• 14 hours ago
Growing up in Silicon Valley, Demi Guo had two loves: technology and poetry. “Because of these two passions, I was always very excited to do something at the intersection of AI and creative stuff,” she says. Her interests came together in 2023 when she cofounded Pika with fellow Stanford Ph.D. student Chenlin Meng, 27. The AI tool turns written prompts into cinematic videos in a plethora of styles—including realistic live action, computer animation and anime. The startup has raised $135 million from investors, including former GitHub CEO Nat Friedman, at a reported valuation of nearly $470 million. This sort of tech has become a hot-button topic amid concerns that AI will threaten showbiz jobs (a central issue in the Hollywood writers’ strike in 2023). Guo argues Pika isn’t out to replace human artists: “The human artist is the one to guide AI in the right direction and eventually become a masterpiece,” she says.

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.
Transcript
00:00We all want to build a very powerful AI video model
00:04that can really help creators realize their creative ideas
00:07in a more efficient way
00:09and to make video creation more accessible.
00:15Hello, Demi, thank you so much for joining us today.
00:18You are the founder and CEO of Pica.
00:20I'm super excited to talk about all your building
00:22in the AI space.
00:23Before we get into the nitty gritty of AI right now,
00:26I'd love to know a little bit about you.
00:27Tell me where you grew up,
00:29where you're based now,
00:30and just a little bit about yourself.
00:32Yeah, thanks so much for having me.
00:33So I was born actually in Silicon Valley.
00:39And then I spent some years in Asia
00:43and then went back for college.
00:46So I did college at Harvard.
00:47And before doing Pica, I was doing AI PhD at Stanford.
00:52Okay, and how did that kind of experience
00:55being born in Silicon Valley impact your interest
00:59in the tech space, in the AI space?
01:01Did that have a big effect
01:03on what you were interested in as a kid?
01:05Yeah, for sure.
01:06So I was always going to Stanford summer camps
01:10since middle school or high school.
01:13I actually chose Harvard
01:15because I know I'm going to come back to Silicon Valley
01:18to build a company and probably do an AI PhD at Stanford.
01:25So definitely growing up in Silicon Valley
01:27and also surrounded by a lot of friends
01:29who were also doing startup,
01:30who were also in Silicon Valley,
01:33definitely inspired me more about building a startup
01:39and also helped me more to understand the space.
01:43Yeah, when you were a kid and you were seeing
01:46these peers or family friends
01:47or just people in your neighborhood
01:48and your community building startups
01:50and building in the tech space,
01:52did you always want to be a founder?
01:53Did you always want to go into the tech industry?
01:56How did that kind of impact
01:58which route you wanted to go down?
02:00I think I definitely got a lot of inspiration
02:05from surrounded by a lot of other founders
02:08and being in Silicon Valley.
02:10I think it's really cool
02:12that everyone is working on something
02:15they're really passionate about
02:16and is really trying to work towards this idea
02:21of using technology to make something
02:24that's very impactful,
02:26that can impact everyone's life
02:27and that can really push the frontiers of technology
02:32and really change the world.
02:33How did you find that you're passionate about AI
02:37and the media space?
02:39Yeah, so I was always really interested
02:43in arts and creative stuff when I was young.
02:47I grew up in a very creative family,
02:50I would say artistic family.
02:52I have two sisters, one sister studies film
02:56and the other sister studies fashion design.
02:59My family are also really into art,
03:02so we have a cute little art gallery.
03:05So I always grew up wanting to be a writer
03:08and really interested in poetry.
03:13That piece of me was basically with me when I was young,
03:17but I also started programming since elementary school.
03:19So I've been doing AI, I've been doing computer science
03:23and AI for seven or eight years now.
03:27Because of these two passions,
03:28I was always very excited to do something
03:30that's intersection between AI and creative stuff.
03:35So using AI, I'm also never a professional artist,
03:39I'm not professionally trained.
03:40So I always wanted to use technology,
03:44use AI to really help me to realize
03:48some of my creative ideas.
03:50Yeah, and it's such a fusion
03:51of the two kind of passions of yours.
03:53So today Pika is an AI platform
03:56that allows people to create videos, correct?
03:58Out of basically either a prompt or an image that they have.
04:01Can you walk me through exactly how it works,
04:03what people are using Pika for today?
04:06So for our company, we train an AI foundation model,
04:09meaning like a AI model that can do anything around video.
04:13They can understand video, can generate video
04:16and can edit videos.
04:18To be more precisely, the way people interact
04:20with our model is there could be a lot of different ways.
04:25But the most basic way is user can enter a description,
04:30which we call a text prompt
04:32to describe what video you want to generate.
04:34So for example, you can say a dinosaur dancing in a shower.
04:38So the video, then it will output,
04:40the AI model will output literally a dinosaur
04:44dancing in a shower.
04:46You can also use image inputs to,
04:49and then to ask the model to animate an image.
04:52So for example, you can upload your own painting
04:56and then tell the AI model
04:58the way you want to animate your own painting.
05:00To be more broadly, you can also edit your own videos.
05:04So for example, you might upload a selfie videos
05:07and you can add an explosion effects
05:12inside your selfie videos.
05:14So how does this technology work
05:16for someone who like does not understand the AI space
05:18and they don't understand that when they input something,
05:20they get an output.
05:21How would you describe it to people
05:23who are just like not in your industry
05:25and are looking for a little bit of insight
05:27into how this is like magically turning
05:30what I'm typing in into a video?
05:33The way it works is we train like a large AI model.
05:37So basically the model learns from a lot of data,
05:42which is a lot of videos, for example.
05:45So the model will learn by seeing, just like us,
05:51by reading a lot of books, we have more knowledge.
05:54Totally.
05:54So the AI model is similar.
05:55So by seeing a lot of videos,
05:58the model start to understand what is a video.
06:03Like when you have a ball, you drop it,
06:06you will fall onto the ground.
06:07So the model starts to understand these physics
06:10and understand, predict what people react
06:15and like all these things by seeing a lot of data.
06:21And did you guys build that AI model from the ground up?
06:23Is it on top of a platform that already exists
06:26or how did you kind of come up with this model
06:28that's then facilitating all this?
06:30Yeah, so we're a very technical team.
06:32So we actually build our model from ground up.
06:35And what was that process like?
06:36What does the team look like?
06:37Who did you have to bring on to make this thing happen?
06:40Yeah, so both me and my co-founder,
06:43we're all AI PhD at Stanford.
06:46So I work on AI for content creation at Stanford.
06:49She worked on diffusion model,
06:51which is the class of models behind image and video models.
06:58So we also, we were basically doing AI for video space.
07:04And since we started company,
07:06we obviously brought a lot of more people to the team.
07:10So a lot of engineers, a lot of scientists.
07:12So including people who were like top researchers
07:17at the top academic labs, such as Stanford, MIT,
07:20and also top researchers from the top industry labs
07:25like Google and Meta and Microsoft.
07:30What was your pitch to these other experts
07:32in terms of like why they should come and work with you?
07:36I think that's a really good question.
07:38I think, I would say usually
07:41because we share the same passion.
07:43So we all want to build like a very powerful AI video model.
07:50This is very, at the same time,
07:52I was also building a video, AI video product
07:56that can really help creators to realize
08:00their creative ideas in a more efficient way
08:02and to make video creation more accessible.
08:05So we share the same dream.
08:06And I think it's very exciting to us
08:08because we're pushing the frontier of the technology
08:12by building the next best like AI video model,
08:17but also trying to redesign the way people make videos,
08:20which is all very innovative and very exciting.
08:23What do you see as the biggest use case
08:25for Pika right now?
08:26Is it for just an everyday user, an influencer,
08:29a content creator on social media, advertisers?
08:32What is the biggest kind of cohort of people
08:35who are using the platform?
08:36Yeah, so right now we see a lot of,
08:38mostly are people who are interested to do creative videos.
08:42So which could include like a film
08:45or it could also include like a commercial.
08:48You could also include,
08:49like this could be like more professional filmmaker
08:53who are exploring using AI to make a low budget film
08:55or make a social media videos.
08:58It could also be like someone who is like,
09:02who are from like completely non-tech
09:05and non-AI background,
09:06such as like accountant or like IT worker
09:10who are interested to make a short,
09:13create a piece as a side project or for their children.
09:16Do you have any examples of projects
09:19that have been created on the platform
09:20that you have just been blown away by?
09:22Yeah, definitely.
09:24Like there are a couple of examples.
09:25I wish I can share some visuals,
09:28but we do see people who are like,
09:29for example, reproducing a commercial
09:32that they made a long time ago.
09:33And now they're using AI, they're able to,
09:36I think the commercial was like 20 people
09:38for like two months with like,
09:41I don't know, more than 100K budget or something.
09:43And then using AI, it can be like,
09:45one or two people team for like one, two days
09:48with like below a thousand dollars.
09:51So, which is very exciting.
09:53Yeah, we also began to explore some use cases
09:56that could even be more consumer
09:58and more accessible for people
09:59who have no creative background.
10:02Do you see that efficiency in terms of like
10:04the finances being cut significantly,
10:07the team being cut significantly,
10:08the time being cut significantly?
10:10Is that the biggest opportunity here
10:12or are there other things that people are looking
10:14to these platforms to improve on?
10:16I definitely feel like the way I think about it
10:18is there are like people who are professionals.
10:20So in that setting, it's more about like
10:21making it more efficient, right?
10:23And like if you're an artist,
10:26the hope is that the AI model can really serve you
10:30to assist you so that you can focus
10:33on the creative idea part, not the manual work part.
10:36So really make it the process from idea
10:39to execution in a much more efficient way.
10:41But we also are starting to explore and use cases
10:45that are a bit more, even more accessible
10:47for like more consumer users.
10:49So for them, like it might not be like making efficient,
10:53but more about like making them be able to create stuff.
10:57So helping them to be creative and to express themselves,
11:02express ideas you have.
11:04Maybe like, for example, like I think we all have ideas
11:07and maybe we all have dreams to be able to, you know,
11:11realize your wild ideas into some visuals.
11:15Totally, and bring that thing to life.
11:17What are your thoughts?
11:18Because I think a lot of people in the creative space
11:20in journalism, for one, but also in video,
11:23in photo, across, you know, many sectors of business,
11:26there's a big hesitancy towards AI
11:27because they don't want it to replace the people
11:29who are doing it currently and doing it in, you know,
11:31more of a manual way.
11:33What are your thoughts on that?
11:34What would you say to people who are maybe hesitant
11:35about the industry because of that?
11:37I want to tell a story for that question.
11:40Yeah, please do.
11:41So that's actually why I started Pika.
11:43It really inspired me to start Pika
11:45because I was trying to make a AI film myself
11:49before starting Pika.
11:50And at that time, we were trying to use AI.
11:54We do live action.
11:56We do live action shots.
11:57And we're trying to use AI to make it into animation.
12:01So that was like before Pika.
12:03That process really makes me realize two things.
12:06One is, you know, the filmmaking process
12:09is still very, very difficult.
12:10Like maybe like even green screening
12:12for a couple of second clips can take a lot of time.
12:15But more importantly, we were a group of,
12:19because it was an AI film project,
12:20we were a group of AI PhD students
12:22and computer science students.
12:24We don't have any visual creative director.
12:26We have a really good script writer.
12:29We don't have a visual director.
12:31And we really, really struggle on the visual.
12:34And in the film, even we have the most technical team.
12:38You know, the film doesn't look good.
12:40And that's something I like firsthand realized
12:43the importance of artists.
12:44And first I realized AI is not going to replace artists.
12:48Like the way I think about it is
12:51AI is really enabling all possibilities.
12:53But the human, the artist is a person to guide AI
12:57in the right direction and eventually become a masterpiece.
13:00And that's also become something we really value,
13:03we really care about in our company.
13:06We have a lot of artists.
13:07We have, you know, our first hire
13:09is actually a creative director.
13:11We have, you know, one third of people
13:13from a creative backgrounds.
13:14And we are really trying to build,
13:16we're not trying to build like automatic AI
13:20video generation platform.
13:21We're trying to build an AI video tool for creators
13:25that can serve creators and enable people
13:27to do more interesting stuff.
13:30Definitely.
13:30Where is the AI model that you have
13:32getting this information?
13:33You said it's learning from what exists today
13:35to kind of, you know, put that into its own work.
13:38Where is it learning that information from?
13:40We learn from a lot of various sources,
13:43like videos, data that we curated.
13:46Okay, so you guys have curated the information
13:48that then you're feeding into the model
13:49that it learns from?
13:50Yes.
13:51Okay, very cool.
13:52And then just in terms of like the future of AI,
13:55obviously video generation, photo generation,
13:57like there has been a lot of movement on it,
14:00but there's also a long way to go.
14:01You know, you see the pictures of, you know,
14:03AI generated portraits and it has like, you know,
14:05multiple extra hands or, you know,
14:07the eyes are, you know, crazy or whatever.
14:09So how do you, what do you see as the future
14:11of AI video and content generation?
14:14Where are you going?
14:15What are the next things that you're really preparing for?
14:18We definitely see like, you know, the technology
14:20and a lot of the AI video product
14:21is still in a more prototype phase
14:24than more like a fully commercialized phase.
14:27And, you know, this is, but at the same time,
14:31you know, this is because the technology is improving,
14:33right?
14:34We're still improving the technology
14:35and our company is also trying to push the frontier
14:37of technology so that maybe we can commercialize even more.
14:41And this is a process, but we see there's so much progress,
14:45even last year.
14:47So the acceleration speed is very, very high.
14:50And I cannot, like, you know, one year ago,
14:52I think maybe like our company just started one and a half,
14:55more than, less than one and a half year ago.
14:57And I think like two years ago,
14:59like AI video is just not a thing.
15:01You cannot actually generate any sensible videos.
15:04And now it's such a leap forward from that time.
15:08And it can, like, I'm very certain
15:10that in the next one year or two years,
15:12hopefully like we can already have, you know,
15:15more, even more commercialized, you know, AI videos,
15:21not the model, but also the product in reality.
15:26Yeah, definitely.
15:27The last question I have for you is you and the company
15:30have raised more than $140 million in investments
15:33and you have millions of users on the platform.
15:35And like you said, that's just in, you know,
15:37a couple of years timeframe.
15:38What's your best piece of advice for other founders
15:41in the AI space right now that are trying to embark
15:44on this rapid growth like you have seen?
15:46For me, I think what's important for me is to work
15:50on something that you're really passionate about.
15:51I think, cause I really care about, like myself is,
15:55you know, it could be our kind of the user.
15:58I use the user tool to create our marketing videos.
16:01I think it's really important you work on something
16:03that you're really passionate about.
16:04So you are very motivated to make it happen
16:07and you also really enjoy the process.
16:10Yeah, well, thank you so much for joining me today.
16:12This is super cool to hear about all your building.
16:13So thank you for taking the time.
16:14Thanks so much for having me.

Recommended