Sakya Dasgupta: A Journey from Gaming To Enabling Embodied Intelligence

Summary

In this episode of Silicon Grape Vine, host Nittin Dad interviews Sakya Duppa, CEO and founder of Edge Cortex, a company focused on creating energy-efficient AI systems for edge devices. Sakya shares his journey from writing a Tetris game at 9 to his PhD in brain-inspired computing and his work at IBM. He discusses the inspiration behind Edge Cortex, the importance of power efficiency in AI, and the future of AI hardware. Sakya also reflects on his personal interests, the significance of taking risks, and the concept of embodied intelligence in AI systems.

Takeaways

• 😀 Sakya Dupta is the founder and CEO of Edge Cortex, a company focused on creating energy-efficient systems for AI applications.
• 🌟 Edge Cortex aims to bring powerful AI models like GPT-7 and multimodal models to edge devices, inspired by the human brain's efficiency.
• 🌐 The company is headquartered in Tokyo, Japan, and operates globally with a presence in the US, Singapore, and India.
• 🎓 Sakya's journey in AI began with an early interest in computer science, inspired by his mother, a computer science professor, leading him to write his first game at a young age.
• 🏫 Sakya pursued higher education in computer engineering and machine learning, with a focus on applying AI to real-world devices and systems.
• 🏢 His professional experience includes working at IBM Research, where he co-invented the concept of Dynamic Bayesian Networks, used in time-series processing.
• 💡 Sakya emphasizes the importance of taking risks and learning from failures, which has been a core part of his entrepreneurial journey.
• 🌱 He highlights the cultural mix and government support in Japan as beneficial for startups, reflecting a growing and diverse tech ecosystem.
• 🔍 Sakya predicts that AI hardware will evolve towards more adaptable and domain-specific architectures to meet the changing demands of AI models.
• 🧠 He is intrigued by the concept of embodied intelligence, suggesting that the physical environment and body can influence intelligence, a concept he aims to apply in creating intelligent systems.

Q & A

What is the primary focus of Sakya's company, Edge Cortex?

Edge Cortex focuses on building energy-efficient systems for running AI applications, especially recent models like multimodal, and bringing that power to edge devices.

How does Sakya's background in AI research influence Edge Cortex's approach to technology?

Sakya's background in AI research, particularly in brain-inspired computing and neuromorphic computing, influences Edge Cortex's approach by emphasizing power efficiency and the development of systems that mimic the human brain's intelligence within power constraints.

What inspired Sakya to start Edge Cortex?

Sakya's inspiration to start Edge Cortex came from his work at IBM Research, where he explored how to make algorithms and software efficient for standard hardware, and his experiences with power-constrained environments like robotics.

How does Sakya's early experience with programming influence his career?

Sakya's early experience with programming, such as writing a Tetris game at a young age, sparked his interest in computer science and laid the foundation for his journey into AI and entrepreneurship.

What is the significance of the Sakura in relation to Edge Cortex's hardware?

The Sakura, or cherry blossom, is significant to Edge Cortex as it represents the Japanese culture of renewed joy and inspiration, which is reflected in the company's hardware and work culture.

What is Sakya's perspective on the future of AI and high-performance computing?

Sakya believes that the industry is still scratching the surface when it comes to AI. He anticipates an evolution in Transformer models towards more brain-like efficiency and adaptability, with a focus on domain-specific reconfigurable architectures.

What advice does Sakya have for young entrepreneurs?

Sakya advises young entrepreneurs to learn to fail fast, take risks, and learn from those failures. He also suggests finding the right partners and investors to help in the growth of a company.

What is Sakya's view on the importance of embodied intelligence in AI systems?

Sakya views embodied intelligence as a significant aspect of creating intelligent systems, where the hardware environment plays a crucial role in intelligence, just as much as the software substrate.

How does Sakya's personal interest in watches reflect his approach to technology?

Sakya's interest in mechanical or electromechanical watches, which use silicon alongside mechanical systems, reflects his appreciation for the integration of technology with traditional craftsmanship, mirroring his approach to combining traditional hardware with advanced AI systems.

What hobbies does Sakya enjoy outside of his work in AI and technology?

Outside of his work, Sakya enjoys outdoor activities like hiking and running, which allows him to explore and stay active, reflecting a balance between his professional and personal life.

What is Sakya's opinion on the current state of AI models and their efficiency?

Sakya believes that while today's Transformer models are impressive, there is room for evolution towards greater efficiency, similar to the human brain, which is the ultimate goal for AI systems.

Outlines

00:00

🌟 Introduction to Sakya Duppa and Edge Cortex

The video begins with host Nittin Dad introducing the guest, Sakya Duppa, a researcher in AI, entrepreneur, and CEO of Edge Cortex. Sakya's background in brain-inspired computing and his journey from his PhD at the Max Planck Institute to founding Edge Cortex are discussed. Edge Cortex focuses on creating energy-efficient systems for AI applications, particularly for edge devices, inspired by the power efficiency of the human brain. The company, headquartered in Tokyo, has a global presence with operations in the US, Singapore, and India.

05:02

🎮 Sakya's Early Inspirations and Career Path

Sakya shares his early inspirations, including his mother, a computer science professor, and his own interest in computer games, which led him to write a Tetris game at the age of 9. His academic journey from India to a Master's in Edinburgh and then to the Max Planck Institute is detailed. Sakya's professional experience includes time at Microsoft and IBM Research, where he explored machine learning and neural computation. His entrepreneurial spirit led him to found Edge Cortex in 2019, aiming to address the challenges of power efficiency in AI hardware.

10:03

🌐 Sakya's Move to Japan and the Founding of Edge Cortex

Sakya explains his move to Japan, initially for his PhD research and later for work at IBM Research, focusing on brain-inspired computing. He discusses the cultural and professional aspects of living in Japan, including the country's growing startup community and the revival of semiconductor interests. The decision to found Edge Cortex in Japan in 2019 was influenced by the global AI hardware trend and the opportunity to innovate in a market with a history of technological advancement.

15:03

🏆 Sakya's Achievements and Vision for Edge Cortex

Sakya reflects on his career achievements, including co-inventing Dynamic Bayesian Networks and their application in finance. He expresses pride in Edge Cortex's approach to hardware and software design, which prioritizes software efficiency before hardware architecture. Sakya's vision for the company involves creating adaptable, domain-specific architectures that can evolve with the changing AI landscape, aiming for brain-like efficiency in AI processing.

20:04

💡 Personal Life, Hobbies, and Advice to the Next Generation

In the final segment, Sakya talks about his personal life, including his love for watches and outdoor activities like hiking and running. He shares his views on Tokyo's vibrant culture and how it inspires his work. Sakya offers advice to the next generation, emphasizing the importance of taking risks, learning from failures, and finding the right partners for growth. He also discusses the concept of embodied intelligence and its relevance to creating intelligent systems, highlighting the importance of a holistic approach to AI development.

Keywords

💡AI Research

AI Research refers to the scientific study and development of algorithms and systems that enable computers to perform tasks normally requiring human intelligence. In the video, Sakya Dupta's background in AI research is highlighted as a foundational aspect of his work, leading to the creation of Edge Cortex, a company focused on energy-efficient AI systems.

💡Neuromorphic Computing

Neuromorphic Computing is an approach to computing that mimics the neural structure of the human brain. Sakya Dupta discusses this concept in the context of Edge Cortex's mission to create systems that are not only brain-inspired but also highly power-efficient, akin to the human brain's operation on a mere 15 to 20 watts.

💡Edge Cortex

Edge Cortex is a company founded by Sakya Dupta with the aim of developing energy-efficient systems for AI applications, particularly for edge devices. The company's focus is on leveraging AI research to create technology that can operate with minimal power consumption, reflecting the script's emphasis on innovation at the intersection of AI and hardware efficiency.

💡Brain-Inspired Computing

Brain-Inspired Computing is a concept that involves designing computing systems that draw inspiration from the structure and function of the human brain. Sakya Dupta's work at Edge Cortex is rooted in this concept, as he aims to develop AI systems that are not only powerful but also operate within strict power constraints, similar to the human brain.

💡Power Efficiency

Power Efficiency in the context of the video refers to the ability of a system to perform tasks using minimal energy. Sakya emphasizes the importance of power efficiency in AI systems, especially for edge devices, which aligns with Edge Cortex's goal to create systems that are both intelligent and energy-conscious.

💡Entrepreneurship

Entrepreneurship is the process of designing, launching, and running a new business, which starts as a startup. Sakya Dupta's journey from AI research to founding Edge Cortex exemplifies entrepreneurship, showcasing how his research background and innovative ideas led to the establishment of a company focused on AI hardware solutions.

💡Machine Learning

Machine Learning is a subset of AI that provides systems the ability to learn and improve from experience without being explicitly programmed. Sakya's early interest in games that applied AI algorithms and his subsequent work in AI research highlight the significance of machine learning in developing intelligent systems.

💡Hardware Substrate

Hardware Substrate refers to the physical components that form the foundation of a computing system. In the video, Sakya discusses the importance of creating efficient hardware substrates for AI, which is a core focus of Edge Cortex's work in developing systems that can support advanced AI applications with minimal power.

💡Deep Learning

Deep Learning is a branch of machine learning based on artificial neural networks with representation learning. Sakya mentions the challenges of running deep learning algorithms, which are compute-heavy and power-hungry, on edge devices, reflecting the video's theme of balancing AI capabilities with power efficiency.

💡Embodied Intelligence

Embodied Intelligence is the concept that the body and environment in which an intelligence operates play a significant role in its cognitive abilities. Sakya references a book that influenced his thinking on this topic, suggesting that the design of AI systems should consider not just the software but also the hardware and environmental context.

💡Multimodal Systems

Multimodal Systems are computing systems that can process and analyze data from multiple different types of input, such as vision, sound, and touch. Sakya discusses the evolution of AI models towards more brain-like efficiency and the potential for multimodal systems, indicating a future where AI can interact with the world in more sophisticated ways.

Highlights

Sakya Dupta, CEO of Edge Cortex, discusses the company's focus on creating energy-efficient AI systems for edge devices.

Edge Cortex aims to mimic the human brain's power efficiency, which operates on 15 to 20 watts.

The company was founded in 2019 and is headquartered in Tokyo, with a global presence.

Sakya's background in AI research at the Max Planck Institute in Germany influenced Edge Cortex's inception.

His early inspirations include his mother, a computer science professor, and his own experiences programming games.

Sakya's journey from India to Edinburgh for his Masters and then to the Max Planck Institute shaped his AI research.

He co-invented the concept of Dynamic Bayesian Networks during his time at IBM Research.

Edge Cortex's approach is to start with software design before creating hardware architecture.

Sakya is proud of Edge Cortex's achievements in a short period, particularly in software-first design and hardware systems.

He envisions AI models evolving towards greater efficiency, similar to the human brain.

Edge Cortex is working on adaptable domain-specific architectures for AI efficiency.

Sakya shares his personal interests, including a fondness for watches and programming.

He emphasizes the importance of taking risks and learning from failures in both entrepreneurship and life.

Sakya advises finding the right investors to support a company's growth and maintain control.

He discusses the concept of embodied intelligence and its relevance to AI systems and personal growth.

Edge Cortex is focused on creating intelligent systems that are power-efficient and sustainable.

Transcripts

[Music]

hello and welcome to this edition of

silicon grape vine my name is nittin Dad

and today my guest is sakya dupta now he

is uh you will find if you look at

Google search he's a is known as a

researcher in Ai and stist but he's

actually also an entrepreneur and a

founder and CEO of a company called Edge

cortex hello

sakya hi naan thanks

for good Saka let's um maybe just let

let's start about um uh what you're

doing right now Edge CTIC uh so it sort

of stems from your whole background in

AI research and and sort of bra yeah I

call it brain inspired Computing but I

think it's more more that sort of

neuromorphic computing just tell us uh

the of edge cortex and you how that came

about and and what what you're doing and

then we'll go into a little bit of your

journey sure absolutely very happy so

you know at at let me just start with

what you're doing a little bit first and

then I'll walk back time so our primary

focus is to build energy efficient power

efficient systems uh for running uh AI

application especially recent models

like J VII multimodal and bring that to

powerin devices like Edge Etc uh which

is not very different from the brain

given that the brain is extremely

constrain in terms of the power it's

around 15 to 20 wats but it uh clearly

even now we canot match the intelligence

of a typical human brain so uh you know

Ed um with that goal of power efficient

AI processing at the edge we started in

2019 uh so roughly around five years um

we are headquartered in Japan uh in

Tokyo uh where I'm speaking to you from

but we are pretty much a global company

we operate out of the US Singapore and

recently we kick started a new center in

India in Hyderabad okay to tell you

about how we you know kind of the

background of otex I think um I have to

go back a little bit in time um when I

was doing my PhD at one of the max

planks at uh in Germany uh the key area

of focus was uh how do you take

inspiration from the as you call it

brain inspired Computing uh both From

algorithm's perspective as well as kind

of the efficient Hardware substrate that

it has and be able to apply that to

systems and we were doing that on

robotic systems like walking machines

and when you're on a robotic system is

very constrain in power size area it's

not unlike a typical Edge environment

yes and lle that's of a focus now the

critical question was how do we make the

algorithms and software efficient

and make it work across standard

Hardware that was the first thing that I

was trying to solve and then much later

years later when I was at IBM research

uh I came to kind of explore same

similar kind of questions that again if

you scale up the algorithm problem you

run latest deep learning algorithms lot

more compute heavy power hungry in the

DAT how do you bring that the and so

that's kind of the I would say starting

point for H cortex be able to solve that

problem and hopefully someday we will be

able to achieve the level of efficiency

that the brain does okay well um I mean

we'll go into a little bit about um uh

your journey and I think what we we talk

about inspiration I think you started uh

actually you inspired by your mom your

mother uh who was a computer science

professor and you you you create wrote

your own game in I think in GW basic

something like Tetris at 9 years old

tell us about that Journey yeah sort of

what uh what made you do

that that is correct wow yes so I think

my mother definitely I've had a number

of Inspirations over the years but I

would say my mother was my first

inspiration in terms of she's not only

computer science and of you know I would

call her a computer scientist in that

sense um so early

on I was I grew up in calata in India

and was exposed to at that time if I

recall correctly MS Doss uh very early

days of computers uh at least in India

uh and games was a major part of so I

started getting exposed to different

types of Dos games which I found

extremely interesting given the amount

of limited amount of Graphics but you

could add a lot of intelligence so once

I learned my first programming language

I believe it was gwpc at that time uh I

in fact the first major project I ever

did and this is I'm maybe n or 10 years

old was writing a tet game uh at that

time if I remember correctly there was

also a Nintendo Game Boy that was there

which also had a similar Tetris so I was

trying to kind of reproduce that myself

and you know that really got me started

with computer science um and much later

I found myself doing Computer

Engineering and kind of pursuing that uh

in a journey uh in fact my uh foray into

AI or machine learning was through games

there were a lot of games that exposed

to that was applying uh AI algorithms

and uh IED that uh for quite a long time

until my masters uh trying to understand

the use of machine learning in games and

then how do we bring that to a real

world devices that we interact with yeah

and and I think that that's sort of took

you from doing your your B bachelor in

India to going to the Masters in

Edinburgh which is renowned for you that

kind of um work there as well uh so and

and and then the max plank Institute so

I think you you you you basically built

your whole uh sort of foundation for for

doing all this AI research that then you

carried on doing with

IBM that's exactly right in fact in fact

after I finished my uh Bachelors I

actually joined Microsoft for some time

and this was pre aor days as well uh but

what I was really attracted towards was

to do machine learning and at that time

at least I was not getting enough

opportunity so I decided that you know

let me go and pursue a master's and

during those days there were not of not

many universities offering AI Masters in

Ai and Edinburgh were one of the few

places okay so so what year was that

what year was that this is um

2006 2007 yeah that time so pre- de

learning wom alexnet was not yet

published um so when I was at Edinburgh

I was actually focusing on Gan processes

and Asian systems uh but then uh one of

the professors um kind of got me into

neural computation and one of my all

time you know talking about Inspirations

alongside my mother I would say one

nyman has always been quite an inspiring

figure for me uh and in fact one nyman

also in early days has talked about kind

of brain inspired Computing um and you

know how we need to be able to get there

with Hardware as well yes so somehow I

found myself posting a PhD in that area

so I was never so much a researcher as

much you though Google says that I would

say I've always been more of an engineer

at creating systems um and uh I think

the entrepreneurship side also always

existed um the idea was to implement

something and make a functional business

out of it and much you know years later

in 2019 I I think the timing was correct

there was a huge taal vent happening uh

in especially in US I would see in China

as well and many of the other parts of

the world Japan was a little slow yeah

but it was clear indicated that a

hardware for AI was going to become a

big thing it it is absolutely important

and I could relate with it in terms of

the challenges I had faced with robotic

systems and others in constrain

environments so we started 2019 the

company and um but I mean how did you

end up in Japan of all places I mean

that's also a interesting story in

itself that that itself is as well I

would say so I think I keep getting that

question you why Japan I think one of

the unique non-japanese who running a

company in Japan uh it was a mixture so

I think uh my PhD kind of brought me to

Japan through an organization called Ren

if you know them uh who have the fastest

superc computer in the world for a for a

time at Le fugaku and that's how I

actually moved to Japan where I was

again focusing on brain inspired

Computing and how do you build systems

out of it um and then you know I've have

always been being attracted to Japanese

culture so that's how I ended up here

but interestingly being in Japan working

at IBM research as well gave me an

exposure towards the demand for AI as

well as

semiconductors unfortunately in the last

10 to 20 years the focus on

semiconductors in Japan had slowed down

yeah uh but I felt that really an

opportunity in 2019 so we jumped in and

started the company and L and beh we

started 2019 I think 2021 tsmc announced

the kumam plant and 2022 we had

foundaries talking about 2 nanometers so

clearly there's a Revival of sorts of

semiconductors in Japan right and I mean

culturally I think it must be very

different to sort of what you had in

India and then Europe uh and and sort of

like all your counterparts are all in

the

US absolutely I think uh you know it's

it has been uh challenging as well as

extremely rewarding Japan is the sixth

country if you believe that I'm living

in now so I've kind of been a little a

Glo Trotter in that sense but Japan

really has an Eclectic mix of um kind of

today at least is strong government

support there is a growing uh startup

Community there is lot of influx of um

capital from investors across the world

globally so there are us investors

European investors investing in Japan uh

and uh you can have a Silicon Valley

Style company so at a cortex we pretty

much run a company where English is the

common language and we have people from

10 11 different countries so extremely

an etic mix and that's kind of

reflective of today's Japan I would say

okay and uh I just change a little bit

to sort of um what are the things that

you think you've sort of greatest

achievement so far I know it's you're

still young in your career but uh uh

what are what what are you really proud

of in in your

career I would say the first thing that

I was immensely proud of was the first

game that I wrote as we talked about

because that I I recall the amount of

joy that I felt from that but you know

years later uh of course in terms of my

own Journey um you know I one of the co

inventors of uh a concept called Dynamic

busman machines which has been heavily

used I invented that when I was at IBM

research and has been used in uh time

series processing at that time lstms

were the norm in the industry but we

were focusing on how do you make a low

latency significantly low latency and

low power perform its sequences uh which

has implications much later now to

wordss temporal programming with uh uh

other types of models latest lstm as

well as current generation Transformers

so I'm very proud of that and uh you

know we had a very successful

application that in business in finance

domain prior to starting at cortex that

what I was doing and was a very

successful Endeavor in its own right uh

and then coming from that uh in at H

cortex that was really the starting

point where I had worked for a long

period of time on software first design

of hardware systems and I I would say we

very uh immensely proud of the way we

have been able to kind of achieve that

in a relatively short period of time

alongside some of our peers in the AI

industry uh I think they established

Norm has always been Hardware first you

start with the process architecture

design and then software becomes an

afterthought you kind of go and solve

the compiler problem yeah me coming from

software background as well as many of a

team we started with that software

mindset first compiler became the

prominent aspect and then we built art

architecture out of that and so I think

that we take quite a lot of pride in

that aspect where we can now see we can

achieve really significantly better

performance per what compared to many

other uh competing uh let's say gel

purpose architectures

okay um and uh you know um the I mean

the big thing right now is AI and high

performance compute and and um what's

your view on sort of where where we're

heading uh I mean I think part of your

story is obviously everybody's story is

AI theh and uh small language models but

uh what's what's your perspective on all

of this being somebody who's a real

scholar in this area as well I don't

know if I can still call myself a

scholar in that not an act s more but I

think the way I see the you know from

the perspective of edge cortex and you

know where we are seeing the industri I

think we just still scratching the

surface when it comes to latest I think

today's Transformer models are great uh

but I truly believe that there's going

to be an evolution uh continuously on

those types of models and I think the

Endeavor is towards getting lot more

brain likee at least in terms of

efficiency that's the ultimate goal and

I think uh St the industri is making

Headway towards there where today's

Hardware that we are creating uh it will

be not completely

neuromorphic uh but even the

neuromorphic hardware that we have today

even driven is not completely suited in

my opinion to PR like efficiency so it's

a middle ground and we as a company also

we are steadily moving towards that and

innovating how do we make a lot more uh

malleable so to speak plastic that it

can adapt to the changes of the changing

AI landscape and I think that's what I

expect move to happen uh today's very

fixed architectures where once you bring

a chip to the market and the AI model

has changed it's going to be very

difficult to you know adapt and G

purpose Hardware by definition you know

no free lunches in this world yeah so by

definition is going to be much less

efficient so I think we will move

towards a little bit more adaptable uh

domain specific architectures uh with

the software first I think adaptable

domain specific reconfigurable I think

those are the areas that we're heading

towards I guess to make make that

efficient to a certain extent you know

we have tried the industri has tried

things like fpgs or field prr great Aras

which are completely reconfigurable but

then again you sacrifice you leave a lot

on the table uh what we have seen is

that you know data flow architectures um

in the industry many have called it cgas

who G reconfigurable area so there's a

middle ground uh and you know we as a

company we are also in a similar space

and there is a lot of innovations that

can happen there where you can get

towards brain like efficiency and when

you combine that with latest things that

are happening in the uh system scaling

side for example not just monolithic DS

looking at uh multi- diet

disaggregation in combination of system

Skilling efficiency and this kind of

architectural efficiency I think that's

where the industry needs to move to be

able to achieve lot better power

efficiency and we as a company you will

hear a lot more from us we are also

moving in that direction okay let's move

to a little bit about you your personal

side um uh what's what's a piece of

technology that you can't do without uh

what what do you use in your daily life

you think okay you really wouldn't uh be

able to function if you would or just

what your favorite piece of technology

you know it's very interesting um so

there are two things that I absolutely

cannot live with up the first thing is

actually not electronic at all I guess

it's partially electronic so I'm a I

love watches so and I'm kind of I cannot

live without a watch okay uh and the

best watch that I like to use is not a

smart watch but rather a mechanical or

electromechanical watch which uses

essentially silicon inside yeah

alongside the mechanical systems I

definitely cannot do without that and I

find that a really a Marvel look at all

the time yeah and the second is perhaps

um unfortunately the smartphone yeah

given the lives that we so those would

be the two things I could imagine uh and

then you know I still like program you

would find it hard to believe that I

still try to squeeze in time and write a

piece of quote on our Hardware whenever

I can

so the third technology would be just

standard programming languages okay any

any new games you written which which

people might want to

play it's been a while I think uh you

know there's a collection of different

games that you I've worked on for a

number of years but um yeah it hasn't

been the case unfortunately haven't

found the time to work on a game

recently myself but I think the closest

that we came is applying reinforcement

learning with our own Hardware um and

essentially um using simulated

environments for for example um typical

self-driving cars or autonomous driving

there's a data problem so that's

something that I've been quite enamored

by looking at using simulation to

generate a lot of data but then use

Hardware like we create to speed up that

or some part of that simulation uh which

RS reinforcement learning for example

can run on Hardware like C so that's

something that still exploring to a

certain extent but I would say not as

actively given my full-time uh

day-to-day responsibilities see year

okay and and what kind of uh Hobbies do

you have apart from programming anything

interesting you go I mean what do you do

in in

Tokyo to is fantastic if you're an

outdoors person uh and I do like to kind

of explore quite a bit so both my wife

and I uh we tend to kind of go out

hiking quite often uh you know so when I

was much more more younger than that I'm

now uh I was into bouldering uh and you

know climbing so uh I don't get that

much opportunity but you know we tend to

hike I would say or go for runs Tokyo is

fantastic for going for a run at night

and just soaking in wow okay well the

Tokyo seems to be like a city like

doesn't sleep here if you if you watch

the media over here in the in the west

so it seems to be very active all night

long it's it's it's a very eclectic um

city but I would say it has multiple

sheads to it it has a mixture of uh

complete modern as well as it preserves

uh you know the the history as well yeah

um and it's funny that you mentioned

that you know behind me you see this

Sakura yeah which is also the name or

inspiration of our Hardware okay and

that really stems from this Japanese

culture as well kind of this uh renewed

joy and inspiration uh which comes every

year and then kind of brings a new

outlook to life and Industry and that's

kind of the ethos for our own products

and work culture as well very good um

final question you so what would you

tell a young Su sua what you know now uh

in terms of Life advice or you sort of

how would you Mentor a young sua that's

such a tough question know I have a few

little kids and I constantly ponder that

question but I think the first thing

that I would see is learn to feel uh and

feel fast and quite often and I think

all the feel is little feel is that I've

had in my life has you know taught me

immense Les lessons so you know that's

in our company also we uh tend to take

you know another way of looking at

feeling fast is taking risks as

entrepreneurs we are always taking risk

and internally as a company culture as

well we tend to kind of think fast uh or

rather think big Implement fast and then

scale so you take risk and then you feel

you learn from that and you accelerate

and you grow so that's an EOS um I would

definitely share that with the Next

Generation if I had the opportunity

myself another thing that I would you

know caution people to be honest is you

know historically when we started the

company uh we were kind of bootstrapped

as well as um we never took funding from

any Venture capitals for a very long

period of time in fact up until first

few $5 million or so that we rais was

completely friends and family ourselves

Angel Investors uh that gave us a lot of

uh ability to control the way we wanted

to grow the company but also it brings

challenges so if in hindsight I would

say um you know if you find the right

partner if you're starting a company and

you know we are very lucky to have got

greed investors today it's good to find

those kind of investors and work

together with them because that might

help the journey um so that's another

lesson that we have learned uh which has

made us stronger in a

matter and and in in terms of like the

learning side obviously those are work

and career but uh in terms of other you

know on on the learning and the life

life life

Journey you know life journey I think

I'm still young comparatively so there's

a lot more wisdom to come from you know

kind of learn um I would see the as I

said you know feeling in general and

then learning from that is definitely a

life lesson that I've had several years

you know another thing that I think

about intelligence which is very

interesting is there's a book that I

read when I was in college it's from rol

Fifer if I remember correctly how the

how the mind shap or how the body shapes

the mind uh which is very interesting

which talks about embodied intelligence

you know the aspect of uh um your uh the

hardware environment or the kind of the

body that you're in uh plays as much

role in intelligence as the software

substrate and that's kind of a life

lesson so okay can

think perspective of um you know how you

kind of kind of the company that you

keep the way you work has an impact of

the way you think and at the same time

you can bring that to the systems that

we engineer uh sometimes we have to look

at it from the perspective of how that

system is going to get implemented from

the system perspective and uh take a lot

more higher up look rather than the M

micro side of what the semiconductor

does interesting I mean there's a

CEO who's talking about embodied AI at

the moment who's in this autonomous

driving area but I think it's quite

interesting that embodied intelligence

uh I think is is is something that is

going to be happening more and

more absolutely see I I would like to

think that we are in the business of

creating or enabling intelligent systems

that's the whole idea the reason why uh

we are creating software and Hardware so

we want to make a lot more power

efficient and sustainable and green but

ultimate goal is that these power

efficient systems should be a lot more

intelligent than what we have today and

I think we constantly think about those

kind of challenges embod intelligence as

well as not just making everything

Vision Centric make it a lot more

multimodal uh and you know the great

thing is today with uh recent transform

models we are now at least scratching

the surface of bringing such multimodal

systems alive in a lot more efficient

well s thank you very

much thank you Nathan lovely questions I

must say it was a great um speaking with

you thank you

[Music]

[Music]

[Applause]

[Music]