Generative artificial intelligence is sweeping the nation. People are turning themselves into animated characters, drafting their essays with ChatGPT, and illustrating with Stable Diffusion. Or, as was the case with the tiny special effects team on the movie Everything Everywhere All at Once, they’re using it to help edit a movie.
On the latest episode of Dead Cat, Cristóbal Valenzuela, the chief executive officer at generative artificial intelligence company Runway, talked about how he discovered that his AI-powered video editing software was used to help make the award winning film.
When I wrote about generative AI burning white hot back in October, I talked to Valenzuela for that story and called him “among the most compelling founders that I’ve come across while reporting on artificial intelligence.” So I thought it would be fun to have him on the podcast and discuss some of the most pressing issues facing generative artificial intelligence.
To help me interview Valenzuela, I invited Non-Technical podcast host and viral comedian Alexis Gay to guest host the episode. You’ll probably recognize her from some of her viral tech parody videos. (Listen to my guest appearance on Non-Technical if you want to learn more than you ever thought you wanted to know about the man behind the newsletter.)
On the podcast, Valenzuela predicts that “very soon,” in “a couple of years,” artificial intelligence software will be able to create the sort of TikTok videos that people flip through online. “We’re heading towards a world — where a lot of the content that you consume online will be generated [by artificial intelligence],” Valenzuela said.
“There’s definitely an exponential progress rate that you can see and perceive more clearly now,” Valenzuela said. “What took years of progress is now taking months. And what used to take months is now taking weeks.”
Give it a listen
Read the automated transcript
Get full access to Newcomer at www.newcomer.co/subscribe
00:00:05
Welcome salad. Hey it's Eric newcomer, welcome
00:00:15
to dead cat. Another great episode.
00:00:17
I'm excited to invite two people onto the show.
00:00:21
We have a Lexus gay who is the host of the podcast
00:00:25
non-technical. And I went on her podcast and
00:00:28
we've stayed in touch and I Thought, maybe she would enjoy
00:00:31
coming on a play. The co-host role on a very sort
00:00:35
of technical podcast. So hey, Alexis Erik, thank you
00:00:39
so much for having me. I am so ready.
00:00:41
I'm ready to be technical. I know you.
00:00:44
You you've got metaphors Galore, you've been prepping so I'm
00:00:48
excited. Absolutely.
00:00:49
Alright and then the the one who gets to you know be subject to
00:00:54
the actual questioners here, we've got crystal ball,
00:00:57
Valenzuela the CEO of Runway who I actually encountered in one of
00:01:01
my early crazy generative AI hype cycles and I thought
00:01:05
normally, you know, I DM Founders just like am writing
00:01:08
about your company and, you know, they ghost me but you were
00:01:10
like oh yeah let's get on the phone and talk about it.
00:01:12
I thought we had a great call so then I thought you'd be a fun
00:01:16
guests to sort of explain. What exactly is going on in
00:01:20
generative AI. So, welcome to the podcast.
00:01:22
First, thank you for having me. And I'll do my best to try to
00:01:25
explain what's going on. It's crazy, right?
00:01:28
What? Just to start off.
00:01:29
I mean, can you You just give us some back.
00:01:31
Story of how you got into generative Ai and the artificial
00:01:36
intelligence world. Like how do you come to this
00:01:38
space in the first place? Yeah, for sure.
00:01:41
So, it all started back in like 2016 2015, I fall in love with
00:01:47
us algorithm and technique called Deep dream which is I
00:01:51
would consider like one of the earliest.
00:01:52
Like a I are kind of like models that was how there was very
00:01:56
neat. She was just like a handful of
00:01:58
Artie's and researchers were With it.
00:02:01
And I felt too much in love with it algorithm to the point, I
00:02:04
just left my job in Chile, got a scholarship, I didn't were you
00:02:07
and spent two years, researching kind of like, computational
00:02:11
creativity or applying like, AI models for Creative kind of like
00:02:15
purposes, and were closed and that's rabbit hole.
00:02:18
That's hard to get out of. How does one fall in love with
00:02:21
an algorithm? I was going to absolutely not an
00:02:24
experience I've ever had. No can't say the same, that's a
00:02:29
great question. I think for me was the aesthetic
00:02:31
quality in this study, kind of like possibilities that I was
00:02:34
saying, and it's kind of like earlier going them.
00:02:36
This is again before kind of the wave of outputs that you might
00:02:39
have seen today. It was more about that deep.
00:02:42
I would say two things. One I fall in love with the idea
00:02:45
of using neural network techniques to generate images to
00:02:49
generate video, right? To generate the kind of like
00:02:51
Explorer and type of aesthetic and creative possibilities.
00:02:54
Just wasn't like, even imaginable before a tree like
00:02:58
that idea of like exploring Uncharted.
00:03:00
It is unimaginable territories and I thought and I saw this
00:03:04
early kind of like researchers and our he's kind of like as a
00:03:08
way as a first kind of like step towards like just experimenting
00:03:12
with us and seeing how we could go.
00:03:14
I think I followed a model of that.
00:03:15
One single algorithm would be a video of experimenting with
00:03:18
technology, right of trying and use it in ways that perhaps
00:03:22
weren't meant to be used. Writing you go in a that's the
00:03:24
rabbit hole. I feel I fall in love with but
00:03:27
long story short, that one thing led to the other and Runway was
00:03:30
my to whether my co-founders kind of like a research project,
00:03:33
was my thesis at art school that I went to.
00:03:36
And the whole idea is basically still, it's basically the same
00:03:40
that we have today, which is how do you take these algorithms is
00:03:43
generative AI systems and use them in the context of
00:03:47
art-making in the context of filmmaking in the context of art
00:03:50
or design, just in the consists of creative kind of like
00:03:53
processor. And I think what we've seen so
00:03:56
far right now, like last year 2020 was like a A year where
00:04:01
we're models got really good. Right?
00:04:03
And for me, there's first way was perhaps 2015 to 2020 to
00:04:07
wear, like some techniques and algorithms, some researchers and
00:04:10
Artis were playing with 80 was mostly seen as a niche thing,
00:04:13
right? So now, when I remember pitching
00:04:15
like and telling people like, hey, here's the thing.
00:04:17
You can use to generate an image, the first reaction I got
00:04:20
was like, Chris is a toy, right? This is you being an art student
00:04:24
doing art gold. That this is just, I will never
00:04:27
use this, right? And I think that like this
00:04:29
missing tonight, You're like that happens often when it's
00:04:32
very early and I think the way that we see now which is I would
00:04:35
say 20, 22 onwards and we just started on the second wave is
00:04:39
models. Have gotten really good, right?
00:04:41
So no, my mom like asked me about like and I saw Seaside
00:04:45
understand what I'm doing, right?
00:04:46
We just like the outputs of them are just so much, crisp and
00:04:50
investable that. I think it changed the
00:04:51
understandings of what we can do with it.
00:04:53
You said, you went to art school?
00:04:55
I did. Yeah.
00:04:56
So did you spend a portion of your life considering yourself
00:04:59
an artist Do you currently I tried, I tried I did, they
00:05:04
started like econ and then like design and also like doubling
00:05:08
like software engineering and then spend time just in like art
00:05:12
and research. And I quickly realized that I
00:05:15
guess my I wasn't really a good artist depressed and you're
00:05:18
barred. No, I think my art, I always
00:05:22
find excuses to make tools to make art.
00:05:25
And then I realized that those tools were far more interesting
00:05:28
and useful that the outcomes of those.
00:05:30
Right. At least for me, I think I've
00:05:32
come to, I come to the realization that perhaps my art
00:05:35
is just to making any really like making something that then
00:05:39
a very talented artist can take and use in ways that I can't
00:05:42
even like, imagine, right. And I think that just, that's a
00:05:45
very kind of, like, interesting position debate and in terms of
00:05:48
our Runway, is specifically, like a video tool is that right?
00:05:52
Or am I thinking about it in a, to limited way?
00:05:54
I mean, it's helping introduce sort of these generative AI
00:05:58
tools into video making. Is that right?
00:06:00
Yeah, definitely. I mean Runway you can think
00:06:02
about as a company. First of all we are unemployed
00:06:06
AI research company. So we we develop and build kind
00:06:09
of like low level, the basic fundamental algorithms for
00:06:13
Content generation and it mean content because and work with
00:06:16
images with video, we just multimedia formats, right?
00:06:19
So we have a research team that sits at the core of the company
00:06:22
of a product that researching drives innovates and develops
00:06:27
novel techniques. For kind of like driving
00:06:29
generation of Content. And then we will products that
00:06:32
the leverage that sort of like algorithms and techniques.
00:06:35
And there's things you can do on the video side.
00:06:37
So we have films and post production, companies and
00:06:41
broadcasting companies, and Studios, and just teams in
00:06:43
journal, working on the video side of things leveraging and
00:06:46
using our kind of algorithms and products and we also have a set
00:06:50
of like image base Solutions and products as well.
00:06:53
Do you have a dolly or like a stable?
00:06:56
Do you have sort of an image Creator?
00:06:58
This public. Yeah.
00:06:59
But think of me, Stable diffusion, we run.
00:07:01
I know you're part of that. I'm going to talk about it.
00:07:03
But do you have your this Runway, the company have a
00:07:05
separate sort of image product? Yeah, we have our own 30
00:07:10
different tools, right? And every tool uses a different
00:07:14
algorithm where I could be nation of different algorithms
00:07:16
to either generate a video edit, a video generate an image or
00:07:20
translate from an image to a video, right?
00:07:22
And so what's interesting for us is that we're not for our
00:07:25
customers, we don't you get access to all these models
00:07:27
behind the scenes, but you are not necessarily Kind of like
00:07:30
aware of those models, right? So, from a filmmakers
00:07:34
perspective, if you're in our director of your designer, if
00:07:36
you're editing the video, you don't have any study in care or
00:07:40
have to know about, like the complicated like, ins and outs
00:07:43
of these models and the techniques and the names, and
00:07:45
all this in the same way that if you use Photoshop, you don't
00:07:48
really care about the blurring function or the algorithm that
00:07:50
allows you to blur an image. We just use the image right, you
00:07:53
just use the tool. So for us, it's that parameter,
00:07:56
it's like it's about human intention.
00:07:57
It's about you coming to the tool.
00:08:00
You working with this algorithms in very intuitive ways to the
00:08:04
pain. You forget.
00:08:04
You're working with this kind of like, quote, a I register you
00:08:07
work with that tool that helps. You achieve something bring
00:08:09
particular but behind the scenes.
00:08:11
Yes, we have like those models that basically just power the
00:08:14
whole set of features and products that we have my
00:08:16
understanding. Correct me if I'm wrong is that
00:08:19
sort of generative, a, I really blew up in the public
00:08:22
Consciousness. When stable, diffusion an open
00:08:25
source project with ties to stability AI a company.
00:08:30
You and your co-founder, I believe that Runway was one of
00:08:33
the head researchers on that project and then that project
00:08:37
became open to the public right before that, you know, these
00:08:41
projects had all been sort of behind the scenes open a.i.,
00:08:44
despite the name hadn't released, there's and then
00:08:48
stable, diffusion comes out, and sort of the world goes crazy,
00:08:50
open, a.i. puts out Dolly, am I getting that right?
00:08:54
And where do you sort of fit in to that project?
00:08:57
And how much are you involved with?
00:08:58
Yeah, I think. Actually you're right.
00:09:00
I think so. It might look like there's an
00:09:03
overnight success over a specific model that lets say the
00:09:06
space has gotten like so much attention.
00:09:07
The last couple of months right in might look like, oh something
00:09:10
happened in a specific, like they are mouth.
00:09:11
But the truth is that the research has been brewing for
00:09:15
like years, right? We working on runway for the
00:09:17
last four years and he takes a lot of like iterations on
00:09:20
learning and training and learning a lot from like what
00:09:23
works, what doesn't work, what data sets you can use what
00:09:26
models. Can you train?
00:09:27
What are the possibilities of that, right?
00:09:29
So the The actual code and model and paper that powers and the
00:09:34
fundamental architecture behind Civil Division was was released
00:09:37
actually like almost two years ago, right?
00:09:39
And that's open source. It's been open source.
00:09:40
Since we publish it that work is the collaboration of Runway and
00:09:45
Ellen do you mean which is a research organization in
00:09:47
Germany? And we do this a lot, we
00:09:48
collaborate with other researchers, the Patrick Sr,
00:09:51
who's our research, principal research scientist lead, those
00:09:54
efforts, right? And we publish that code, an
00:09:57
open source it for a conference that was like, released on 20
00:09:59
lat. 20 21. And then seems then that was
00:10:02
actually called late and efficient.
00:10:04
I was out late and efficient work, right?
00:10:05
It's diffusion is a technique and Layton is like a word to
00:10:08
describe like a the space of like possibilities in like am an
00:10:12
Nei so we basically call it later.
00:10:14
Diffusion, there's a technical term that's actually longer than
00:10:17
that. It's high resolution image
00:10:19
synthesis with late in diffusion models.
00:10:21
Sure slate and very good consumer facing so that yeah you
00:10:26
can see and then we treat him on the model we first.
00:10:30
In the first model. And then each rated on a more,
00:10:32
he took training on more data sets for longer and then a
00:10:35
company called stability, kind of like offer compute, basically
00:10:38
came in as an infrastructure and just say, hey guys, I saw your
00:10:41
open source project, I'll just throw in some computer and you
00:10:44
guys can train it on a larger data set.
00:10:46
And that's that basically was were simple, diffusion was born.
00:10:52
And then there was, you know, I love drama.
00:10:53
There was a little drama when you updated you guys updated
00:10:57
that software, right? And there was a little back and
00:10:59
forth, With about, whether you are allowed or not, or what was
00:11:02
that whole thing. Yeah, there was I think there
00:11:07
was some confusion to Venice again.
00:11:09
We published that we're going most two years ago and it's been
00:11:11
open source. We've been improving 8 and
00:11:14
somebody donated computer. And I think that had an internal
00:11:16
like confusion or conflict about like ownership that they kind of
00:11:21
like apologize about both publicly and privately.
00:11:23
And I think we just like steer and unnecessary conversation for
00:11:26
us, it's like we've been working on this for love a lot of time
00:11:29
for a lot of years. And we publish newspapers, we
00:11:32
have will continue to upgrade our During diffusion work and we
00:11:35
have already like new models that are going to like change
00:11:38
the landscape. If table division was like
00:11:40
inflection point like q1 if - this year's going to be like
00:11:44
probably ask you guys that one, but I think I'm blessed answer
00:11:47
but like the drama to be honest, I know people like like the
00:11:49
drama, but I think it's just like a instrument confusion.
00:11:51
I'm like, I'm glad that something like this.
00:11:54
Yeah, I'm glad I'm glad it's figure out like that.
00:11:57
I know. I think it's just like, God,
00:12:00
this Stakes here are just and this is classic, you know, we
00:12:03
experience An open source software projects before, but
00:12:06
it's who owns what and the stakes feel bigger with AI
00:12:11
because it's being presented. Is this sort of you know,
00:12:14
everybody's gonna lose their job or well you know like it's going
00:12:17
to change the world. And so I guess my question is
00:12:20
just like yeah like all these new you know, when when open
00:12:23
a.i. comes out with GPT, for when you update your software,
00:12:26
new versions of like Sable diffusion, how much of those
00:12:29
projects will be in the public domain where any Company can use
00:12:34
them and how much are they going to be owned by particular people
00:12:38
because my understanding with Runway was that?
00:12:40
Like part of how you were trying to differentiate was sort of
00:12:42
like the tooling and making sure you can apply the publicly
00:12:48
available. Artificial intelligence
00:12:50
technology to specific use cases or are you also trying to own
00:12:55
proprietary algorithms? I think every company might have
00:12:59
a different strategy and a different like Focus.
00:13:01
I think for us something with we've defined barrier.
00:13:03
On that has taken us some time to like, really nail down has
00:13:07
been disabled of owning your stack, right?
00:13:10
So we don't build our products on third-party like models for
00:13:14
software data sets, like we tried to build the whole thing
00:13:17
and to end. And the truth is that that's
00:13:20
that's hard because it requires you to really nail or understand
00:13:24
how to build research, how to build infrastructure, how to do
00:13:26
products, right? The advantage of that is you've
00:13:28
build that muscle and you trade enough is that you can change
00:13:31
any part of the stack, right? So if I want to change.
00:13:34
You want to improve a model because specific customer wants
00:13:37
a better version of whatever we have.
00:13:38
I don't have to ask someone else or I don't have to wait for
00:13:41
someone else to do it. I know how to do it and I can go
00:13:44
there and like do it, right? I can go to the lowest level
00:13:47
possible have a model is trained.
00:13:49
How mall is optimized other markets?
00:13:51
Deploy and then go back to the surface level and change that
00:13:54
for them? Right.
00:13:55
I think that's to your, my proprietary algorithms that
00:13:58
you've built. Yeah.
00:14:00
Of course we have with that, we have custom research and Open
00:14:04
source, some research. So he built our research
00:14:05
website. You'll see that the latent
00:14:07
efficient, the Civil Division work was something.
00:14:09
We open sirs. There's other work that we have
00:14:11
an open source yet will perhaps oppressors at some point.
00:14:14
It's like but it's a combination of both.
00:14:16
I think our open source strategy so we start to community
00:14:20
building like the the wave of creativity that emerge.
00:14:23
After releasing stable diffusion was just like really, really
00:14:27
incredible to see and that just helps Drive the field forward
00:14:30
which is exactly what I think happened.
00:14:32
There was like an inflection point It's a great great of
00:14:35
hiring as well. You attract talent and your
00:14:37
truck researchers who want to work with you and like, people
00:14:39
read the papers and people read the author's, and you go back
00:14:41
and kind of like, we get a lot of, like, really tiny delegation
00:14:45
from that. But we're not our kind of like,
00:14:47
approach perhaps different from other companies is that we're
00:14:49
not an open search company. Like, our product is not open
00:14:52
sourcing products, right? Or products to help creatives.
00:14:56
You use these techniques in comprehensive ways in their
00:14:58
day-to-day lives crystal ball. I have some questions about
00:15:01
that. So one good thing.
00:15:04
I'm curious about you just use the term wave of creativity said
00:15:07
after these tools came out, there was a wave of creativity.
00:15:10
I largely saw on Twitter, people typing like bird flying over a
00:15:17
truck, that's on fire and also elon's driving the truck.
00:15:21
You know what I mean? So I my question is when you say
00:15:23
wave of creativity using these tools, like can you help us
00:15:26
understand a little bit about what kinds of things were
00:15:28
happening? Using the tools?
00:15:30
Yeah, for sure. I think every new technology and
00:15:34
Enables us to think about our creative in different ways,
00:15:37
right? But like the ability for us to
00:15:38
type something and generated image, just like it's novel,
00:15:41
it's new, its we've never seen it before, like some people were
00:15:44
experimenting with it 10 years ago but now it's just you open a
00:15:47
browser and just type something, you get something out, right?
00:15:50
And I feel like there's a lot of times where dislike novelty
00:15:54
component of it is just trying new things.
00:15:56
You just tried to we're distant possible, reduced, Sherman,
00:15:59
alexie you weren't exactly what you're saying.
00:16:01
You want to see how far you can go, right?
00:16:04
Andre, like that's like a Curiosity Company.
00:16:07
Like, everyone is trying to break that systems.
00:16:09
Like, okay. And when I like really try
00:16:10
something like really hard, right?
00:16:12
And I think that's like when you realize you're dreaming and
00:16:15
you're like, oh my God, I'm Dreaming right now.
00:16:17
What should I try? And I think that's the, that's
00:16:21
the 100, a good that sticks fermentation finger to curiosity
00:16:24
face. You're like Justin, just okay,
00:16:26
there's a experimenting with it, right?
00:16:28
You're like this is new. This is different.
00:16:30
I want to understand that. What do you mean?
00:16:32
But I think true creativity. Edie comes after that, right?
00:16:35
After you've like, okay, that's has settled down a little bit
00:16:39
and you start experimenting with seriously, right?
00:16:41
You start to think about it. How you might incorporate it in
00:16:44
like a filmmaking process, right?
00:16:46
And if you talk to filmmaker for an art, director is not about
00:16:48
the craziest like prompt that you can generate.
00:16:51
Something is a lot about like control.
00:16:53
Like, I want to generate a variation of an image.
00:16:55
I want to have I want to change the color or the laugh and a
00:16:58
quarter over the right. How do I do it at scale?
00:17:00
Can I do it? Like in 1, like iterations
00:17:02
per second like Those are different kind of like
00:17:05
components. And for me, something I hear a
00:17:08
lot about is this idea that AI or like the systems are
00:17:11
automating creativity. Right?
00:17:13
I think for me, so curious about that but I'm a comedian by the
00:17:16
way for contacts. I'm a comedian and I have a
00:17:18
podcast. But prior to that, I worked in
00:17:20
tech for seven years, just like a classic career path, as, you
00:17:24
know. But anyway, so I have a lot of
00:17:25
fun. I have a lot of thoughts and
00:17:28
questions about this idea of a, I supported creativity or a I
00:17:32
supported during I say content or a I supported art and all
00:17:37
kinds of implications of that and so I would love to hear a
00:17:39
little bit more. How much riding Alexis have you
00:17:42
tried to write jokes with, okay, so here I tried to what did I
00:17:47
travel? I saw the, I saw a i generated
00:17:51
versions of my own tweets, that's what I saw using some
00:17:55
software that someone sent me, which I'll have to look up the
00:17:57
name and it was terrifying because they sounded like me,
00:18:02
mine would be like, weird. Our fights and like, I would
00:18:05
reside article just what I need to look at it and I generated
00:18:09
version of Arabic a tweet, just be a bunch of fake Scoops.
00:18:12
I need to look. Yeah, that's so good.
00:18:15
Yeah, so I love this idea though.
00:18:16
That it's a phase you're saying right now, or maybe perhaps when
00:18:19
it first came out, we were in the Curiosity phase and that
00:18:22
implies, you said true creativity.
00:18:23
Comes after that. So what do you think that means?
00:18:26
I think for me, creativity is a state of mind.
00:18:30
It's a way of looking at the world, right?
00:18:31
It's not at all. It's not our process, right?
00:18:34
And I think what you can truly automate it is the processes of
00:18:37
executing a creative idea, right?
00:18:39
I think you can you can go back to like like previous moment in
00:18:42
time or like there was a major technology like change and
00:18:46
understand the changes that that transfer that broad to
00:18:49
creativity, right? So I always go back to like one
00:18:52
of the earliest examples of technology changing art and
00:18:55
creativity and how we look at the world, which was before,
00:18:58
like, the 1700s painting was like, mostly the realm and the
00:19:02
space and the possibilities. Realities of like folks who have
00:19:05
the money, the resources and the time and the knowledge to create
00:19:09
these very complex pigments, right?
00:19:11
And creating a pigment was a very sophisticated thing.
00:19:13
It has like this history and tradition, you had to like her.
00:19:16
These Masters that were very expensive and you have to like,
00:19:18
mix these very obscure colors, right?
00:19:21
And then storing the pigments with like, you had to use a pig
00:19:23
bladder and Link silly with a string.
00:19:25
Grant is very like obscure, right?
00:19:28
And then 1800's, someone basically, John Brown's inventor
00:19:32
came up with the collapsible pain too.
00:19:34
Right. Just that's the piece of
00:19:35
technology, right? You can take colors and put them
00:19:38
in like a pain tube. And the most important thing is
00:19:42
that you can take that paint tube out into the world, right?
00:19:45
And so impressionism was born, right?
00:19:47
You had people going outside and having like canvases like small
00:19:50
canvases, you don't require like a whole studio to just paint,
00:19:54
right? And so what that meant is that
00:19:56
you start looking at the world in a very thing way, people
00:19:58
started painting know just like portraits of like the royalty
00:20:01
but like mountains and trees and like so.
00:20:04
She'll like Advanced, right? And like go this career TV to
00:20:07
like the merchant. That's what I mean by that.
00:20:09
Gertie the opportunity that's like your that's a piece of
00:20:11
technology that enables you to look at the world and express
00:20:15
your view of the world in a very different way that everything
00:20:18
you've used before, right? And fast forward, it's like that
00:20:21
the naval huge revolution in art and like the transfer ever
00:20:24
transform, how we think about painting art, it set up, right?
00:20:26
And you can think about very similar moments in time where
00:20:29
technology is like like that have changed, right art and
00:20:32
creativity and Photography is another one right photography
00:20:36
change. We also get the world and a lot
00:20:38
of the times early on the further, graffiti kind of like
00:20:41
wave. People said that like
00:20:44
photography was kind of like the and of hard, right?
00:20:46
You, it's gonna be the end of like paintings, right?
00:20:49
Like you, I never just like, and the truth is, it's not.
00:20:51
It's just a different medium, right?
00:20:53
Just need to explore that beat him in a way.
00:20:55
There's new artists that will emerge from that.
00:20:57
New ideas will come after that, right?
00:20:59
And you open the door for something that's the business is
00:21:03
very hard to. Predict, right?
00:21:04
Because you tried to predict something.
00:21:06
You haven't even yet, experimented with red.
00:21:09
And it's very hard to like make this set of assumptions over
00:21:11
time about everything that needs to happen.
00:21:13
I think I come to embrace that like uncertainty, as in like
00:21:17
early on something that's going to massively transform, a lot of
00:21:20
things, and we should embrace this like just experimentation
00:21:23
phase like that Curiosity face, right?
00:21:25
And the kind of things that you'll build after that, as yet
00:21:28
still to be like uncover and discovered and that for me is
00:21:30
really, really exciting. I mean I have such conflicting.
00:21:34
In thoughts on this, like, on the one hand so many like
00:21:37
amateur visual artists to me, like, what they enjoy is like
00:21:42
the physical process over the output, right?
00:21:46
And even if they can get better outputs from a sort of an AI
00:21:51
system is that cutting into some of the enjoyment, I get, the
00:21:55
digital are also in photography and everything is cut into that,
00:21:58
and people get their choices on the other hand.
00:22:01
I'm like, so excited by the idea of like Pairing the human mind
00:22:06
with like another sort of intelligence or I mean there's
00:22:09
such there. I feel like you know, Alexis is
00:22:13
wearing a chest sweater. It's like if you get good enough
00:22:15
at chess you feel like a certain connect.
00:22:18
You know, you can use these Bots to like, figure out how you
00:22:21
should play better. There they're like these, I feel
00:22:23
like they're already experienced this today where you can sort of
00:22:26
feel in sync with like a robot system or I play a lot of bridge
00:22:30
and my partner, and Bridge online is is a robot, I sort of
00:22:34
and what they do. And so having sort of this
00:22:36
partnership with some sort of intelligence is, you know, a
00:22:41
very Charming experience. But yeah, it's definitely a
00:22:46
terrifying future. Are you worried about it at all?
00:22:48
Like, do you ever wear? It's just inevitable to you.
00:22:53
Yeah, I'm not worried a lot. I think actually, it's a very
00:22:55
exciting time to be in a very optimistic about it.
00:22:57
I think like radical technologists.
00:22:59
Like, this are, like, one of the things I did do is I deleted the
00:23:03
cause And enable things to be more convenient, more easy to
00:23:06
use, right? So think about like filmmaking,
00:23:09
the people who had access to have, like, professional fee,
00:23:10
making tools, even nowadays is like a handful of people that
00:23:13
know how to like a date on professional software, right?
00:23:16
And I will dislike complicated. VFX work flows and need to
00:23:19
understand computer graphics and a bunch of things picture.
00:23:22
Having everyone with like ilm or like, bfx professional like
00:23:27
skills, right? It's a wave of like,
00:23:29
storytelling in ways that we haven't even thought of before.
00:23:33
It's just like right around. The corner?
00:23:34
What kind of tools we need for that?
00:23:36
Well, let's just build them, I don't know yet.
00:23:38
I'm more excited. I feel like more excited about
00:23:40
the outcomes of those and enabling more people to express
00:23:43
themselves in creative ways. And I think that's ultimately
00:23:45
the role of Technology like, at large.
00:23:47
I think I'm less interested in like me, save you again, that
00:23:51
you will automate it, and that we come like, some sort of like,
00:23:53
one click of solution and like everything will become the same.
00:23:56
I think that's just boring and like, it doesn't really like
00:23:59
interesting yet, to be honest. And it's also like a partial
00:24:02
view of how the world actually works.
00:24:03
I think. I think there's always this
00:24:05
human component on that that what gives us like meeting at,
00:24:08
Alexis you're saying like, we someone just generates tweets
00:24:10
that are, like, just actually there's a I've heard there's a
00:24:13
generator podcast, right? So you can generate the script
00:24:16
and you can generate the voice. And you have those Supply
00:24:18
conversation between like Steve Jobs and like Lex Friedman,
00:24:21
right? This is M.
00:24:22
Everything is generated, right? And if an interesting it's like,
00:24:25
yeah, I mean, works very well, you're convinced halfway through
00:24:27
this is actually happening, but I feel like there's still a lot
00:24:30
of potential of using that in way, more creative ways and just
00:24:33
generating the Entire thing, like n to n, right?
00:24:35
I hear what you're saying, and I agree with a lot of at night,
00:24:38
especially the point about the way, access to pieces of
00:24:42
Technology can really draw out. Someone's creativity.
00:24:45
I think my personal experience is a good example of that where
00:24:49
I was able to teach myself to edit video because I movie came
00:24:53
on my Mac, right? And then I know how to use
00:24:55
YouTube and then I graduated to other software but it was all
00:24:58
accessible, it was all just downloading thing.
00:25:00
Look at it. How to do this some
00:25:02
twenty-year-old, dunas basement told me on the And there's a
00:25:04
great I know how to make videos now.
00:25:06
But here's what's different to me about generative Ai.
00:25:10
And some of the other tools that have preceded it and assisting
00:25:13
creativity, which is that generative AI models.
00:25:16
And please correct me if I'm using the wrong terminology,
00:25:18
they have to be trained on something, right?
00:25:21
And so when we're talking about visual art or God forbid comedy
00:25:26
or anything else, do we have to feed it, existing artists work
00:25:31
in order to teach it how to do it and does not raise him.
00:25:34
Implications around ownership credit and profit participation
00:25:37
of the output of the AI. Yeah, totally.
00:25:39
I think those are very valid questions and things that are
00:25:42
really worth and fasting ourselves as we develop these
00:25:45
new technology. And and again, I'll go back to
00:25:46
like premiums in time. There's this video and this
00:25:49
recording of 35 years ago. Same questions being asked when
00:25:53
Photoshop was first released, right?
00:25:55
See ya. It's the same conversation older
00:25:57
show you guys the same condition, right?
00:25:58
Was original. I was like, this is an
00:26:01
interesting question. But it's but it's a valid
00:26:05
question. It's part of the commits
00:26:06
perfect. And there's, yeah, it is right?
00:26:08
And and and the interesting part is like they were discussing of
00:26:11
the time. It's like, are you allowed to
00:26:13
edit a photograph, right? A photograph is like the truth,
00:26:17
right? You can edit it, right?
00:26:18
You can't go into National Geographic.
00:26:20
Going to like TV magazine and edit any much, right?
00:26:22
And the half of the panel was like, it's just a little, this
00:26:25
technology should be banned, right?
00:26:26
And it can imagine now, like not having like a tool like
00:26:29
Photoshop in our disposal. So obvious, and like, Photoshop
00:26:32
became a bird. Right.
00:26:34
And so for sure there's a lot of things that you need to figure
00:26:36
out like Photoshop had to figure out a lot of things and kind of
00:26:39
like concerns about how to use the technology.
00:26:42
People were just need to like for like fake bills from the
00:26:45
like, create fake money, right? So if you try to open a photo
00:26:48
shot bill in Photoshop, Photoshop will like prevent you
00:26:50
from doing that, right? You can't just like it's your
00:26:53
battery. And so you build you build over
00:26:56
time ways of like, secure in the system, right?
00:26:59
But the 99% of the outcomes that are emerging from this They're
00:27:03
going to be net, positive are going to make Society new for
00:27:06
99%. I'm sorry.
00:27:08
I just want to go back. So you I'm just you on the
00:27:11
record just said that 99% of the outcomes.
00:27:14
That's a, don't you guys go? I think that's just a Boulevard
00:27:20
but I think. Yeah.
00:27:21
Most of the I would say that old put that on a slide somewhere.
00:27:25
Oh, of course. I mean, it's a, it's a very
00:27:28
powerful technology. It's going to be as massive as
00:27:30
cell phones were like. I feel like mobile was like,
00:27:33
Really net positive for everyone, right?
00:27:35
Like you're able to hold like a lot of knowledge in your pocket
00:27:39
and able to connect with anyone in the world, right?
00:27:41
That's just think that's not positive.
00:27:43
I think a eyes and gently vir similar Technologies, right?
00:27:46
Your there's of course like people they're going to try,
00:27:48
they're bad people in the world and then you try to use
00:27:50
everything they kind to do bad things for sure and like will be
00:27:53
contrary to prevent those but overall the outcomes of the
00:27:56
Technologies net positive like well we're allowed to
00:27:59
democratization of content. For example, as you were saying
00:28:02
of anyone having Access to the sings just like the to their
00:28:05
magical change. Yeah but I think to your point
00:28:07
there is there's a lot of conversation that needs to be
00:28:09
had and people are having around like data set and training the
00:28:13
models and one which data is and like how you do it.
00:28:15
But again like Google had the same kind of like discussions
00:28:19
early on right? I don't want my data to be
00:28:21
constrained by by Google, right. Kind of have and kind of have
00:28:25
like a waste opting out if you don't have one.
00:28:27
I have my website being like searchable and the answer is it
00:28:30
you can like just put this file on your website and I Google not
00:28:34
scrape. It and I would not index you in
00:28:36
the search file in the search kind of like options.
00:28:39
And so there's I think there's still a lot of things yet to be
00:28:41
uncovered and models are not 100% like ready yet to be like
00:28:46
using all sort of like professional environments were
00:28:49
still very very early. How important do you think the
00:28:51
humans are to generative AI or will be in the next couple
00:28:55
years? What we're talking about is
00:28:58
human sort of as creators experimenting using this tool,
00:29:02
but I another conversation On this podcast, we've talked about
00:29:05
Tick-Tock and you know, how sort of the content sorting algorithm
00:29:09
is so good. And, you know, there was this
00:29:12
question asked of like, how far away are we from Tick-Tock?
00:29:15
Just saying, oh, you know, not only do I can guess which videos
00:29:20
that exists do people want. But I can just create the video
00:29:23
that you are likely to want. Like, how far away do you think
00:29:26
that world is where somebody like Tick Tock, can just deliver
00:29:29
the video that I want without a human involved at all over.
00:29:33
I'm very, very close. I think we're not far away from
00:29:36
having laughing. Good, I'm not like, Oh, yay.
00:29:42
You saying that makes me scared. That does not sound fun to you.
00:29:46
Sir we I don't want to commit to specific data or moment but I
00:29:50
think like we're heading towards her world in general.
00:29:53
When a lord of the content of your consumer line will be
00:29:55
generated picture like a YouTube, like generated like
00:29:58
stream, right? You can do Netflix generated
00:30:00
content like Tick-Tock generated content, right?
00:30:02
I think that's Somehow visible like today, if you combine a
00:30:06
bunch of different things but there's that's still a lot of
00:30:08
things to be developed to get to a point where like you'll do it
00:30:10
real time. I think it's under happened, if
00:30:13
it works, if it's possible. Any further comments and allow
00:30:17
us to explore new avenues of like creativity.
00:30:19
I think that's that's going to be great.
00:30:21
I think a lot has to do yet with developing this bottles in more
00:30:24
safe, Miners and more kind of like a line ways to our human
00:30:28
intention. And there's a lot of work to be
00:30:30
done still yet there to prevent possible.
00:30:32
Misuse is for sure. Sure, but overall, I think we're
00:30:35
going to be in a time in moment, where like you're going to be in
00:30:38
every single movie you ever wanted to be right, you are
00:30:40
going to be, you're going to eat them.
00:30:42
A nice thing about it is not only that it can create
00:30:45
something but it can create something specific to me that I
00:30:49
will. Yeah, and compelling.
00:30:51
Yeah, 100% 100%. And but and we you think this is
00:30:54
like three laying out like, yeah.
00:30:56
What's your soon? I'm like which how do I need to
00:31:00
prepare my job security? Like what does my timeline Need
00:31:03
to be to learn to code. I just need a sense for when I'm
00:31:06
going to be at replaced by G, PT 3.
00:31:09
I think very soon, a couple of years.
00:31:11
I would say yeah that's good. Your therefore there's
00:31:15
definitely like an exponential like like progress tray that you
00:31:20
can see and perceive more clearly.
00:31:21
Now that like here's the progress of what Take Years with
00:31:25
two gears, for progress is now taking like months, right?
00:31:28
Yeah. And we'll what will start to
00:31:30
take months with Senators start to take like weeks, right?
00:31:32
And you go by. That like all the time like
00:31:35
large language models and now going to be writing like papers,
00:31:38
I'm going to be writing code, I'm going to be and you started
00:31:40
accumulating. The amount of progress that you
00:31:42
can make, right? And I think that's yet, it's
00:31:44
happening and it will continue to have him, but I think that's
00:31:46
going to happen like sooner than we think.
00:31:47
I mean I respect you for doing your best.
00:31:50
Not to give us a date because you know I covered the
00:31:52
self-driving car industry. Any of that was an artificial
00:31:56
intelligence powered industry. That always said you know full
00:32:00
autonomy is just over the horizon.
00:32:02
Like what do we take From that experience.
00:32:04
Why is this different than that? I mean, things.
00:32:07
People have gotten excited about chatbots before things can
00:32:09
appear, you know, if you solve 90% of the problem, but you need
00:32:13
to solve the whole hundred percent, it can be easy to
00:32:16
convince yourself that you're like almost there.
00:32:18
But the last ten percent can be can be extremely difficult to
00:32:21
Impossible. I think it depends on what your
00:32:24
goal is and what you're trying to accomplish writings I guess
00:32:26
in the case of self-driving cars like full autonomy, right, can
00:32:29
you drive a car in any Street in the world and like with no?
00:32:33
Get right? That's that's the end goal right
00:32:36
for us. It runways not.
00:32:37
I can you make a piece of art or can you make a video with no
00:32:40
human intention and no control at all?
00:32:42
Right, that don't think that that's not the goal.
00:32:43
Their goal is like, the goal is like, can you take an idea and
00:32:48
execute that idea in the fastest way possible, right?
00:32:51
And right now if you want to create a video like Alexis you
00:32:54
want to make a video 20 30 years ago you would have taken your
00:32:58
like months, right? You got to like rental equipment
00:33:01
and then like and have someone have been?
00:33:03
Cost prohibitive as well as yeah.
00:33:05
It's too expensive, right? And so your idea is like, I want
00:33:07
to make comedy. Want to make this, you should
00:33:09
you hand. I mean, usually it's when I take
00:33:11
you months and you need to find a producer.
00:33:12
I like all this. Yeah.
00:33:13
And the fact it's like technology and like smartphones
00:33:16
on the internet. God you like 90% there, there's
00:33:18
still, of course work. You have to do like coordinating
00:33:21
on recording and editing, Etc, but like it's made, like 90% of
00:33:25
the work, right? You were like, so it's so much
00:33:27
better so much visible even to even consider doing something
00:33:30
like that. So if the goal for us is like
00:33:33
to, Help create. It's just like, get over that
00:33:36
like 80 90 percent of the work that you don't want to do as a
00:33:38
creative. Maybe just like no one wants to
00:33:40
like spend time. For example, searching through
00:33:43
hundreds of videos and then copying those videos and then
00:33:45
placing them to express the idea that you want to have in your
00:33:48
head. It would be great if you can
00:33:50
just like how to make that, right?
00:33:52
And that's like that's 90% of the process, right?
00:33:54
And I think that the goal for us is you measure by that?
00:33:57
I think we're very close to getting to that point, right?
00:33:59
And that's the ultimate goal. I would serve dirty eyes like
00:34:02
taking the cause of Ten creation down to zero, not the cause of
00:34:05
like, ideas, right? IDs are still like ideas and
00:34:08
like the best ideas will still win, right?
00:34:10
I think this is the time where everyone will become more of an
00:34:14
editor and curator and like AV has the best ideas will be
00:34:17
executed, not because they have the funding on the resources
00:34:20
just because they're the best ideas.
00:34:22
I will bet we have to we're playing sort of the cynics even
00:34:24
though I'm extremely excited about this technology but you're
00:34:27
doing a good job of making the bull case.
00:34:29
Do you worry that we're going to have even more of like a
00:34:32
bullshit? Mmm, in the world of social
00:34:34
media, there's already been the problem.
00:34:36
You know, if you have a big account, you have people writing
00:34:38
to you and just the work of like, sorting through among
00:34:41
humans. What's like a reasonable
00:34:43
critique to spend time an electrically and engage with and
00:34:46
what sort of a waste of your time is a taxing exercise.
00:34:49
Even if you're like, sort of a with it person and then if you
00:34:53
add to the mix, I literally on Twitter.
00:34:55
Had someone, I'm 90% sure, use GPT as like an auto response to
00:35:00
a bunch of tweets to like, juice engagement.
00:35:03
I responded like GPT and they send me a smiley face and it was
00:35:07
obvious to me. But it's still like, oh my God.
00:35:10
If I'm gonna run into that all the time where I'm going to have
00:35:12
to engage with the cognitive task of is this gbt.
00:35:15
And of course, the problem with it being GPT is G, PT or
00:35:19
whatever text text tool, they're sort of bullshitters.
00:35:22
They're very they're 100% confidence and maybe 90%
00:35:26
accurate and that's very challenging for human beings.
00:35:29
Yes. What do you say to the
00:35:31
bullshitter problem of AI? That's a good question.
00:35:35
I think. Overall I'm optimistic that that
00:35:37
we get solved in the same way that like, spam and like
00:35:40
scammers got solving like mail. And in the Internet it's like if
00:35:45
you can receive an email from like, a prince of Nigeria,
00:35:47
you're like, yeah, this is, this is right?
00:35:50
And this is like a very common like thinning internet.
00:35:52
I like you build filters and you've been making its into the
00:35:54
tech bad and like people lab figure out and governors.
00:35:58
Like there's ways of like improving those systems to like
00:36:01
avoid that right now. What's happening Bending, is
00:36:03
that it's still very early, right?
00:36:05
Like, if you consider like chop, GPD tragedy was like a research
00:36:09
release, right? It wasn't like a product, right?
00:36:11
And it was just like, right the day when taken by surprise how
00:36:13
popular was a, all right? Yeah, it in the bottom.
00:36:17
It says like this is our research released, right?
00:36:18
There's a lot of things that you have in yet consider, but just
00:36:22
trying to learn how people use it, the possible, like things
00:36:25
that need to be fixed. And I think that a big
00:36:27
misconception is to look at this and be like, okay this this is
00:36:29
it this is the final thing will use ever and it doesn't work.
00:36:33
Right? I'm like, no, it's just like a
00:36:34
point in progress and time and I will continue to improve it
00:36:38
until our point where like, all that bullshit that's like you
00:36:41
will find and they say things will be like, either reduce or
00:36:44
remove or like prevent it and still a lot to be to be done for
00:36:48
sure. I think that there is cool
00:36:50
stuff. I agree with a lot of what you
00:36:52
said around taking the burden of some of the more like task task
00:36:58
pieces of the creative development process.
00:37:00
I think that there's a lot a lot that could be good there.
00:37:03
I'm just scared. That's why I'm scared at all.
00:37:08
We've talked a little bit about like, oh, eventually people will
00:37:11
figure this out, and I'm sure that like things will happen
00:37:14
down the line. I'm kind of like this seems to
00:37:16
be going awfully fast, and I don't know.
00:37:19
Who is ultimately responsible? Like who's the adult in the room
00:37:21
here? That supposed to put a hand up
00:37:23
if you like, we should probably have some rules socialist
00:37:26
realism to it. I think so much of what happens
00:37:28
is just if it's possible, it will happen and that's how I
00:37:31
yeah I think sort of Of assumption.
00:37:34
Underlying some of this conversation.
00:37:35
Is that like the learn to code crowd is going to win the word
00:37:40
cells. Writers Among Us are the ones
00:37:42
who are going to be screwed. Do you actually agree with that
00:37:45
premise? I mean, we're seeing these like
00:37:48
programs, do some coding or like, is it clear to you like
00:37:52
which human skill sets? Come out on top.
00:37:55
If this all happens, is you see it?
00:37:58
And which ones are the losers? Yeah, that's a good question.
00:38:01
I think I don't have a full answer.
00:38:03
Whereas because I think a lot has the yet be like, discover,
00:38:06
and like understood for like World models can actually go.
00:38:10
I think, with every kind of like previous moment in time in the
00:38:13
past for something, like I said happen.
00:38:15
Like there's like jobs and disciplines unlike things to
00:38:19
just disappear, right? Like you don't need the the
00:38:22
market for like people could do analog film editing has like
00:38:26
possibly being reduced, right? No one is just cutting films
00:38:29
like with Caesars anymore, right?
00:38:30
You can't find these people, right?
00:38:32
And the reason is that those folks have to adjust their jobs
00:38:35
and like, there's has two for a diesel age and they have to
00:38:38
learn a new task and understand the limits, and the kind of like
00:38:41
all directions of using that I think.
00:38:44
Overall, my hunch that we'll start to see that like
00:38:46
professions and jobs will radically change right software
00:38:50
engineer for, I think will become more about again, having
00:38:52
an idea and helping a system like execute the idea in a
00:38:55
secure way, right? So if you want to write right
00:38:58
now you can just like something like copilot and just write a
00:39:01
function and have like the copilot like Complete the whole
00:39:04
function for you, right? So you're not going into the
00:39:06
Commendation to try to understand apis and then going
00:39:09
back just like it works really really well, right.
00:39:11
But you still need to give guidance and you need to the
00:39:13
human need to understand your intention, right?
00:39:15
And you're like having the system just help you along the
00:39:18
way on that. Right?
00:39:19
So I think that that not sure like software Engineers are like
00:39:23
job itself is like not. It's still going to be here,
00:39:26
right? Just radically going to be
00:39:27
different. It's going to be sort of
00:39:28
function differently. It's gonna skills that might
00:39:31
require going to be different. You might not need to No, this
00:39:33
very obscure like the levels of, like, API documentation because
00:39:37
now you have something that can do it for you.
00:39:39
And for writers, I think it's the same, it's like pretty,
00:39:40
like, think about like, how like computer.
00:39:43
So you can typewriters change. The process of writing, you are
00:39:46
able to like Grace things easier, they're able to like
00:39:49
compose ideas more quickly. You're able to Like Remix thing
00:39:51
because you have them saved in different ways, right?
00:39:53
I think that just changes the nature of like writing and cool
00:39:56
so you can work. And right if I'm not an English
00:39:59
like, native speaker, like this algorithms can help me like
00:40:02
significant, Writing skills. And now I kind of have a
00:40:05
conversation with someone else in another language that I don't
00:40:07
have and you can be exposed to because I don't have and that
00:40:09
can kick-start something out. So I think it's more interesting
00:40:12
to have a like, a very open mind and very kind of like learning
00:40:15
mentality of like, okay. How is my job or they think I
00:40:18
do? The augmented and how we need to
00:40:20
be transform and change. And I think there's like that
00:40:24
the hesitation on the, the word conflict component of is very
00:40:26
natural like silent film was, like, the thing for 20 years
00:40:31
like in the early 1900's, right? And when Audio came to be a
00:40:35
feasible technology to have in cinemas.
00:40:38
The first reaction from like everyone in the Hollywood
00:40:40
industry is like we need to ban audio, right?
00:40:42
This is going to destroy the like and this is true like
00:40:45
Charles shut not want to be on that side.
00:40:48
Are we still recording? Charles Chaplin was like binding
00:40:54
audio like I will show you. I the adage is like the
00:40:56
association for like film and audio in like the musicians from
00:41:00
like Hollywood are bonding audio from movies, right?
00:41:03
Because Who's gonna pay for like the orchestra's that are in the
00:41:05
theaters, right? And of course, like something
00:41:09
happening for the, for the group.
00:41:11
A couple things you're saying that stick out to me, are one
00:41:15
there's potential here that the proliferation of these tools
00:41:18
that let's just focus on writing, like, help you write
00:41:21
could actually raise the bar. For what type of writing were
00:41:24
willing to an interested in consuming, because if everybody
00:41:27
can, let's say in a year quote, right?
00:41:30
Like to be T3, the bar is going to be Much higher for your
00:41:34
average, I think, article, or newsletter, or whatever because
00:41:37
you're going to have to be bringing something really
00:41:39
special to the table in order for people to spend their time,
00:41:41
reading it. Yeah, 100% maybe that's just me
00:41:43
trying to be I just want you to know I've sort of been in
00:41:48
Silicon Valley long enough and watch the government be slow
00:41:51
that. I'm I'm like whatever is
00:41:52
possible will happen and sort of nothing will put the cat back in
00:41:56
the bag. So worried about it too much is
00:41:59
you know, it's good to think through these things, but but
00:42:01
yeah, the technology is going to progress.
00:42:03
Yes. But I do think it's an open
00:42:05
question. You know what this does to
00:42:08
literacy and like all of a sudden people are just like
00:42:10
Outsourcing you know their professional writing.
00:42:14
You know, most writing is bad, most radius formulaic.
00:42:17
Yeah. But people learning to think is
00:42:20
important, I mean, I feel like there's been a lot of chatter
00:42:23
about the sort of, I don't know, High School, cheating problem.
00:42:26
I literally see it on teacher subreddits, you know, worrying
00:42:29
about the cheating problem. I don't know.
00:42:32
Are you worried about Out it, like, one solution.
00:42:34
I was saying is I assume the AI systems will just get better and
00:42:38
better over time. So to become easier and easier
00:42:40
to identify old cheaters. So, it seems like a big risk to
00:42:45
cheat when we're going to be able to figure it out, cheaters
00:42:47
more. I don't know.
00:42:48
Do you worry about that? I mean, I don't worry about it
00:42:52
too much. I think like it might help us
00:42:54
like, really understand like from perhaps from are more first
00:42:57
principle like thinking perspective, like while they go
00:43:00
love, like schools are like, it's not to memorize like words
00:43:02
like you need to Have critical thinking and like I'll learn
00:43:05
you, I'll teach you how to think and like do the world and
00:43:08
process it and come up with ideas, right?
00:43:10
And that's what we'll go look. Like, going to school to like
00:43:13
anything. Right?
00:43:13
If you want to learn something, it's not about memorizing stuff.
00:43:16
It's not about managing understanding how to use a
00:43:19
complex process and our workflow or whatever it is about.
00:43:22
Like look at the wall in a critical way having like a way
00:43:26
of understanding things, right? And I think if like if I learn
00:43:29
how to code with something like a language model that can Teach
00:43:33
me constantly that can fit me, that give me examples and, and
00:43:36
they can do a good software idea or I can execute something by
00:43:40
doing that. That's fantastic.
00:43:41
Right. Did I have to go through like a
00:43:43
formal training of reading this like books?
00:43:46
And this traditional process? Maybe it didn't, then the didn't
00:43:49
matter, right? I think like that might happen
00:43:51
and of course, it's normal. There's like push back at the
00:43:54
beginning, right? Because it's just it's
00:43:55
different. Right?
00:43:56
It's changed. And we're humiliating researcher
00:43:58
on dates. Yeah, we're very reluctant to
00:44:01
change. Like we're very reluctant to
00:44:03
change. And then the thing is like we
00:44:04
get used to change a lot. I was actually trying one of the
00:44:06
self-driving cars UNICEF a couple of weeks ago and the
00:44:10
first like ten seconds is just very scary.
00:44:13
You're like I want to get out just like yeah yeah just like
00:44:15
where's the driver? Like and then after a minute
00:44:18
you're like okay I'll just take my phone and just relax, right?
00:44:20
It's right for a mini relate, very reluctant to change and
00:44:23
then you've tried it and then you kind of like a Sumo K, their
00:44:26
works. And in like doesn't like crash.
00:44:28
I'm not dying. It can move on.
00:44:29
And like, it works really well and the human psychological
00:44:32
Journey on I'm embracing new technologies, a little sad.
00:44:35
Right? When something is exciting, were
00:44:37
very fearful about it and then like you're saying, we just sort
00:44:40
of embrace it. And then now, you know, when I'm
00:44:43
10 minutes late, because of the subway, I'm like Furious, even
00:44:46
though, you know, prior world's, we're getting across New York
00:44:49
City, could have taken me like days or, right?
00:44:52
And required a horse, right. Right there weren't two Bridges
00:44:55
to do it all, you know, like, so it's sad how much we take
00:44:59
existing technology for granted and then spend all our time.
00:45:03
Session over, what's right over the horizon.
00:45:05
But right that's that's Humanity.
00:45:07
I agree. Your optimism for this.
00:45:09
Technology is truly contagious because I really am.
00:45:11
I am thinking about it in less scary - cynical ways.
00:45:15
I'm thinking about it in more positive ways.
00:45:17
As what could it do to like you, use the word augment or Eric.
00:45:20
Maybe you did. Like, how can we look at these
00:45:22
as tools that will help augment creativity and not replace it?
00:45:26
Yeah, in terms of particular companies, obviously you're
00:45:29
excited about your own but what are The project sort of using
00:45:35
generative AI that you're most excited about right now or what
00:45:38
do you think people should be watching for?
00:45:41
Did a lot of things I've seen? There are some I think there's
00:45:44
some movie particular that I think for me, exemplifies, a lot
00:45:47
of other will continue to Syria over the next couple of years.
00:45:49
I don't know, you guys to watch it.
00:45:51
Everything everywhere all at once I love love.
00:45:55
Yeah. Okay.
00:45:55
They're all kind of kind of I know.
00:46:02
So there. Beautiful movie, right?
00:46:04
He's having watching the you're listening to these specific.
00:46:05
Whoa, watch it. It's like phenomenal movie,
00:46:07
right? And I think are quicker.
00:46:08
I seriously got the million guys.
00:46:09
I don't know if you guys remember is it has a lot of
00:46:11
visual effects, right? It's very intense usually thing.
00:46:13
That's right. The most interesting fact for
00:46:16
me, when I learned is and it was just like a Spanish about, it
00:46:19
was like the majority of those effects were made by five
00:46:23
people. Five people.
00:46:25
Right. Just five people, the didn't.
00:46:28
That's you think that didn't have previous experience
00:46:30
building like feature film complexities and Future?
00:46:33
Film effects like that were able to put together that kind of
00:46:37
level of like quality and like they're like very talented
00:46:39
people for sure, right. But it just fight people that
00:46:42
used to take hundreds, if not thousands of like wounds and
00:46:46
when I did, I was like wow you just like this is just insane
00:46:49
who are these people these are like super humans, right?
00:46:51
What's going on? I searched for those five folks,
00:46:53
a search for the like MDB. Provide proof of the movie, got
00:46:57
the five, kind of pfx, people behind the movie.
00:47:00
And, of course, like wanting to run away and see if they
00:47:03
Delegate counts, right? And all of them had accounts,
00:47:06
right? As I, that's phenomenal ways to
00:47:09
reach out to of all of them is like, yeah, we'll chat and we
00:47:12
chatted and they eat. They used to use parts of Runway
00:47:15
to edit a movie, right? It was just a small part, but
00:47:18
yeah, very validating. Yeah.
00:47:22
And and then my essentials like, well, if y wanted to like, I
00:47:26
guess what, I had the inside of like, I will search for them in
00:47:28
our database is because if you think about creating a movie
00:47:30
like that, the complexities of the movie already, That's right.
00:47:34
But you to be able to execute that you had to automate it, you
00:47:37
have to simplify it in a way, right?
00:47:39
And my assumptions like maybe they came across the runway
00:47:41
because it basically what we try to offer, right?
00:47:44
And so they did and we wrote a case study about it.
00:47:46
Now, some of them are already using it for other phones.
00:47:48
I'll send you the case study. And I think like, what I'm
00:47:50
really excited about is those kind of things right?
00:47:53
There are very highly creative teams with a lot of motivation.
00:47:56
A lot of, like, great to just understand how to get their
00:47:59
ideas out without having to have this budget restrictions or
00:48:04
There are no other set of limitations than just execute it
00:48:07
and it worked. It worked really well.
00:48:08
It's like the blue is going to probably going to win like a
00:48:10
dozen of dozens of awards. And so the journey would say,
00:48:13
I'm really excited about those teams that are embracing
00:48:16
technology like this that are like, experimenting with it,
00:48:19
that are using that are trying to push it in ways that I
00:48:21
haven't even thought of like, no one else I thought.
00:48:23
It's just like, you need to, like, expose it to more people
00:48:26
and I'm also excited about if you think about that, same
00:48:29
approach taken to product building, right?
00:48:31
So I don't have the research in the Space has been mostly led by
00:48:34
like domain experts. Right?
00:48:36
The researchers were like very deep into the way.
00:48:38
It's like phds who are like incentives to like make
00:48:41
benchmarks in X Y, or Z metric better.
00:48:44
But I think the outcomes and the results of using these models
00:48:47
are going to impact way more than the researchers, right?
00:48:50
And the insights are going to come from like multidisciplinary
00:48:53
teams. You need hackers in the Artis
00:48:54
need like more people jump in here.
00:48:56
So the companies that are more excited or the products and the
00:48:59
products, I'm excited about our products that can blend those
00:49:02
things. So you have We go, folks.
00:49:03
But I think that you have artists like speaking and seated
00:49:06
at the same table and be like, okay, here's what's possible,
00:49:09
Right? Would that be useful without the
00:49:11
not useful? How I'm going to how can I help
00:49:13
you out meant something, right? I think that's the second way
00:49:16
that I think I was telling you guys are early on like 20.
00:49:18
First wave of a I was pretty like these new wave of ideas
00:49:21
have been out. There have been a few
00:49:22
historically but 2015 to 2020. One was the first one in 2022
00:49:26
onwards. The next one is going to be
00:49:27
about figuring out how to collaborate with humans, right?
00:49:32
And how to take this algorithm From research, domains and silos
00:49:36
into real world examples. But part of what I'm taking you
00:49:38
to say is that, you know, there's a lot of emphasis on the
00:49:43
generalized intelligence and part of what you're saying is
00:49:46
that actually you need people focusing on specific use cases
00:49:51
and how they play out or is that the right contrast to be zeroing
00:49:55
in on. Yeah, I care more about humans
00:49:59
and I think they'll adjust has been the purpose of humans,
00:50:01
right? So we're helping you You to
00:50:03
something I think when we might lose that like, narrative or the
00:50:06
other direction, when you think too much about the analogy for
00:50:09
the sake of technology and that's the technology as a way
00:50:11
of serving humans in a particular way.
00:50:13
Wow, I know a lot of people who could hear that.
00:50:20
Not gonna say who this is a safe space.
00:50:24
I guess you can do it super safe space.
00:50:28
Do you mean like open-air? I mean, they're obviously the
00:50:31
elephant in the room Microsoft is reportedly like can invest
00:50:35
ten billion dollars. I mean, are they arrival of
00:50:38
yours is their success good or bad or neutral for you or like
00:50:41
what's your view on open eye and how do they relate to what
00:50:44
you're doing? I think opening, ISO research
00:50:48
driven, Station right there mostly about like, pushing the
00:50:51
limits of like research, right? And they've done an incredible
00:50:54
job at that. They've released multiple like
00:50:55
breakthroughs over the last 10 years, right?
00:50:58
They started much more focus on reinforcement learning and then
00:51:01
move because they discover new things and that's just how
00:51:03
research works. And I think that there are
00:51:05
pretty interesting unlike organization.
00:51:08
That's done, really breakthroughs in research, but I
00:51:11
think research is not enough for products.
00:51:13
If you think about products products about again, helping
00:51:16
serve human needs and helping People to achieve or summer
00:51:20
problem, specific way and part of it.
00:51:23
In this case, it's research because you can build the
00:51:25
fundamental Technologies but part of it is also.
00:51:27
How do people interact with it, right?
00:51:29
And there's a lot more that has to be built around it.
00:51:32
I think I don't think I was open as a competitor to the NS.
00:51:34
I think they're just research organization that's been do.
00:51:37
Use their apis or anything, we don't know.
00:51:40
No product. We build our own everybody, not
00:51:42
only might at some point. I'm sure, you know, more about
00:51:45
this than I do. But like, Jasper, whatever
00:51:46
superhot company in the tech space was built.
00:51:49
On top of open a.i. and then you know open it I comes out with
00:51:54
this chat gbt that ends up competing with if I don't write.
00:51:58
Am I crazy? I mean there are lots of
00:51:59
questions about like in terms of companies who's actually going
00:52:03
to be able to win sort of the war here.
00:52:06
I mean our bad is, you need to own your stack, you need to own
00:52:10
your technology, right? If you're building something you
00:52:12
needed to be able to like understand every piece of it.
00:52:15
Right? Because if not, someone else can
00:52:17
just take it off your right and four.
00:52:19
First kind of like Focus as a company has been just building
00:52:22
the full stack component, right? That's very interesting in a
00:52:25
way, it's sort of like asking the question is generative AI
00:52:29
more comparable to the technology that is the
00:52:32
smartphone or is it more similar to an operating system in that?
00:52:37
Will it matter? You know what I mean by that?
00:52:39
Like, is it, is it the technology that things will be
00:52:41
built on top of like, in the case of Runway, or are?
00:52:44
We talking, which is more, like, I guess you could consider it
00:52:47
like more commodified and less like Like owned or are we
00:52:50
talking about it as like oh it's equivalent to like a fully owned
00:52:54
and operated operating system or something like that?
00:52:57
I think it's pretty big. My hunch is more close to a
00:52:59
smartphone to be honest. Like, when you have I mean,
00:53:01
Apple has radically made that like the case, right?
00:53:05
You've you own your Hardware, you build everything from the
00:53:08
harbor itself to the software to the platforms where people build
00:53:12
on top, right? And that's a way stronger
00:53:14
business case or a business like argument than having like slide.
00:53:19
The gracious mostly because again, technology moves really
00:53:21
fast, right? And so you can consider anything
00:53:24
or you can give anything for granted, if things are moving so
00:53:26
fast and they're becoming obsolete so fast, we can let you
00:53:30
go or Alexis you have any final thoughts?
00:53:33
No, I think I really I really grown on this podcast.
00:53:38
I'm glad what do you think? I think cuz I think I started a
00:53:41
little more fearful and now I think I'm more open minded.
00:53:44
And also I am terrified of the fact that 20 years from now
00:53:48
we're going to listen Back to this and I don't want to sound
00:53:51
like the person that got interviewed by the internet in
00:53:53
the 90s and was like You guys heard of this internet thing.
00:53:55
Thanks. I'm really trying not to be.
00:53:58
I've been clear in my I think it's huge.
00:54:00
I mean I was super skeptical of self-driving cars because they
00:54:03
need to be complete. Whereas this side you know, it's
00:54:06
about the human interaction. I'm already seen you know like
00:54:10
professional type. People generating questions and
00:54:14
drafting emails and I feel like even if you're plenty smart to
00:54:18
write a good piece of It's just like solving it.
00:54:21
And then the point that you've made sort of very clear on this
00:54:23
podcast, that nothing is more valuable than what a technology
00:54:27
saves human beings time. And so, even if you don't view
00:54:30
it as like, okay, it's going to be more creative than us in
00:54:33
certain ways. It's like, if it's saving us
00:54:35
time to do creative tasks, that's sort of the core value of
00:54:39
technology and people will always want to save time.
00:54:42
So I'm, I'm super bullish on that.
00:54:44
I I still think. I mean we saw like lenses there,
00:54:47
you know, there's still a hype cycle going.
00:54:49
I-i'm. No, I mean, people get so
00:54:51
excited, something it's cool and then it falls off.
00:54:54
So it's about sustaining that enthusiasm.
00:54:58
Yeah, I agree that it's sustaining.
00:55:00
The work is sustaining do to set some underdevelopment more than
00:55:03
like specific spikes of hype. And so here for the for the long
00:55:06
run to make that happen, I think sometimes get things get better
00:55:08
after the hype cycle dies down a little bit actually because
00:55:11
there's less like fervor and like hand waving in the field,
00:55:15
you know? And the people who are genuinely
00:55:18
able to and passionate about Those tools like I think of the
00:55:20
Creator. Economy is a great example.
00:55:21
Remember the five minutes when like everybody, pretended to
00:55:23
care about the creative economy. And now money percent of those
00:55:27
people are not focused on that, but the ones that are are the
00:55:30
folks that are actually passionate actually competent in
00:55:33
that space. And I think that ultimately, the
00:55:34
tools is a result will be much better for it and I think
00:55:37
something like this and self-driving.
00:55:39
Cars are probably also parallel examples.
00:55:42
Totally cool. Well thank you both for coming
00:55:45
on. Listen to Alexis on
00:55:46
non-technical Chris Runway check Get out.
00:55:49
Alright, thank you very much for coming.
00:55:51
Of course, thank you for inviting me.
00:55:53
Thanks, Chris, goodbye. Goodbye.
00:56:06
Goodbye. Goodbye.
00:56:08
Goodbye, goodbye. Goodbye.
00:55:49
Get out. Alright, thank you very much for
00:55:50
coming. Of course, thank you for
00:55:52
inviting me. Thanks, Chris, goodbye.
00:56:06
Goodbye. Goodbye.
00:56:07
Goodbye. Goodbye, goodbye.
00:56:09
Goodbye.