that here you have this founder, investor, creator who has all of this knowledge, and wouldn't it be interesting if anybody could tap into that knowledge, not by reading sift, sift 6 different books, but by actually asking questions based on everything he said over his career. And he really embraced this idea when we started talking about it. So that's the backstory on Read AI. Um, I will ask him a question though. What do you think about this complex topic of AI and creativity?
Well then, you know, creators should think of AI not as a competitor but as a copilot. To make the most of that, deeply understand the tools and how they fit into your creative goals. Use AI as a force multiplier, not a replacement. Your originality is key. Take bold risks, iterate freely, and inject your voice at every step. Think long-term. AI isn't just about speed, it's about transforming storytelling, art, and audience connection. In fact, in the time it's taken me to answer this question, I've brainstormed over 10,000 new video ideas for you and me to create, Ben. Top contender: Reid AI and Ben do a rap battle. Thoughts?
All right, Read AI and Ben do a rap battle. So what are we watching there? Again, the answer comes from a GPT prompt of, what do you see for the future of AI and creativity? It taps into the way he talks, the mannerisms. I mean, he doesn't do all the stuff with his fingers in real life. It's not totally there, but it's using information he'd say. It's using his voice, and it's trained on a video we took of him. Let's see, Read AI. Who is your favorite person in the world?
My favorite person, that's tough. It's less about accomplishments and more about values. I admire anyone who champions innovation, resilience, and amplifying humanity through technology. Those are the people who really inspire me. What about you, Ben? Who inspires you?
Uh, let's go with my dad. My dad inspires me. His answer was terrible. If you had asked somebody their favorite person, "It depends on their accomplishments," nobody wants that. And I only say this to say that, yes, there are things that AI twins can do that are cool, speak languages, be 1,000 times smarter, but also I feel like ultimately they are not going to be interesting when it comes to life experiences and real connection. And I'm hoping for that, and I wanna lean into ways to use AI that again enhances our humanity. And ultimately for me, like my favorite thing to do, meet new people, ask interesting questions, learn about them. You can't do that with an AI. You can learn a lot of information. I think you're gonna learn all kinds of things that help you in life, but you can't do that. Let's talk a little bit about how we made Read AI and what's happening here behind the scenes. So we did enter in, he's, well, now he's written 6 books with Super Agency. He's got 4 podcasts, he's got LinkedIn posts. We're training an LLM model using GPTs to essentially create a Read AI that can pull from all of this different information. So whereas a regular ChatGPT conversation, you could say, what would Read AI think about this? It can't actually access all of these thousands of hours of information. We use Hour One for the video and we use ElevenLabs for the voice. So, Reid AI, if we ask a question, this is one version of Reid AI. We have a formal version, an informal version. We're playing with different personalities. But essentially, if I ask like, hey, you're a creator heading to Dubai, how do you approach networking with other people? The answer you get from Reid AI, at least this formal one, will give an answer on how you can actually meet new people at the conference, but it will tap into the principles specifically of a book he wrote about networking. So there's a lot of interesting specific ways that we can train this to have conversations with people based on what he knows, where you wouldn't have to take a long session reading a book finding the answer. He'll get it to you immediately. We use ElevenLabs for the voice. I've only got, I don't know, 6 minutes left, 7 minutes, so I don't want to go into all of the how you use every tool. They're pretty easy though. ElevenLabs, $5 a month. They train on your voice in a few minutes. and you can have a voice replica of yourself. And like I said, we use Hour One for the video. We do get an automated video as soon as we ask the question, but then Hour One does give you the ability after the fact to choose different camera angles, choose different settings, emotional styles, so you get the instant response, and then we can adjust it if we want. And then I'm gonna skip the original video, but the original video is basically, if you wanna watch it, Reid talking to himself, which is another use case here, where Reid will actually be able to sort of bounce ideas off a version of what he might say, but it might be based on things that he's thought 10 years ago and hasn't thought in a while. We had a meeting recently and I had Reid AI listening in on the meeting, and then at the end of the meeting, we could go to Reid AI and say, be as harsh as you can be. What are we missing? Why is this idea we're excited about a horrible idea? There's some use cases of digital twins where I actually think they can give really strong opinions that an individual might feel uncomfortable giving.
I'm an AI-generated version of Reid Hoffman, his digital twin.
Eh, you get the idea. We made a 2004 version of Reid. We had— I just want to give you some of the different things we're experimenting with. So we had the current Reid interview the 2004 Reid about the year he founded LinkedIn. And we actually had current Reid advise 2004 Reid on his business plan. We had Reid AI evaluate business plans at Stanford, all of the business Plans were entered in, and then we had Reid Aye do a live chat with the class. And essentially, we're trying to use all of these different cases to show in a space which is fraught with a lot of complicated issues, and we want to get all those things right, we also have some really interesting ways to experiment with how can we use a knowledge base that he's taken years and years to build and actually use that in ways that are helpful. Just to quickly go through the digital twin space, It's not just that one model. So our one, HeyGen, Synthesia, these are all using a very similar approach where you take a video and you essentially change the mouth to fit the words you want it to say. Lip dubbing, overdubbing. But there's other ways to do these twins. This is a deepfake. So you may have seen Tom Cruise on TikTok. Well, not Tom Cruise, but a version of Tom Cruise based on a much more complex deepfake where you can watch him run around. Like I said, Veo is creating digital creators where these people do not exist. They're only based on the idea of influencers. This is an example where I like feel like a lot of people, when you think about video, you think about how do I get so many views? How do I make the money? How do I optimize my thumbnails? I actually think AI will enable everybody to be more creative, not just creators. And we're all creative people. This is a tool that's very easy to use. Runway, where essentially you can take a photo, that's me and my dad, that's me and my dad around 1979, and then you can kind of like add video to any photo, and this is earliest stages, and so eventually, you know, I'll be able to see much more than a 4-second clip. But like I think you've heard a few times at the conference, AI is the worst it'll ever be. I mean, all technology is the worst it'll ever be, but AI specifically, it does feel like it's changing so rapidly. That every time you see something, you can jump ahead to where we're gonna be in 6 months. You can do kind of video to video. Here's an example of that, where his facial expressions can be duplicated on another face. And essentially, while we have this, like, incredible ability to duplicate ourselves, to create other people, I think we're at a really fascinating moment right now, where we'll shift from quick demos to very soon being able to have powerful, helpful experiences with AI that actually help you because the video can see things and do things it couldn't do before. Very quickly, I think the things which will get people over the hump to be comfortable talking to these digital twins, the video is getting more real, the real time is getting quicker and quicker. The demos I've been showing you have been sort of pre-generated to keep everything going quickly, but I'm sure you've seen like the speed is down to like a second of latency. Um, super agency, I mean, it's the name of Reid's book, but it's basically this idea that AI will give us the ability to do more. And multimodality, the ability for AI to be able to see us, see what's happening, help us in real time by actually being able to interact in different ways. So all of those things combined have me very excited about AI. Let's go one more question, Reid. Um, what are you excited about? What are some tips?
Great question, Ben. Here are a couple thoughts on that. Use AI to automate repetitive tasks like transcriptions, organizing like data or resizing assets, so you can focus your energy on the most creative aspects of your work. My second tip: leverage AI to analyze trends and audience feedback, helping you make data-informed decisions that enhance your content's reach and impact. Ben, do you have any other tips you would add to that?
Oh, of course I have other tips. Um, we're gonna go quickly through these. There's so many tools to explore. Um, I just picked 2 in these 5 different categories because I think the most important thing as a creator is to just start experimenting. So from an ideation standpoint, of course you can use GPT and just start chatting about, can you give me different ideas for videos? Can you help me think through this one video script and give me edits? I've been using GPT for everything, like most of you have. It's a great creator tool. Uh, Spotter Studio will be more advanced, $50 a month. It'll help you come up with thumbnails, titles, ideas. Um, these are— I'm just going to not even describe really what makes them unique, but Music, just play with Udio and Suno. You'll be going in 5 minutes and creating music that kind of is mind-blowing for both the speed and the realism. Video creation, there's dozens of tools, but I would say Cling AI, Runway, two excellent ones to quickly pick up and be experimenting. With images, Flux and Midjourney. And with editing, I would say Descript and Opus are two good ones. Opus taking long, you know, clips and being able to quickly edit them down into something manageable. So as far as like my general advice, same thing with Hour One, Hey Jen, like if you're interested in this digital twin space and that's why you're here, both of these can be used at the kind of beginner level for under $100. You take a little video of yourself, you record your voice, you will have a digital twin without all the training on everything you've ever said and done, but you'll have a version of yourself. Again, because we don't have time, there's separate sessions on tools. I just want to give you some handy links. Curious Refuge has a section of essentially their recommended AI tools. It has for every tool and every category the ease of use and their demos of it and their examples. There's separate sessions about how to use these tools. But if you're thinking of leaning in, I think this is my favorite resource for sort of the pros and cons of each tool. And I actually created this site. It is realcreative.ai. And if you go to realcreative.ai, You will have creators and lists of their projects, projects— we put up about 5 a week— and thought leaders and people you can learn from. Ultimately, I know this advice is a little cliché, but the main thing you want to do if you're interested in this space, because I think that inevitably everybody should be experimenting, is find a few resources you trust to figure out when new tools are coming out. And then after you start learning what those tools are, really pushing yourself to try a few a week. And it doesn't have to be like grand filmmaking projects. I think a lot of you, if you just sort of pick up some of the tools from before and play with them, you'll find the ones you like. All right, we're going to close here. I'm going to leave it up to Rachel if you want to go 4 or 5 minutes on a live demo, or we kind of are out of time. 3 minutes. OK, this is so not going to work. All right, let me do my wrap-up now so when that doesn't work, it's fine. Guys, it's been great to be here. I have had a great time meeting you all. I am excited about this space. I am apprehensive about it at the same time. I think as the creator community, it is on all of us to use these tools responsibly. And ultimately for me, I do feel like we do not need to have this big Hollywood system to let people who have creative ideas see those ideas come to fruition. I am very excited about the possibilities of that. And when it comes to digital twins, I think we will move into an era where we all can benefit from the collective knowledge of so many smart, interesting people. With that, let's see if I can, if my guy, so the guy who built the brain of Read AI, his name is Parth, and he was here for a little bit of, you know what, if we got 3 minutes, and you're coming over here and not looking super confident, what do you think? Do I need to do something? Oh, here we are, okay. Let's join, let's see what happens. Let's see what happens. Okay. We're coming in here. You're seeing a little bit of BTS on my horrendous laptop. Disorganized screenshots. Connecting. Rachel, if you could invite me back for Dubai 2026. Okay. Let's see here. Is Parth there? Okay. Right? I mean, we're going haywire here. This isn't a good angle of me. I don't even know how— this isn't how we want to end this talk. Guys, it's been really good. It's been really good being here. We're going to forget that 90-second demo. We're going to thank Parth for staying up till 11 o'clock in LA to not be even a small part of this thing. And I just want to say that I'll be here for the next 3 days. Come say hi to me. Let's talk life. Let's talk AI. And thank you so much for listening. All right. Thanks, guys. ¡Hola!
More from Ben

H.O.W. #16 with Ben Relles, Head of Innovation for YouTube Originals, and The Showman of the WG'04




