Advice for Indian Engineers, AI and India, Wrapper Startups, and More! | Sundar Pichai’

Top Leaders’ Lessons
8 followers
1 Video View·Jan 3, 2025
We explore the groundbreaking advancements in AI and how India is set to lead this wave of innovation. Plus, we dive into Sundar Pichai's advice for software engineers, the future of AI startups, creative adoption trends, and more!
Chapters:
00:00 — Introduction
00:26 — India’s Position in AI
02:48 — Sundar’s Advice for Indian Software Engineers
04:48 — Thoughts on “Wrapper Startups”
06:13 — Creative Adoption of AI
07:50 — How Google Addressed Racial Inaccuracies in Historical Characters
09:31 — Conclusion
10:00 — Personal Question
Sundar, thank you so much for having me.
It's an absolute honor to be here.
I've seen so many keynotes of yours, right?
Like, through the screen.
And it's a fanboy moment for me to be here
you know, with you right across me.
And let me just get started, right?
I think somebody here mentioned
that this is AI Coachella.
And it felt like AI Coachella. Madness, right?
So many new things that came out.
The engineer and the creator
in me is super excited.
But I come from India and I feel like
you know, it feels like
this is an AI race
and Silicon Valley is moving really fast.
I feel we're sort of
getting left behind.
If you were an engineer back home in India
or you were an entrepreneur there
what would you do to keep up?
Well, first of all,
good to talk to you Varun and congrats on
a great channel and
the work you do.
Look, I see it a bit differently
when I think about
AI, even as
Google thinks about AI.
For us now, India is
one of our top markets, right?
We have, both for many of our
products,
it's our largest user base.
We have large teams in India.
So we've been thinking about
AI in the context of India.
The second is
unlike past technologies,
the underlying APIs
for AI are available
everywhere, including India.
So I see developers
already using it.
So I think India is a well-positioned.
I think,
in terms of,
what AI offers is for anyone
to use it and solve,
build applications and solve problems.
So I think to me, for India
this is the moment.
India has an extraordinary base of developers,
engineering talent.
So for them to understand this trend
and build applications
in a way that makes sense either for India or
for the world, from India,
I think both are exciting possibilities.
So I would view it that way.
From a Google
perspective, I'm excited because
one of the things we are doing
with AI is really breaking
barriers.
I think the fact
that all our new models are
multimodal, so they can reason across
text, images, video,
which means
the way it will come to India is you could be
watching any video you want in
any language, even if
the content didn't exist in your language.
Or people
who can just look at things
and ask questions.
So that is the future. So if anything,
I view it as part of
continuing to help Digital India,
getting more and more people
there, comfortable with technology
and giving them access to technology.
So I think it's huge.
And you were talking about software engineers, right?
I don't know if you know this, but there's an entire
industry in India built
to help young Indians
crack the fang into you.
And these are generally very
smart kids, but unfortunately
they're just glazing over the fundamentals
and then rote learning, data structures and algorithms.
I don't know if you've heard of this.
And now they see a big shift coming, right? Gemini can write code,
you'll have a pair programmer
available pretty quickly.
Do you have advice for these people?
How do you get
out of the competitive exam mindset with all
these shifts coming about?
I think it's a great question.
I think real success comes
from understanding things
in a deeper way, right?
And
almost tempted to go back to
movie Three Idiots or something like that.
There's a scene in there, I think,
when they ask
Amir Khan the definition of a motor.
And there's a version
you just describe what a motor is.
And there's a version where you actually
understand what a motor is.
So I view it the same way.
It quite doesn't change.
You have to
understand
the technology in a deeper way.
If you do, you can transition.
You can do things.
Programming languages may change.
The tools you use may evolve.
But you're understanding
the stack in a much deeper way.
So that's what I would encourage people to do
always.
Programming is going to get easier
in a way that's better.
All of us are already
thankful we don't have to write an assembly
code or
we're used to higher level abstractions.
AI will
help with even higher level abstractions.
You're going to use natural language. You're going to
tell sometime, hey, can you just debug
this code for me and explain what happened?
And it'll do it.
It's going to make you a better programmer.
So I think it's more important to
go deeper rather than trying to
learn the thing of that moment.
Interesting.
I don't know if you've spent time on Twitter
recently, but a lot of people
there's a word for
startups being built on top of models like Gemini.
It's kind of derogatory
in nature. It's called a wrapper startup.
You've heard of wrapper startups, right?
Do you have thoughts on that? I feel like
if you build for use cases, it should be
commercially viable, but we'd love to hear it from you.
There are going to be APIs
and people are going to use APIs.
I think that
is the way things will get built.
So, you know,
it's not bad to be
a wrapper. It's just you shouldn't be a shallow
wrapper where
I think you have to think about
where the models are going
and make sure you're doing something
which just in the next version
of the model will be done natively
by the model itself. Got it.
That will have a short shelf life.
But if you're using
AI to solve a real world use
case in a way
which involves a few more
things you're doing, or
chaining multiple models,
you have to think about the value
you're adding on top of the model
in whatever way, chaining
multiple things together.
And also thinking about the forward curve
of the models and make sure the models just aren't going to
do that. So I think it's a combination of both of them.
That's actually
a very profound way to look at it.
The value on top of the model is
what people will eventually pay for.
I started a company this year. It was called
AOS.
We were looking for
problem statements. We said models are going to be dime a dozen.
Some are going to be open source.
And you've actually contributed to open source quite a bit.
And one of the
things we realized, and I don't know who to attribute
this quote to, but
hallucination seems to be a bug
for tasks that require
accuracy. But it seems to be
a feature for creative
endeavors. And so slowly
we found that even our client base
sort of, they wanted
us to help them with creative endeavors.
Have you seen more enterprise adoption
by creative companies?
Very much so.
I think particularly
as people are thinking about
like, take marketing
campaigns for example.
We have good projects underway with like creative
agencies, advertising agencies
where they actually
want to push the frontier.
There are two ways you can imagine using it. Using it
to just kind of take the mundane tasks
and make it easier.
But there are also people who want these
models to come up
with things which otherwise they wouldn't have done.
And there you're exactly right.
I think hallucination is a feature, not a
bug. Right? And
that's the way I would think about these models
for that purpose.
And if you go back to the Alpha
Go game, there was a moment in which
the model move 37
I think, yeah, when it made that move
which no human prior to it had done before.
Right? It was a creative, intuitive
way of playing
the Go game. And that's what you hope these models
will do. Because they think in an
unconstrained way. So Google got into a little bit
of trouble recently with
you were
generating historical characters with
some racial inaccuracies.
Now,
I feel like typically that should have
stopped a company like Google. Or at least really slowed it down.
And it definitely has slowed
down some of your competitors. No longer shipping
publicly usable products.
But I feel it's not stopped you
this time. And I feel
today everyone on the internet
is a critic. Even if you do your best
work and you really put intent behind it and you're
being transparent about everything.
People are complaining. Do you feel like just
the number of people complaining on the internet
has gone up? And that's also given you air
cover to ship boldly?
We spoke about it last year at Google I.O.
We encapsulated
it as being bold and responsible.
And I think
with a technology like that
which is developing fast
but is still immature. You have to
ride the curve where you're pushing.
But you also
be as responsible.
But there is some trade-off.
And you're trying to find that optimal point.
I think there's a lot of user
excitement. So that's what is guiding
us to say people are ready for it.
Right? And
we are rolling out AI overviews
in search. People are
really embracing it and
responding to it. So I think
people are finding it helpful.
So that gives us the confidence
that we are in the right direction. Which is
why you see us moving fast.
But there will be moments
when we feel maybe this technology
is important
to spend more time
getting this right. So we'll balance it.
But on the balance, yes, we are moving
boldly. Yeah. I'm
very long-term bullish on Google. I think
a lot of this is going to get commoditized
but you own the means to
distribution. YouTube, Chrome,
Android. So thank you so much
for doing all you do. And
like I said, it's been an honor.
Likewise, Varun. Greatly enjoyed it.
Thanks.
Food in India. What's your favorite
food? Well, it's Bangalore.
I will probably get a dosa.
It's my favorite food.
If it is Delhi, a Chole Bhatura.
And if it's Mumbai, I'll do a Pav Bhaji.
Nice. Alright. Thank you so much.
