I would like to invite.

Matthew Putnam who was C.E.O.
of the co-founder of the Nano Tronics to

take stage and interview is this
company introduce the speakers and

then offer a couple of you
know lectures on the stopping.

One of the thank you so
much Dr Joseph and the and

Georgia Tech Institute of electronics for
for having us.

I can't tell you how happy I am that
this place and places like this exist so

that I've spent my life between industry
and academia I was a professor in material

science at Columbia University and
while we had amazing professors and

students what we didn't have was the
interaction between departments to be able

to make the type of advances that you
can can have here and to bring in.

You know industry and
really of other thinkers.

So this means a lot to me.

And thank you so much.

Tronics is you know I was involved
in a previous business and

then I became an academic had
a material science lab so

I am a materials scientist
more than a microscopist So

even though we're going to talk a lot
about imaging of a much more in your world

in a way you know I studied nanomaterials
I taught a course in reality and

then a materials and in semiconductors
really compounds semiconductors so

I really am imbedded in your
world the I think the thing

I realized in academia is this what would
normally just be considered a company

promotion line is really true to build
a future you need to see it and that's

literal and of course you know trying
to be visionary and see the future but

it is impossible to do these types of
technologies that involve nano scale.

And my kind.

Unless you can image them and

then scale them which is something that
doesn't happen in a normal university lab.

So we did I was hoping that our company
could bring this to industry and we have

done this in a number of industries very
quickly on the company you're not here

to hear about business things but
in case you have questions I think that.

A philosophy of ours is in some
ways similar philosophy to

this institute which is not to just
focus on people like ourselves.

Figure doing what we do we bring
in advisors from many different

areas to help so
we you know we have a lot of computer

programmers a lot of our solutions are
computationally based so all the better to

have the founder of Skype who is excellent
at working with computer programmers

Eric Drexler who you know invented
the word nanotechnology should inspiration

to me with the book Engines of creation
that he wrote in the late eighty's.

Lucky to have him as a consultant
an advisor to the company but

we even have to G.E. vice chair.

Of venture capital as founder of paper and
this is Peter Thiel as our board member

other than my father who's the other board
member who started the company with.

Sixty plus customers in a range of
industries However semiconductors compound

semiconductors new advanced materials
in general are our main customers.

You know over seventy five instruments
in factories we've raised some money

not that important to you and
we have patents.

Were based out of the three.

Nano trans is a collar cells are based
in New York we do manufacturing in

California which we make started
out as making automation and

robotics Now we do vertical
manufacturing for making these to do.

To integrating our
computational systems and

as I go into this I'll tell you
why we think that's important.

To even though we think of ourselves as
providing unique computational solutions

having a full package actually becomes
key to scaling a factory taking

what everything all the work that's done
in places like this institute bringing

them to a factory we can be put
the prototype for that factory and

provide a very simple tool that we
build ourselves into some customization

started out in Ohio where we have
service technicians and some IT admin.

So after I get to seven years of being

in business finally came up with
a sentence to describe what we are and

even that isn't that good for
us Silicon Valley back company like us.

It's what I've always been told that
it's important to have an elevator pitch

it's hard to pitch something like this in
an elevator this is as close as we get

combining optical microscope be even
that's not completely true we also do

atomic force microscope he was talking
about a little bit but optical microscope

be computational super resolution which
is really how the company started but

it's still a work in progress for us and
I will talk by different techniques that

others do that we do that we
aspire to where the world exists

in super resolution microscopy
artificial intelligence this.

TIME You know two thousand and
seventeen is the luckiest time for

I think you know the creation
of nanotechnology.

In the world and it has to do with
the convergence of material science and

artificial intelligence it wouldn't
matter how smart anybody was it wouldn't

it wouldn't matter if these
things hadn't converged and I.

About that convergence and
then robotics not not as crucial but

you'll see why it plays into it so
to make the most advanced microscope

I guess you can be the judge of that but
but you know why why we think it's.

So this is the only graph
that I have goes back

to my days as a professor and
what the goal

of the company became so I worked in
I was lucky enough to work in an I

think incredible materials science lab it
was this lab in the basement where the.

Old lab where the laser was invented and
I felt so

lucky we had a lot of equipment
maybe not maybe not as.

Certainly not as much as
this doctor dose of the labs

However we you know I was a polymer
reality just I put nano of my

background is putting you know the latest
of nano fellers in to truly complex.

Polymer systems so we did everything from

real logical measurements to a lot
of different types of microscopy.

So you would find something
optically then you would take it and

move it to an electron microscope maybe or
in A and

by the time you've done that you've
developed a one small device so

I mean the thing I remember the most is we

became incredibly excited
when you know what a Ph D.

students can we we finally done it we've
developed a full of all which is a carbon

nanotube that acts as an eye and

I are Nano in ten so
it can work day and night.

We are excited.

Until we realized we had to move up to
eighty ninety percent efficiency for

photovoltaic something that
everybody says is not possible

well they're kind of right and kind of
wrong the physics exists we had it but

you couldn't scale it so
it doesn't matter so

nano Tronics was an Al growth of trying
to scale not that technology but

any technology that was not currently
scalable so the idea here is look

from now on a meter and even suddenly on
a meter through looking at meters and

looking at smaller features over that area
if you could do that put it in production

and we'll talk a lot about process control
as being key to this and why artificial

intelligence allows process control in
different ways but this is the goal and

what I found at Columbia was this
was something that it wasn't my

lab that had this issue the bioengineering
lab on the floor above me had

this the chemistry labs had this
everybody had this issue so.

This was something that spurred
starting a business to address it

then so around two thousand and ten.

Now even though this wasn't the first
instrument we developed I told you

a little bit about the background on the
types of materials that I was interested

in at the time or had a background in and
they were three dimensional.

A lot of and they were a mix
of organic and inorganic and

you know it was complex and
doing three dimensional measurements.

Are not generally fast.

So to have a rapid three dimensional
measurement tool was important to

us this was looking at
at fillers that were

covered by a polymer other types
of materials so this was the first

inspiration of a product of a product
it didn't become the first customers.

Of our product.

But one thing you'll notice about nano
tonics is that we like to be like this

institute and look outside of anything
narrow that we're working on and we're

lucky that I think that we're not making
a single device or anything like this

we get this nice picture of what industry
in general is doing and one thing that.

We're noticing now is that three
dimensional measurements apply even

to compound semiconductors semiconductor
next generation electronics and

flexible devices so
we use also technology to do this

that comes from something entirely
different we use a method for

obtaining three dimensional images that
come from the book the film industry for

creating C.G.I. figures from astronomy and

please ask questions about that as
we like to go through but you'll

feel free to ask how we do this because
this is this is kind of a fun thing for

me so that's how we started then
something else was occurring.

Now I thought of this huge gulf in time
where things were going on underneath

the surface but that the public didn't
know about those two things that had that

were existing Were we talked about
Eric Drexler and also the nanotechnology

initiative that was you know two thousand
is that correct was signed into.

So you know Congress and this big amount
of funding this it all started and so

nano tech was a big work.

And exciting one then years went by
where people became skeptical because

we weren't seeing the kind of
almost Sayf I type advances

that were that were cracked only
theoretically and lab proven so.

So there was this golf at
the same time Marvin Minsky and

others in the field of advanced networks.

Had also made great advances starting
in the late one nine hundred fifty S.

through the seventy's in at MIT and
other places but

the idea of nanotechnology I mean
artificial intelligence actually

became you know almost a joke so something
called the AI winter with this time when

artificial intelligence wasn't used Well
both of those things now are realities

now they will be even more
realities as time goes on.

And other than just artificial
intelligence existing in your Amazon AT go

at home or in Syria or for Google search
engines how does that how does AI

in nanotechnology collide in order
to create physical things and those

physical things maybe have computational
power and do a lot of things but

change our our world in a massive way
that was sort of a promise before and

we think that both of these are important
so we're lucky to take advantage of that

to have the opportunity of advances
made in AI and I'll get into that.

So other than just technological
advances Let's go back

a little bit further in my life
I worked in a company that made

starting in the eighty's my father
started this company made factory.

A way to do data acquisition and S.P.C.

on the factory floor using
P.C.'s rather than using as.

Super you know mainframe computers and
things.

And so that goal was to automate in a way.

As we started nano Tronics really realized
that animation is not just a wait of

replacing jobs it's a way of replacing
a necessary task so things that.

Just cannot work due to the limitations
of the human brain no matter how smart we

think we are and this is actually
an image from a factory in China.

So in slave to manual tasks
is actually not that much

of an exaggeration I went to this silicium
work this facility who'd actually

are looking at automating but
thousands hundreds of thousands of

people working twelve hours
a day actually losing eyesight.

Having to stand for this entire time so

this this type of sweatshop labor
now looks like a clean room but

you know instead of you know a lot
of people in a basement sewing but

it's still a sweatshop in a sense and
it's not limited to China.

So a fact of the future that already
exists is more like this mobile

devices different ways to interact
with robots to train systems.

And this job then becomes
still take it's still useful.

In fact more useful because
you're getting the speed and

the power of artificial intelligence and

just calm computation in general that
is more powerful than a human alone but

along with humans can use human
creativity and the desire to discover

in order to make products that
are of better quality but

can iterate technologies faster and this
applies whether you're thinking of what we

consider the most advanced companies
like an Intel or something that follows

Moore's Law to a company that makes
tires that are trying to you know move

to nano fillers rather than carbon
black is a reinforcing filler so

it applies as I just was saying talking
about high tech and low tech it's all.

In the future and all can benefit
from this type of automation.

Very quickly on this as I
sure all of you know but

to get sort of a background on who we are
and how we get that first component and

how we're working towards in
the areas where we haven't.

Computationally doing something
did physics thought impossible.

We don't claim to break
any laws of physics or

to you know there's unlikely
to be a potman limit or

you know this is unlikely
that anything that John or

myself does is going to
rise to this level but.

There are several different ways of
defining diffraction limit of of light.

Abby is I think the most common one.

You know basic you just can't move out of
the visual electromagnetic spectrum and

be able to resolve something that is
smaller than that by using a normal white

light well so the first thing that
we did was come up with a method.

And that became a patented technology
only little bits of it which we actually

even use but I was inspired to
say that you can do things to

trick the out Abbi limit in a sense and
it is called Super resolution microscopy.

You know starting in
about two thousand and

six I started working on this
stuff not with great success but

was very early into this idea
that theoretically it works.

This is all you know taking
Abbie's equations you're you're

you're really tricking the denominator
of the of the diffraction

limit of the Abbey limit
the numerical aperture and

trying to computationally
change that from of.

You know physically defined number
by doing a number of things for

different things we do this for
the integration of light.

By movement of angles and
number of things and because we're

covering such a wide range of things today
this is the fun stuff to talk about so

going to talk to me about it there's also
interesting patterns works papers we've

braided with Cal Tech and can be a number
of places really interesting work

if this is something you're interested in
and doing any type of nano fabrication.

Or bio medicine or anything I think
it will be something interesting

because of the limitations presented by
other methods that are able to achieve

be limit resolution such as.

C.M.T. that are either destructive
tests or time consuming or

only look over very small areas.

So a few ways that people do this that
aren't we do you know fluorescents

microscopy bio samples just make
things light up differently so

you see it see them differently and
you can do some correlations to that

there's something called expansion
Mike Ross could be that King and

Professor Boyd and Boyd and mighty.

And actually this is something that
goes back much further the idea

of expansion my cost could
be this was used for

Neuroscience in sense of
you make things bigger.

So that it can be seen
under a microscope by

expanding them with a certain solvent and
then you using

fluorescence in order to be able to
recognise them in classify those things.

So one of our partners at Cal Tech.

I think talk and
talk very well about one of the methods

that we now have the intellectual property
to and have engineers working on.

Something called for
a polka graffiti microscopy.

Something that had been used for X.

ray diffraction and not been used in.

My craw Skippy in light my cross could be.

One sort of a word winning technology
that I can't take any credit for.

But this group and we've collaborated
with and now we're on the I P four I

think he says this well the challenge
of imaging slides at high resolution so

he won't really was working in the bio
space we work more in reflected light

if we're looking at semiconductors but
we both transmitted and

reflected light but he's thinking of
slides for bio same thing applied.

You know high resolution
wide range of view

in the fact that it requires a physically
perfect microscope it's free of aberration

this is them doesn't exist so F.P.M.
works by collecting a set

of low resolution images so
we assume that any image you take.

No matter how many pixels you have on
a sensor it's such a low resolution

compared to other methods that achieve
some of the Abbey limit resolution.

So you take the low resolution images.

And you sample this from different angles.

Tear a terrible image and I apologize for
this we can write this out if we want

the basic idea as we're collecting
information on the plane by iterating

light from you know one
degree to ninety degrees and

taking multiple images to zero
in on a feature that you're

looking at something not possible if you
can't take images almost immediately and

you know start a match you know really
map the image that you're taking.

And enter find that center really
elegant way of doing something.

That has.

Almost no cost associated with it
it's taking a regular microscope

iterating the angle of light and
this for a fee.

This was a BIOS sample.

You know try to ignore these labels
a little bit is slightly confusing but

if you look at the at this fuzzy
image this is taken with a two X.

objective.

You know this is just you know a Nikon or
an Olympus objective on a microscope that

optically not that different than
a microscope objectives that have existed

for the last four hundred years in
principle not that different and

use you get this fuzzy image due to the
Abbey limit and do due to this numerical

aperture even if even if you're not if
you take this you in a rate this will

you do for me and
you end up with a two using a two X.

objective you end up with this image

to the right of it which is
the equivalent of using a twenty X.

objective.

This is the thing that I
think is the most successful

that even if we get away from talking
about the limit which is achievable and

doable and and works well but the idea
of a lot of what we do it nano Tronics.

Is you know this scaling idea
requires looking over large areas and

not seeing small areas so
if you can do with a two X.

objective what would
normally require a twenty X.

objective identify features and it doesn't
require just wear it take out your feet

we do other things as well in
order to detect features and

then you have a type of throughput that
allows scaling that was never seen before.

So that's why it matters.

It's precise it's fast it's
not nondestructive of course

depending on what type of material looking
you could consider electron microscopy to

be destructive or not but if you're
taking a soft material microfilming

is not only enormously time consuming but
obviously destructive and

not something that is factory
worthy This can be done

it's just a light microscope it
can be done in any environment.

As you get to higher resolutions of course
you need you know some type of isolation

and you want you want to avoid
particles but it's not this is not

existing in a vacuum for instance and
it lends itself extremely well

to automation because of this
you can move samples through.

You can have robots move them
through just automated stages.

Absolutely don't try to read that.

There's something that everybody is
familiar with and some things that

nano Tronics I'm jumping around here by
the way you can kind of get your mind off.

It.

And for kind of a fee by the way I
mentioned because I think it's cool and

relevant and physically possible but
we could talk about a lot of other

technologies that address the inherent
problem in electronics but

this problem actually exists elsewhere
as well we think about the enormous

benefits of Moore's Law This increasing of
computational power due to the shrinking

of transistors in the lowering of
price but a new semiconductor fab.

So two problems with this the first we
address a little bit earlier is that some

day you reach a place where you cannot
no longer do lithography you can No you

longer use white no matter you
know how you try to trick it to.

Use the same process is.

This is which incredibly hard that Gordon
Moore who's also one of our investors and

advisor the founder of Intel by
the way who this was named after.

Basically the same thing that he
defined in one nine hundred seventy but

you know continuing to shrink and
shrink to now a seven and a metre nodes.

This is of great benefit of course we all
know the benefits that's why you have

a computer in your pocket and you probably
get to the point where you could have

an implanted computer still with
Moore's Law Eventually though it gets so

small that you have quantum tunneling that
becomes a problem you don't know how to

deal with a normal silicon semiconductor
process in order to defeat this so

if it that's just that's
the physics problem

there's also a problem that others
don't know it's called rocks law.

Rock being the first investor in Intel.

Who said and turns out as wisely but

more financially rigorous
idea that he realized that

as this happens something called
rocks law now also exists that goes

in opposition to this so the price to
make these even though it's cheaper for

us to acquire and and
use new transistors new I mean

new chips it costs more to make
factories so a semiconductor fab there's

a new semiconductor fab Global Foundries
fab that's being built right now in China

that's fourteen billion dollars the new
Samsung fab is fourteen billion dollars.

Obviously not something that we can all in
this room just start to try to create new

technologies and build them that's
a shame great innovation but

eventually makes even large companies not
sustainable if rocks LA continues along

with Moore's Law So Moore's Law could
come to an end will come to an end using

the same technologies unless we figure out
new technologies some of which I think to

Dr Joseph could even address some of the
additive methods that and you know three.

Three dimensional semiconductors
could address Moore's Law but

how we get how do we address
rocks at the same time so

to build it you have to see
it to make it you know to do

something smaller in a different way that
addressed scale ability a little bit.

Is I think how we deal with
that one way to deal with it.

When we put this side together.

We had you know this is computer
performance we just discussed gets

better and better artificial intelligence
comes in somewhere along the line

I actually think where we are here is
actually lower I'm not sure the arrow

is too high it's really still I would say
a little bit below humans in some things

a bit above humans and other things
sometimes and one of the things that John

will be speaking about later is where some
of the applications we get I think kind of

superhuman many still humans
are better but we have the human curve

going up slightly while their computer
performance was going up act exponentially

but truthfully as we get these
technologies human cognitive power.

It and this is good in a way because
it's not human creative power but

as we get more tools we learned
last week we actually if anything

perform slightly worse as
humans the more tools we get

if you're just talking about computational
performance we could always play the piano

better we can we can do a lot of things
better we can even think of new ideas and

new inventions better maybe but computers
are getting exponentially better and

we're getting at least
slightly worse my opinion.

So the idea of the convergence that we'd
like to have a lot of the things I've

talked about any of them we could go into
in more detail did talk really about

why and we will why some of these you know
robotics we know are improving a lot due

to computation so there's this is if you
kind of consider this a bit did this.

Graph a bit of a Venn diagram in
a way there's a lot of overlap

you can do about it because you have
computation because you have AI and

AI you can do better because you have
robotics and so on and you do this all

better because the improvements of
things like So we use that are you know

have better performance for running deep
learning networks for instance than.

They were even possible using millions and

millions of dollars supercomputers
even even recently so

it's exciting stuff that allowed then
machine learning to happen and allow for

fast computation for doing things such
as such as achieving super resolution.

And for what we do for
what we make that sentence of the micro

optical microscope the robotics automation
so on it is completely agnostic to

industry and John talk about applications
that are not I talk a lot about

semiconductors here because that's this
critical thing of reaching Moore's Law but

it applies to everything you know an AI
system a imaging system doesn't know

what it's looking at so you know we work
and the you know in all of these areas and

we think there's overlap so
with this one tool we we

can do things in a lot of industries that
have generally been segmented by industry.

And you know industry equipment makers and

different industry expertise
now it's all the same to us.

To be sort of the industry agnostic
with the same tool I think kind of nice.

To forget the specifications here but
the tool is the size of this and

the tool involves moving things
taking in this case away for doing.

Action nano inspection and the idea
that it is the size of a table top and

can feed in from production tools
changes the way the whole thing

a whole factory works and we'll talk about
this but this is really about specific.

Nations there's specifications today
none of these on a normal computer with

a normal machine these special occasions
will be different several months from now

and they were very different
several months before so

taking advantage constantly of this.

This doesn't always happen by the way.

I mean this is not due to anything
brilliant about nano Tronics

it has some but there is something stuck
in industry some of the highest tech

companies in the world have interesting
hardware solutions and very expensive

hardware solutions for inspection but
because they aren't competition and

software based are using all operating
systems they're using old chipsets

in some cases they're even
transferring data manually using.

You know I hadn't we hadn't seen a floppy
disk for years until we went into one

of the most sophisticated semiconductor
fabs it's to replace an entire fab and

to replace these tools that they're
used to and set up recipes for

require a lot of legacy unless you're
iterating constantly by using the things

that we talked about earlier so
this should change to the point where when

we're putting the slide together we're
saying we want to have this legacy and

from it this is going to be legacy
information if it stays with you

you know do we want to put this
down even it will be different.

So this is an example of how a factory.

We're worked from our experience and

some of the things that we have changed so
you have manual inspectors which we've

talked about this before and
here that picture had thousands.

And this feedback.

System to go to production tools come back
to manual inspectors maybe go to other

things in or from others and other
types of large scale inspection tools.

But all of this is not so
even if we consider it in line

it is it is it is still kind of a old
way of looking at it that doesn't

produce very fast increases in yield or
fast ways to do interation But

this is the way still we have many of
the customers that we work with and

partners that we work with still have
a kind of factory system like this

now by using our tools
which we call inspect.

It's something like this so
we are a bay or you could also think

of it as being the center of a wheel
with a spokes that go out to.

You know to different production
machines that all feed into our tool so

you have a a leader that is
either training a system or

it can be fully automated
that is dedicated to a group

of production tools so
there's no humans involved but

there's also not a long line
of multimillion dollar.

Inspection tools that lead
to this rock law problem so

this is the way you know more
the way that we envision.

The way a factory works and
I'll get a lot more into this later.

Just to check the time
to make sure I'm OK.

And cut me off when everyone.

Yeah OK Five minutes is fine so we'll
talk about this a lot later mainly but

to stay in your mind anyway this
segment of process and the first graph

on a twenty four hour cycle with all types
of sensors no matter how many sensors

you put into place you're controlling only
one part of the process we want to have

a fully integrated process and we when we
want to make that happen you won't get

all that for me just saying that right now
but we but we'll talk about our philosophy

behind this and this is all of this is
a work in progress we have customers.

That are you know actually pay.

Customers that are doing parts of this but

just like I don't want that
slide that had specs on it

I'd rather have some of these ideas
be you know less specific but

something that we will have specs
on next time we see you and

then different specs become so this some
of these are ideas that we're working on.

So I'm not even quiet five minutes but
on that first part can I take any quick

questions and then we can move on to the
next stuff that may answer some of them

but I'd love to take some questions
is who we are and why we exist.

The role we fill crude
currently in factories.

Nothing.

OK.

OK.