- 5 months ago
Forbes Assistant Managing Editor Katharine Schwab speaks with Decart CEO, Dean Leitersdorf, about how the startup's real-time AI is building immersive digital worlds for fun and creativity. Discover the vision behind their trillion-dollar ambition and how they're set to redefine our interaction with AI beyond productivity, moving past "doom scrolling" into new forms of leisure and entertainment.
Read the full story on Forbes: https://www.forbes.com/sites/amyfeldman/2025/08/12/next-billion-dollar-startups-2025/?ctpv=searchpage
0:00 - Introduction: The Trillion-Dollar Question for AI
1:17 - What is Decart? An Antidote to Doom Scrolling
3:00 - How It Works: Turning Your World into Minecraft
4.26 - The Breakthrough: Real-Time Interactive AI Video
6:28 - Beyond Filters: New Use Cases in VR and Gaming
9:02 - Why AI Video is the Next Frontier for the Internet
12:22 - The Tech That Makes Real-Time AI Video Possible
14:15 - Our 'Kilocorn' Ambition: Building a Trillion-Dollar Company
16:32 - The Business Model: How We Plan to Get to $1 Trillion
19:11 - How We Became a Unicorn Before the Ink Was Dry
21:38 - The 3 Companies Inside Decart
24:08 - Meet the CEO: From a '14/100' on His First Exam to a PhD at 23
28:55 - Addressing the 'Scary' Side of Real-Time Video AI
31:02 - My 10-Year Vision: Making the Internet Fun Again
Subscribe to FORBES: https://www.youtube.com/user/Forbes?sub_confirmation=1
Fuel your success with Forbes. Gain unlimited access to premium journalism, including breaking news, groundbreaking in-depth reported stories, daily digests and more. Plus, members get a front-row seat at members-only events with leading thinkers and doers, access to premium video that can help you get ahead, an ad-light experience, early access to select products including NFT drops and more:
https://account.forbes.com/membership/?utm_source=youtube&utm_medium=display&utm_campaign=growth_non-sub_paid_subscribe_ytdescript
Stay Connected
Forbes newsletters: https://newsletters.editorial.forbes.com
Forbes on Facebook: http://fb.com/forbes
Forbes Video on Twitter: http://www.twitter.com/forbes
Forbes Video on Instagram: http://instagram.com/forbes
More From Forbes: http://forbes.com
Forbes covers the intersection of entrepreneurship, wealth, technology, business and lifestyle with a focus on people and success.
Read the full story on Forbes: https://www.forbes.com/sites/amyfeldman/2025/08/12/next-billion-dollar-startups-2025/?ctpv=searchpage
0:00 - Introduction: The Trillion-Dollar Question for AI
1:17 - What is Decart? An Antidote to Doom Scrolling
3:00 - How It Works: Turning Your World into Minecraft
4.26 - The Breakthrough: Real-Time Interactive AI Video
6:28 - Beyond Filters: New Use Cases in VR and Gaming
9:02 - Why AI Video is the Next Frontier for the Internet
12:22 - The Tech That Makes Real-Time AI Video Possible
14:15 - Our 'Kilocorn' Ambition: Building a Trillion-Dollar Company
16:32 - The Business Model: How We Plan to Get to $1 Trillion
19:11 - How We Became a Unicorn Before the Ink Was Dry
21:38 - The 3 Companies Inside Decart
24:08 - Meet the CEO: From a '14/100' on His First Exam to a PhD at 23
28:55 - Addressing the 'Scary' Side of Real-Time Video AI
31:02 - My 10-Year Vision: Making the Internet Fun Again
Subscribe to FORBES: https://www.youtube.com/user/Forbes?sub_confirmation=1
Fuel your success with Forbes. Gain unlimited access to premium journalism, including breaking news, groundbreaking in-depth reported stories, daily digests and more. Plus, members get a front-row seat at members-only events with leading thinkers and doers, access to premium video that can help you get ahead, an ad-light experience, early access to select products including NFT drops and more:
https://account.forbes.com/membership/?utm_source=youtube&utm_medium=display&utm_campaign=growth_non-sub_paid_subscribe_ytdescript
Stay Connected
Forbes newsletters: https://newsletters.editorial.forbes.com
Forbes on Facebook: http://fb.com/forbes
Forbes Video on Twitter: http://www.twitter.com/forbes
Forbes Video on Instagram: http://instagram.com/forbes
More From Forbes: http://forbes.com
Forbes covers the intersection of entrepreneurship, wealth, technology, business and lifestyle with a focus on people and success.
Category
🤖
TechTranscript
00:00I think that at this point, if you ask a random child somewhere around the States or somewhere in Europe or somewhere around the world, what's ChatGPT?
00:09They'll tell you, well, it's the AI I use to do my homework.
00:11But if you ask them, okay, well, what AI do you use when you're not doing your homework?
00:15The answer is, well, I don't know.
00:17What is an AI when I'm not doing my homework?
00:20That's a trillion dollar question that is going to be answered over the next year.
00:23Someone's going to answer that.
00:25And if that answer is the cart, then the cart is a trillion dollar company.
00:30Hi, I'm Catherine Schwab, an assistant managing editor here at Forbes covering technology.
00:35And I'm really excited to be here today with Dean Liedersdorf.
00:39He is the co-founder and CEO of Descartes, which is building AI models that generate digital worlds.
00:46Dean, welcome.
00:48Thanks for joining me.
00:49Hey, Catherine.
00:50Thanks for having me over.
00:51It's going to be, hopefully, it'll be a very fun ride.
00:53Yes, I can already tell.
00:56So let's just start with high level.
00:58What does it mean to be building an AI model that can generate digital worlds?
01:04What are you building at Descartes?
01:07AI has completely changed our lives over the past three years.
01:10It started with chatbots just completely taking over what it means to know stuff or to ask questions.
01:19And, you know, over the past year, we've all seen AI video just completely take over all our feeds.
01:24Our TikTok, Instagram feed is just covered in AI-generated cats.
01:29Now, what we're really trying to build is what's a completely new experience that allows people to be creative using AI video and AI audio that just lets us feel something different than continuously scrolling on the platforms that we know.
01:44Just like chatbots completely changed what it means to find information online or to create information, and they didn't just make better search.
01:53They made a completely new experience.
01:55What is that new thing that will make us feel different when we're not doing our homework, when we're not doing our, you know, work and all the commitments we have?
02:04But just something that lets us be creative and have fun and is not endless doom scrolling.
02:12Okay, so that's very high level and conceptual.
02:15So what, in practical, concrete terms, what are you building right now?
02:20So I'll give you an example.
02:21We just launched Mirage a few weeks ago.
02:23So what Mirage lets you do is you can open, I can, as you can, you know, check out for yourself on MirageDecart.ai, you can just take any video stream, whether it's your camera, whether it's a screen share, and just pass it to any different world.
02:36So I can have this Zoom call.
02:39Instead of it being a Zoom call, we can have it where I'm in Frozen and you're in Star Wars, and I can completely see what's in my imagination, just show it directly to you.
02:48So the first model that we released in this series was a demo a long time ago, roughly back in November, which allowed us to simulate just an entire virtual world.
02:59And you could kind of play a game that was completely generated by AI.
03:03Now, what we released a few weeks ago really just lets you take your real-life environment and just cast it into any dimension that you can imagine.
03:10It's super, super early.
03:13We have a bunch of launches coming up in the next few weeks building on top of this model.
03:17That really just let you take what you see in front of your eyes and just change it with your imagination.
03:23Okay, so can you give me an example of how your users have been playing around with this and how they're using it?
03:31So something we've been seeing that's really mind-blowing.
03:35Lots of people like to take Mirage and just apply it on the real world.
03:38So we'd see people just taking broomsticks and hitting each other with them and then applying with Mirage a Star Wars filter.
03:45And everything would turn into an insanely cool lightsaber fight.
03:49Or we would see people, we saw a bunch of people just, you know, like running around their house, usually kids.
03:55And they'd turn the entire house into Minecraft using like a Minecraft blocky filter.
04:01And they would just run around the house, start punching like their furniture and seeing if it would break.
04:06It just lets people completely jump into this new immersive kind of experience they just couldn't describe before.
04:14So just, does that mean like if you're on Zoom with someone, you can kind of apply this filter onto your live video, right?
04:25That's what we're talking about.
04:25Or like on FaceTime or something like that?
04:27Exactly. So what we're allowing for, as you were saying, as an example, you know, with Mirage, with the previous version of Mirage,
04:34you can take, for example, live video screen, whether it's Zoom, whether it's a FaceTime call, and just apply any filter you can imagine.
04:42Now, this is enabled by a new kind of tech that we released just a few weeks ago that lets you basically do real-time video models.
04:49So if you imagine all the video, AI video that you see on the internet, it usually takes a lot of time to generate those videos.
04:57With our tech, you can actually generate this in real time.
05:00So you can just say, hey, I want this to change into Frozen, or I want this to change into Spider-Man, and it just happens instantly.
05:08We're also going to be launching very soon with this.
05:11It's also just completely generating videos from scratch.
05:13So you can describe anything that your imagination would just appear here.
05:17I could, for example, in this interview, say, hey, see my hand?
05:21It's empty now.
05:22Give me a sword.
05:23And then it's just a sword would pop up here.
05:25And I could say, okay, and you know, we should, can you add a monkey that sits right here?
05:29Like, no, no, a bit lower.
05:31Okay, perfect.
05:33And now, like, monkey, high-five me.
05:34And then, like, the little monkey here would high-five me.
05:37So it's a key technological breakthrough that lets you just interact with AI-generated video as it's generating.
05:47And not say, hey, here's a text prompt and wait a few minutes for the AI to spit out a video.
05:52Okay, got it.
05:53So, you know, it reminds me a lot of when Snapchat released Filters.
06:00Like, this is a long time ago at this point.
06:04And it was, like, a huge thing.
06:07And people loved playing with them, like, turning themselves into cats or adding ears or whatever.
06:13You name it.
06:14So it sounds like the kind of real-time equivalent of that in some ways.
06:19Is that, am I getting this right?
06:21So I definitely think that there's a filter-like aspect here that that's, you know, that's a thing we know as people because we've seen this and existed.
06:29And, you know, so far we could use static filters here.
06:32You can create any filter that you can think of just with your imagination.
06:35Just talk to it and it just happens.
06:36But what's very interesting to us is that it seems people are finding completely new use cases for this tech that didn't exist before.
06:46Something that is very, very interesting is VR.
06:50You've seen people put it inside VR headsets and just walk around cities and change.
06:54You know, you can just walk around New York City and change it to a Godzilla fight.
06:59And then the entire city would change as you walk inside it.
07:02So while we do get some familiar concept that we saw with filters and that lets consumers already instinctively know what to do,
07:11we also get a completely new dimension here of what it means to interact with AI video and what we can do with this.
07:19Okay.
07:20So help me, given that this is kind of a totally new way of using AI, you know, those of us, I mean, most of us use some form of chatbot at this point.
07:29Um, how do you see people using this as part of their daily life or, or what are, what are some more of those kind of use cases?
07:39It seems like mostly in like entertainment and gaming.
07:41Is that, is that right?
07:44What we can do with this, chat GPT was the first interface that people and computers had that was real-time text.
07:51We could say what we think and we'd have a computer speak back to us in text.
07:56For the first time, you can actually interact with a computer real-time in video.
08:02So whether it's a chatbot that you can actually see.
08:05So chat GPT, you know, you can, you can text it.
08:07You can talk to it over a phone.
08:09You can't actually see it.
08:10So it's whether seeing a chatbot and actually, and, and, and visually interacting with it, having a FaceTime call with chat GPT, for example.
08:17And all the way to letting you jump into creative worlds that you imagined.
08:25So whether it's taking your living room and changing it into Antarctica or letting you imagine a completely new game and just play it on the spot.
08:32So what we, what we're, what we're looking at here is that if you look at the internet pre-chat GPT at the old internet, it used to be basically four things.
08:41It was knowledge and everything that was related to search and maps and docs and all that.
08:47It was creativity, Instagram, TikTok, YouTube, gaming, Netflix, and all those sorts.
08:53And then we had communication and then shopping.
08:55That's basically the internet.
08:56The first category was completely taken over by chatbots.
09:01Chatbots has completely decimated how we get information, how we use information, how we share information that we, that we create.
09:09Nothing really has happened to the second category.
09:11Creativity has stayed exactly the same over the past three years.
09:15We just got cat videos added to our TikTok feeds.
09:19With, by having the ability to just generate any video on the fly.
09:24We can simulate all these experiences that we had and potentially bring even new ones to users.
09:33So sure, there's the obvious use case of just infinitely generate your TikTok stream.
09:37I can just continuously show you a stream that just changes based on, on how you like it.
09:42If you're smiling or if you're crying, we'll know if you like the video and just continuously generate the next frames.
09:47But it's very interesting to see what people will create with this new kind of technology.
09:52Okay.
09:54Okay, great.
09:55So you mentioned that there were some technological breakthroughs here to enable this.
10:01How is training this kind of model different from GPT-5 or a text-based model?
10:07Like, walk me through kind of the training process and what, from the tech perspective, what breakthrough was required to get here?
10:15There are generally two big categories of AI models.
10:20One is all the text-based models.
10:22And those are basically just trained on getting all the text on the internet, just dumping it on a bunch of GPUs and getting an LLM that knows how to create very, very good text out of that.
10:31That's where, you know, we see GPT-5, Claude, XAI's models, GROC, and so forth.
10:36But the other kind of models that we're seeing are what's called diffusion-based models, and they're usually used for generating video or audio.
10:43So you take a bunch of videos or you take a bunch of audio files, and using them, you synthesize, you create a model that knows how to synthesize new content based on a prompt.
10:52So I could tell it, for example, create a video of mammoths in Antarctica, and you'd wait a bit, and then you'd get a video like that.
11:00What used to be before Mirage, Mirage was the first real-time video model in the world.
11:07What that basically means is instead of having to tell an AI, hey, here's a prompt and generate a video based on that,
11:15and then I'd have to wait usually 10, 20, maybe even a minute, seconds, or maybe even a minute till we get the response back,
11:22with Mirage, you can just tell it's something, and roughly 40, 50 milliseconds later, it starts creating it.
11:29That actually lets you be able to continuously interact with AI to really shape the generation to be what you want.
11:37In order to get there, the main two breakthroughs we had to achieve were, one, completely reducing the latency,
11:44so it doesn't take tens of seconds to generate the video, but roughly 0.04 seconds to generate it.
11:49And two, also, instead of just generating short clips, so the AI model so far could generate, say, 10 seconds, 20 seconds, maybe 30 seconds,
11:58Mirage can generate an infinite clip, so I can generate an hour-long clip, I can generate 10-hour-long clips.
12:03And those were the two key breakthroughs that we had to do in order to be able to get this.
12:07Interesting. So I'm curious, what drew you to wanting to build models when you're competing against all of the big tech companies?
12:19Of course, they're all moving into different types of models, video, audio, all, like, you name it.
12:25And I think GPT-5 can now generate apps.
12:29You know, what is your thinking around this really significant competition that could wipe you out as a small startup?
12:42I love this question. You know, when we started the cart, it was roughly two years ago.
12:48And, you know, we're ideating for a long time. It took us, like, three, four months until we found what we wanted to do.
12:54And we looked at everything. We looked at everything from cybersecurity to starting a solar panel company, AI for healthcare,
13:01which really went through all the different ideas.
13:04And just nothing stuck. Nothing was interesting enough.
13:11It took us, you know, my co-founder, Moshe, he was fed up with me at a certain point.
13:16He's like, dude, everything we're thinking of, you're excited about it for two days, and just the weekend, you're like, okay, this is not fun anymore.
13:23What do we do? What do you actually want to do?
13:26And so, it took us a few months to admit it, but we finally realized, once in your life, you got to take, you know, five years, maybe 10 years, and try to build a kilocorn.
13:39You know what a kilocorn is, right? You have a unicorn that's a billion-dollar company.
13:43Then you have a decacorn, a centicorn, a kilocorn.
13:45Okay, trillion-dollar company.
13:47Once in your life, try to build a Google. Try to build an Apple. Try to build a Facebook.
13:52Give it five years. If it works, that's amazing.
13:54If it doesn't, you have the best ride possible.
13:57Now, to do that, the field you have to compete in is in building completely new consumer experiences based on very, very deep technological breakthroughs.
14:10It's incredibly hard, as you were saying.
14:12You know, there's so many great competitors out there, and this field is just so, so, so hot and challenging right now.
14:21But you certainly have to give it a shot.
14:24Now, if it works, it's incredible.
14:27If it doesn't, you really had the best journey you can possibly imagine.
14:32Okay, so why do you think this idea is a trillion-dollar idea?
14:39Like, where would your revenue come from?
14:43Like, what is the business plan to get to?
14:47You are now a billion-dollar company.
14:49How do you get to become a trillion-dollar company?
14:52So, I think that with AI, the rules are slightly bending.
14:59Companies used to have to be just consumer companies or selling to enterprises.
15:05With AI, you can kind of do both.
15:06It's like when, you know, ChatGPT, they're both selling it through ChatGPT itself, but also through the API.
15:13We're seeing something similar with what we're building.
15:15So, we're building our real-time video models, and we're providing them both to enterprise customers.
15:22We're paying a lot of money for access to the tech that we have.
15:27And frankly, we've barely burnt any of the money that we raised, I think.
15:32But at this point, we've raised over $150 million, and we've burnt less than $10 million of investor funds, simply because we're generating a lot of revenue from licensing our tech.
15:46So, that's one stream.
15:48And, you know, when we launched Mirage, we had inbound from everyone.
15:51Even hotels came up to us.
15:53Hotel chains around the world.
15:54We're like, hey, can we, you know, can we stick up Mirage on, like, our lobbies so that when people walk by this, they see themselves in a completely imaginary world?
16:04All the way to hospitals that were like, hey, can we, you know, when we do, like, surgeries on kids and we have to distract them, can we show this live stream to them?
16:13Because it's so captivating and draws their attention.
16:18So, that's on the one hand.
16:19At the same time, what we're targeting is how do we just create a completely new experience that makes us, that makes every person on the planet just feel better and feel something fun when they're out to be creative, when they're out to have their leisure time?
16:40I think that at this point, if you ask a random child somewhere around the States or somewhere in Europe or somewhere around the world, what AI do they use?
16:51What's ChatGPT?
16:52They'll tell you, well, it's the AI I use to do my homework.
16:55But if you ask them, okay, well, what AI do you use when you're not doing your homework?
16:58The answer is, well, I don't know.
17:01What is an AI when I'm not doing my homework?
17:03And that's a question, that's a trillion-dollar question that is going to be answered over the next year.
17:08Someone's going to answer that.
17:10And if that answer is the cart, then the cart's a trillion-dollar company.
17:13It was a little bittersweet for us because we had just put you on our next billion-dollar startups list.
17:21We closed the magazine.
17:22It went to press.
17:24And then before we had the chance to publish it online, you had already raised an over-billion-dollar valuation, which makes us both look smart and also like we just missed it.
17:36So, congrats on your funding round.
17:40If you're not burning any of that investor money, how are you planning to use this new funding round to kind of get to the next stage of the business?
17:48First of all, super fun to have the previous interview that we had roughly a month or two ago.
17:55But before all this happened, it seems like you guys are fortune tellers.
17:59So maybe on your next $10 billion company list, we should also interview for that as well.
18:04But frankly speaking, we have many new models coming out soon.
18:12We spent the past seven months in very deep research to bring Mirage and to make it work.
18:19And so far, it's the only model in the world that can generate video in real time that's actually available to people.
18:25Also, the other big tech don't have this yet.
18:28We're at least several months ahead of them, except for Google.
18:31Google seems to have this as well.
18:34But we have lots of new releases for the model that are upcoming, which will enable drastic new capabilities that we've seen internally over the past few weeks.
18:44And we're just mind blown and we can't wait to share them with the world.
18:46And at the same time, we're going to have many new products released on top of this.
18:51So we're releasing a VR product.
18:52We're releasing a FaceTime-like product.
18:54We're releasing a Snap-like product.
18:57We're releasing products that kind of resemble stuff that we just have with a completely new AI twist on top of them.
19:03And we're going to iterate a lot until we see what people actually like doing with us.
19:08What we're actually going to do with the money is
19:10When we have these launches, if they do take off and they can take off instantly, we'll need to be able to serve many, many millions of people at the same time.
19:24Obviously, GPUs are very expensive, which is where all this fits in.
19:28So you're planning to launch consumer apps, like you just mentioned, a FaceTime-like app.
19:34So we're talking about an app that lets you call your friends and family where you can have this kind of like AR live video experience in that app.
19:44That's coming?
19:45A hundred percent.
19:46So we have that coming.
19:47We have a few more cool stuff coming that we'll be sharing in the next few weeks.
19:50And at the same time, we're providing our tech to enterprises.
19:54There's lots of expansion there in the gaming and the real estate and entertainment sectors.
19:59And so Descartes is kind of operating like three companies.
20:02You have one deep tech lab that just produces new foundational generational models every few months.
20:09And now it's going to be a few weeks.
20:11And at the same time, the second part of the company is releasing new consumer experiences with that.
20:16And the third part of the company is actually giving that tech, giving access to that tech to enterprises to enable them to create new experiences for their consumers.
20:26Gotcha.
20:26Okay.
20:26So we reported as part of the list that you brought in $20 million in revenue in 2024.
20:33I'm curious if you can be more specific about your customers.
20:39You've mentioned gaming, real estate.
20:41How are Descartes models being used right now by your customers?
20:46So in the previous year, we licensed lots of our tech that enabled running models efficiently.
20:53So with our tech, we have basic infrastructure that allows us to run models, whether it's our models or other people's models, just faster or more cheaper.
21:05And during 2024, up until now, we've been providing access to lots of the tech to enable other customers to be able to run their own models much more efficiently.
21:18Starting next week, we're also going to be providing public access to using the models that we have that we're creating ourselves via an API.
21:30And so that's going to create a completely new revenue stream for us in this upcoming year, which we're very excited by.
21:37So, so far, it's been providing other companies that have their own models with the tools to be able to just execute them much more efficiently on GPUs.
21:46And this is going to be complemented by an additional revenue stream that we're starting now, which is actually giving access to Descartes models that we build in-house.
21:53So I do want to rewind a little bit.
21:55Will you tell me a bit about your background and kind of how, what brought you to starting this company a few years ago?
22:04Yeah.
22:07I'm not saying it, but I was born in Israel, Switzerland, then Palo Alto.
22:14When I got to Palo Alto, I was 15.
22:16That's, that's when I realized my, you know, everyone lied to me my entire life because, because they told me I was smart.
22:22And then I get to Palo Alto and I see a 14-year-old kid doing his undergrad at Stanford and a 17-year-old one in his third bootstrap started.
22:31This is like actual real people.
22:32And I was like, what did I do my entire life?
22:35Okay.
22:36Like I was sitting around playing RuneScape and Minecraft and getting hundreds on tests that didn't matter.
22:41And suddenly I get to this completely new world in Palo Alto where you realize you're in the smart people Olympics.
22:48You realize that so far, you know, you've been playing in your minor leagues in whatever city you grew up in and you thought you were smart.
22:56And then you get there and suddenly you see you're not the only one.
23:02And for me, that was when I got to Palo Alto, first semester of sophomore year in high school, it was a complete disaster.
23:10I got a 14 on my first exam.
23:13Okay.
23:13Like there was an exam that the week I arrived in computer science and I got a 14 out of a hundred.
23:18Now, my teacher was so kind that he wrote that I was sick on that day so that I won't appear on, you know, my report card.
23:27Second semester in sophomore year, I got my act together.
23:30I was like, okay, I need to start catching up.
23:31So I went to my guidance counselor and I told her, well, you know, I just have to skip a grade.
23:35I'm going to do junior and senior year of high school at the same time.
23:38And so that happened.
23:39Then I went to college.
23:40I said, well, you know, I flew back to Israel to my home country.
23:45I said, well, I'm going to take the four-year undergrad and do it in a year.
23:48That was the goal.
23:50The TLDRs ended up doing my undergrad and master's PhD all together in five and a half years.
23:54It was the fastest ever in the Institute's history.
23:57I finished my PhD when I was 23.
23:59So my PhD was distributed computing, which is the foundations for how we take huge technological problems and distribute them over lots of different GPUs.
24:12That used to be very, very niche.
24:14And today it's under stage.
24:16It's the foundations which actually enable AI to work.
24:19I won the ACM award for the best PhD in distributed computing in the year 2022.
24:24Now, when we started off the cart, there were a few random kids that someone one day told them they were smart.
24:37Okay.
24:38I was sitting in my co-founder's living room and really thinking, you know, what do we build?
24:43What do we do with, okay, we're apparently smart people.
24:47Someone said that.
24:48And we can apparently hire other smart people to do amazing stuff with us.
24:53What do we actually build?
24:55And it was just a few months after ChatGPT had launched.
25:01We all knew that at one point a product like ChatGPT will come, that AI will, you know, actually just completely change how we interact with each other, how we interact with the internet.
25:15We just didn't know it would come this early.
25:17And it took us, it took us a few months to realize that ChatGPT is actually a thing and that we need to completely change from whatever ideas we previously had and make sure we're focused on how do we build the next AI giant.
25:31And the rest is history.
25:32So I have to tell you that I find the idea of real-time live video and models that can manipulate it both, it sounds very fun in the ways that you've described it.
25:45It also sounds like a little bit scary if you could manipulate what people are saying or doing in a realistic way in real time.
25:55And I'm curious how you have thought about that problem as you've developed this model.
26:02Oh, I think, I think that's a very important problem.
26:07And it starts with setting expectations.
26:11So if you're creating a platform where people expect all the content to be authentic or to believe that it came from the real world, then it is definitely a huge problem.
26:24If you have content that's generated there and you can't tell it apart.
26:28And most of the platforms have been trying to deal with it by labeling content as it has generated.
26:32If you start with the expectation that you know lots of the content you're seeing is someone's imagination and not necessarily fact, then that definitely completely changes how you as the user perceive what you're seeing.
26:49At the same time, this is a very, very tough problem that I think as an industry we have to address and we have to understand how we adapt to a world where lots of the content that we're seeing is imaginatory and not real.
27:04So it's, it's, it's a common, it'll probably be a combination of how we as an industry tackle this at the same time, how consumers respond to the content that they're seeing.
27:17Okay.
27:18Okay.
27:19So can your models right now kind of manipulate what someone is saying in real time?
27:24Oh, we're, we're, we're not touching, we're not touching audio at all.
27:28Okay.
27:29Yeah.
27:30Audio manipulation is, is something that's very dangerous needs to be dealt with, with care.
27:35Yeah.
27:36Okay.
27:37Um, so Moonshot, 10 years from now, what is your great hope for Descartes?
27:49Being on our phones used to be very fun.
27:52And we, we used to have, you know, a lot of fun with, with entertainment on individual, in the, in the virtual world.
28:00Over the past few years, as a consumer, it feels like doom scrolling has taken over lots of what we do on our phones.
28:08And it's just not that fun.
28:09There's so many different apps out there that you have an instinct to go to and spend an hour with them.
28:14And you're not actually enjoying your time.
28:17I think that we have a real opportunity here to, to create a completely different experience with AI.
28:23That would actually allow us to relax and to, to have a good time and to just channel our inner creativity while we're using it.
28:35Instead of seeing continuous, continuous streams that don't really mean that much to us.
28:41If we manage to touch hundreds of millions, a billion people with the tech that we create, whether it's directly or whether it's indirectly by selling it to partners.
28:50Then that would, that would have been what we set out to do.
28:55And that would, that would make me very, very happy to see the world having a bit more fun in this, in this entire new, in this, in this very stressful era that we're living in.
29:06Dean, thank you so much for your time.
29:08I really appreciate all your thoughts on, on the future of entertainment with AI.
29:12It's great.
29:14Thanks so much.
29:20Thanks so much.
Comments