Transcription
Welcome to First Time Founders. I'm Ed
Elson. Artificial intelligence has
become one of the most heavily funded
sectors in the world. More than 30
startups have raised over $100 million
this year alone. And as AI becomes more
embedded in how the world operates, a
handful of firms have emerged as the key
players behind that transformation.
Among them is a company building the
kind of AI most people don't see. That
is the AI that is powering the systems
that run businesses and governments.
Founded in 2019 by three former Google
engineers, this company has focused
squarely on the enterprise market,
developing large language models for
clients like Dell, SAP, and Salesforce.
It even recently signed a deal with
Canada's government to bring its
technology into public operations. Now
valued at nearly $7 billion, it has
earned a place alongside giants like
OpenAI and Anthropic, helping define
what the next era of AI will actually
look like. This is my conversation with
Nick Frost, co-founder of Coher.
All right, Nick Frost, good to have you
on the program.
>> Thanks for having me.
>> Uh, so for those who don't know uh what
Coher is, I think we should probably
just start there. What is Coher? What
does Coher do? what are you guys
building in AI?
>> So, we're a foundational model company
and we are uniquely and singularly
focused on the enterprise.
So, there's there's about 10 companies
in the world that can make foundational
models. So, foundational models the
large language models that are largely
these days synonymous with AI. If
somebody's talking about AI, they're
probably talking about large language
models. Uh there's about 10 companies in
the world that can make them. We are
unique amongst them in our singular
focus on the enterprise. So we make
large language models that are good at
the stuff that enterprises need them to
be good at. We make them uh easy to
deploy and efficient to deploy for our
enterprises. We deploy them securely and
privately so that we can't see the data
that our customers are passing into the
model that allows them to to access the
truly useful data out there. And we make
them easy to work with via an agentic
platform. So we do kind of the whole
thing in order to get AI to work at
work. So these foundational models, I
think most people who are interested in
tech kind of know what they are, but
just at a very basic level, the
foundational models are the models that
all of these AI startups are building
off of or all of these companies, if
you're building AI or you're building AI
products, you need the foundational
model which companies like Open AI and
Anthropic and Coher, your company are
building. Um, what am I missing?
>> If you talk about AI companies, there's
a lot of companies building stuff with
AI. Mostly when people say they're
building stuff with AI, they mean
they're building stuff with foundational
model. These these days there's still a
huge amount of work being done on more
traditional machine learning, smaller
systems. Um, and a lot of those people
working on that will still very
rightfully call so that they're an AI
company. But if you are talking to
someone and they say, "Hey, I've got a
startup and it's an AI startup or
something." Chances are they mean
they're building off of a foundational
model. They're making getting a large
language model to do something useful
for their customers. Uh there's a
relatively smaller number of companies
that actually make the foundational
models. So that actually make the large
language models that take in a bunch of
words and then predict the next words
that should come next. Um yeah, there's
about 10 of those. So there are 10ish
companies building foundational models
uh which is basically the backbone of AI
or at least it's one of I don't know the
vertebra of AI. Maybe we could call the
chips the backbone. But what is so
striking is there are thousands of AI
companies and we're we're we're seeing
many of them and so many of these
companies building AI and yet there are
only 10 companies that are building
these foundational models. Why is that?
>> So in short, it's really hard and it's
enormously resource inensive.
Building large language models is a lot
more like building a rocket than it is
like building other computer science
projects.
>> It requires a huge number of really
smart people who have experience doing
it, working in tight unison. there's a
whole bunch of different things that
need to go well in order for it to be
successful. There's a whole bunch of
experimentation that needs to get done
and there's still, you know, uh, and
there's huge amounts of resources that
need to get put into it in order to make
the thing work, right? So, you have to
get a huge amount of compute. Um, so
rent all those chips, you know, that
that you were talking about. You need to
get a huge amount of data. You need to
have a huge amount of people helping you
create that data, getting data
annotators. you need to have a whole
bunch of really smart engineers working
together in order to make it go well and
even then it it's still challenging. So
yeah, so there's really only about 10
companies in the world that are doing it
because of that reason. In the same way
that there's there's not that many
companies building rockets either,
right?
>> So how did you end up being one of the
people who built one of these rockets?
Take us back to the beginning. How are
you? How did you get into this?
>> Before co-founding cohhere with Aiden
and Aiden Gomez and Ivan Zang, um I was
uh a researcher at Google Brain. So I
worked with Jeff Hinton for a few years
there working on explanability and
adversarial examples and capsule
networks and stuff. Um which was really
fun. And it was there that I met Aiden.
And Aiden uh was just finishing up a
stint in Google Brain in California
where he worked on the paper attention
is all you need which introduced the
architecture that we still use today. So
that he helped write that paper in 2017
and you know almost 10 years later we're
still using the same the same
architecture. Uh so after he worked on
that and when I met him in Google Brain
Toronto he you know he was obviously
very excited about the architecture uh
and about what it could do and he showed
it to me and I thought it was also
really really exciting. So you know we
noticed something about the nature of
this new model that created an
opportunity and indeed a need for
companies to make foundational models.
What we noticed was that for the first
time in machine learning's history, if
you wanted to solve a task like a
language task, the best model to solve
that task was not a model trained on
that task alone. It was a model trained
on a whole bunch of tasks.
>> Mhm. So that was really exciting and
that made us realize, hey, like there
there's going to be a need if companies
are going to actually make this stuff
useful and get this to work for them.
There's going to be a need for companies
to create really big and really good
foundational models that other companies
can use. So we had that realization in
2020. Um, and we've been delivering on
that since then. We've been trying to
make language models useful for the
enterprise by making them really
effective at the things that they care
about. You mentioned uh Jeffrey Hinton
there who you who you studied under who
for those who don't know is considered
to be the godfather of AI. Um why is he
the godfather of AI? What did you learn
from him? And I mean I think people
generally recognize him and his name
maybe if you're into tech but perhaps
they don't um if you're not super
plugged into what's happening in AI. So
what was his role in the story of AI and
what did you learn from him? So I
studied with him as an undergrad. So I
only have an under I don't have a
masters or a PhD or anything. Um, so I
had an undergrad from uft and I did take
his course while I was there and I sat
in the front and asked lots of annoying
questions. Uh, and and I really only
worked with him closely when I was at
Google. So I was a research uh
researcher at Google brain and I worked
with him at for I was uh working out of
water for a little bit and then I found
out that he was working in Toronto and
then we started to work together and
then I helped start up the Toronto brain
group with him uh and worked there for a
few years and it's during those three
years four years that I learned most of
what I know about research and machine
learning and neural nets and I learned
it from him. So I learned a huge amount
from him, but I I'm not I don't have a
PhD or a masters or anything. As for his
contribution, I it really can't be
understated.
Neural nets
neural nets have been an idea for a
while. People have been thinking about a
neural net architecture and in
particular Jeff's been thinking about
neural net architectures since the like
you know mid80s.
There was a long time where people
thought they were not going to work
and there was this whole wave of you
know first perceptrons and that which is
just a single layer neural net um and
people thought that was kind of
interesting for a little bit and then
some work came out to show that they had
some fundamental flaws and that really
cooled people down on them. people
weren't excited about neural nets. Uh
and then people started working on
multi-layer neural nets or multi-layer
perceptrons and that solved some of the
the critiques but still people were not
excited about it and they generally
thought it was a bad idea and if they
wanted to build AI they were much better
doing things like search or or symbolic
reasoning or things like that. Uh and so
very few people worked on it and largely
they thought it was dumb except for a
few people Jeff being one of them. So
Jeff tirelessly worked on neural nets in
the face of general ridicule for decades
for decades until around 2011 2012 they
were finally able to show that neural
nets were suddenly the best at image
recognition. That was the first thing
that they really they really knocked out
of the park on. Um and that was done at
uft with a bunch of other brilliant uft
students. the reason we are where we are
with neural nets in general which of
course is the precursor to transformers
right so there's kind of if you think of
it broadly there's like AI as a concept
there's machine learning as one strategy
for doing that neural nets as one
strategy for machine learning
transformers as one type of neural net
it's kind of where we are so the neural
net part in particular uh Jeff can claim
a huge amount of responsibility for and
it's really his tenacity uh that that
and his dedication to continuing to work
on it even when everybody else around
him was saying no this is a bad idea
it's not going to work um that we have
to thank for where we are today
>> so when we look through the when the
history books are written about AI
um I mean AI is having its moment right
now what changed I mean AI had been
worked on and neural nets had people had
been working on this stuff for decades
Jeff Hinton had been working on it for
decades he makes this breakthrough with
image recognition in the early 2010s now
It's ubiquitous. Was chat GPT the the
breakthrough moment? Like what will the
textbooks tell us about what changed
when AI became mainstream?
>> There have been other AI moments. There
have been other times when people are
when the whole world's really thinking
about AI. This is the first time that it
I would say it's been this dominant
narrative
of the economy for the past few years.
And that's a first like and techn that
tech it's been the dominant narrative of
technology for the past few years and
it's been the dominant narrative of the
economy even more for the past few
years. So that's that's kind of a first.
But there have been moments where people
have been as really excited about AI and
thinking that they're in some kind of AI
moment before. You you got to separate
AI as a property versus any
implementation trying to get at that
property.
So people have been thinking about
artificial intelligence like what
happens if a machine has intelligence
the way a person has intelligence for a
really long time. There's a myth that I
cite pretty often that was written in
the like a the around you know 1500s
1400s I believe um a Yiddish myth about
the the Gollum which talks about you
know some rabbi imbuing intelligence
into a clay man and then he asks the he
asks the the Gollum to go get fish from
the river and then he leaves his house
for a little bit and when he comes back
the house is filled with fish and the
river is empty and like like it's a joke
right like it's a it's it's effectively
you know, a comedic story that's told at
that moment. And the joke is, oh, like
intelligence is complicated and there's
nuance in language and if we gave an
artificial thing language, maybe you
wouldn't understand that nuance. That's
like a 500-year-old joke.
>> Yes.
>> Right. So, people have been thinking
about this for a really long time. More
recently, you know, after the computer
was invented, you know, there was a
whole wave of people thinking about
that. Now, Alan Turing was thinking
about the Turing test, thinking about
intelligence. After that there was
search. There was the deep blue moment
when when uh search algorithms beat
Kasparov at chess and that had a similar
moment. So people have been thinking
about this all the time.
This is different. This is a different
moment and it's different in its scale.
And when people write the history of AI,
this is certainly going to be a pivotal
moment. And I'm convinced that neural
nets are certainly going to be a central
component of of machine learning and AI
going forward. like they're so good,
they're so fantastic at they do all
kinds of things that there's no other
way we could get them to do yet. Um, and
transformers in particular, large
language models are very easy to use for
the average person. And that is I think
really
why this feels different. So if you look
at the other moments when people were
talking about AI like deep blue, let's
look at that one as an example, right?
Like there you can read tons of articles
about people talking about what's
happening with the machines or are
computers getting as smart as people.
They beat the best chess player in the
world. Like what's going on? But if
you're an average person, you couldn't
really interact with that. Like maybe if
you're good at chess, you could try the
chess bots and that and people did. And
actually, you know, chess in some ways
chess is more popular than it's ever
been before. And in part that's because
you can be at your home playing against
something better than a grandmaster.
But you could interact with it that way.
you you couldn't really interact with a
search algorithm like an AAR search
algorithm in anything else. So your
experience of it is pretty limited. Same
with machine learning. Like when we made
image recognition, the best image
recognition model, suddenly, yeah, your
phone, you could go on Google Photos and
you could search up pictures of, you
know, dogs and see all the pictures of
dogs you've seen over the years. Like
that's new. That's cool. But you
couldn't that's still directive. That's
still like somebody made the model that
does the thing. It it's telling you how
to use it. Transformers are the first
time that any person without any
experience in computer science or AI can
go up to the model, you know, open up a
chat window, ask it to do something and
it'll do it or it will not do it and
that'll be interesting itself. But you
can interact with it without it being
prescriptive of how you interact. And
that's I think the the reason why this
is suddenly so much bigger. it's
suddenly so much more interesting, so
much more widespread, and why it's
become the dominant narrative of tech
over the past few years. So when people
write the history of AI, and I want to
be clear that I think the history of AI
is not done, I think, yeah, I don't
think transformers are going to get us
to artificial general intelligence. So I
think there's going to be more waves of
new independent spontaneous inventions.
I'm sure that's going to happen. But I'm
convinced that the transformer is going
to be a central component of that. And
when the history of AI is written a 100
years from now, a thousand years from
now, this moment will be talked about as
relevant and interesting and a moment
when a lot of stuff happened really
quickly as a result of the tenacity of a
handful of people. Yeah, it's
interesting that in a way it was the
consumerization
that really took things in a in a
completely different direction. Um,
which is almost a testament not
necessarily to the underlying
technology, but almost to like the
productization and being able to put
this kinds of technology into the hands
of millions and then eventually hundreds
of millions of people. Um, is that when
you see all of these big tech companies
that are spending hundreds of billions
of dollars building out their AI
capabilities, building out data centers,
renting compute, buying chips, uh, and
then spending money on on on models like
like the ones you've built to build
their own products? Do you think it was
sort of a moment where they kind of woke
up to what the capabilities and what the
prospects of AI could be because they
just saw it a lot or was it something
else? Was it that the technology changed
in a fundamental way? I mean to to what
extent was this sort of the narrative
that suddenly captured people's
imaginations versus something in the
technology actually changed which made
Mark Zuckerberg think now we need to get
on this. I think everything we're
experiencing today is largely
predictable from around 2020 2019.
Now that's not a coincidence. That's
when I left Google to start coherent
with uh Aiden and Ida. So that's you
know the reason why I think it was
largely predictable around that time is
because that's when I predicted it. So
I'm sure other people predicted it
before but I got on board at that time.
Um at the time I think when I remember
telling people I'm going to leave to go
create this foundational model. I don't
think we used the word foundational
model. We just said large language model
company. We're going to be a large
language model company. We're going to
make large language models. I remember
everybody saying yeah that's probably a
good idea. I don't think anybody was
thinking like, oh, that that makes no
sense. The question was not like the
question was like, oh, you know, is
Google is Google just going to do it?
Are the other are the other big
companies just going to do it? Um, but I
think at the time it made sense. Now, it
it really still wasn't popular. And when
we had conversations for the first few
years of coheres history, the
conversation was this is a large
language model and here's why we think
it can help you. The conversation now is
okay cool like why you know why your
large language model or like how can
this actually help me get into
production? How can I have it access my
my private data without giving that
away? Like how can I how can I deploy it
in a secure and safe way so that I can
handle regulated industries? Um how can
I connect it to my specific data in an
enterprise? Like those are all the the
questions we answer now. So it's changed
a lot and what changed in particular and
a thing I did not predict at the time
was the success of chat fine-tuning. So
you train this big generic language
model. Uh when language models were
first created what they did was they
just completed the ends of sentences
because they were trained on the web. So
you really can think of it as like a a
web they were calling it a large
language model but at the time it wasn't
a large language model. It was a web
text model.
>> Yeah. It's like a Reddit language model.
Yeah.
>> Yeah. So, if you wrote the first part of
a website, it would write the second
part of a website. Not even the HTML,
just the text on the website. Um, and
you could do a lot of stuff with that,
but it was confusing and weird. And then
OpenAI and a few other companies at the
same time fine-tuned that large language
model on chat dialogue.
And that suddenly suddenly people
understood it. And at the time actually
I remember thinking I was surprised at
how efficient that was because when you
think about it you're training a model
on the entirety of the web. So a huge
amount of language and then you
fine-tune it on a relatively small
amount of chat and yet actually it
learns how to chat pretty well.
>> Yeah.
>> So that is I think responsible for for
the difference between 2020 and 2022. It
was the data efficiency of chat
fine-tuning that allowed people to to
like for the model to meet them where
they're at. They kind of expected users
kind of expected chat to work when you
told them it was a large language model
and it it didn't. It was this like weird
text thing. So then making it work in
the way they expected it to work seems
to have really gotten like woken people
up to the effect and the the utility of
these models.
>> Yeah. And I'm sure it was also the
volume too. The idea that if you keep on
chatting with this AI, you're you're
contributing more and more data for it
for it to train itself on. Um, I I want
to get to the specifics of cohhere in a
moment, but you know, it's it's
interesting. You're describing
that the model gets better when it's
subjected to or when it's fed large
amounts of data and also like diverse
forms of data. And originally, we were
kind of just limited to the web. But the
web isn't all of life. There's more
beyond the web uh that that these models
could be trained on. And so too, you
could say the same thing about these
these chats. I'm wondering if there's
other forms of data that you think will
be prevalent for uh model training in
the future. Um you know,
things in in the physical world. I mean
I mean typing words onto a onto a
keyboard and seeing words on a screen
isn't everything. But to AI right now it
seems to be close to everything. So is
there a way are there other forms of
data that you think in the future AI
will be fed and therefore that would
sort of uh take us I guess on the path
to AGI.
>> Let me first talk a little bit about the
way the way we train these models.
>> Okay.
>> So the first step is to train them on
everything on the web everything on the
open web. So you have you create a a
data set of all the text um that's
available for training from the web and
that turns out to be a huge amount of
text orders of magnitude more text than
you will ever read. I I like like a
thousand people a thousand years reading
24 hours a day volumes of text like
that's how you know that's how much
text. Um so first step is train on that.
Then you make a data set with people. So
you have people create like talk to the
model and if the model gives a good
response they say that's great. If it
gives a bad response they say that's bad
and they write what the model should
have said. If you do that process you'll
create both uh ratings like is it a good
response or a bad response and you'll
also create what's called supervised
fine-tuning data SFT data. So that's
like here's the input to the model and
here's a gold standard of what a person
wanted like they wrote out the sentence
like that's what the model should have
said. That's called SFT data. So then
you train the model on that SFT data.
After that you can do reinforcement
learning which was a type of machine
learning invented before transformers
where the you're training a model
without access to the to the right
answer. The model kind of tries stuff
and then you say hey this was better or
this was worse and you and you update
the weights of the model based on that
that signal. So then you can do
reinforcement learning. Now we do a
whole bunch of reinforcement learning
with synthetic data. So now we use the
model itself to generate data and then
do reinforcement learning on that
synthetic data. So that's a big
component of training now. So there's
like the data you get from the web, the
data you make with people, and then the
data you make with the model itself. And
those all of those are super relevant
for making the models that people use
today. Your question about models being
restricted to the to the web um and
missing the stuff in the real world. Is
that a blocker to AGI? like, "Yeah,
definitely that's a blocker to AGI." If
when you say AGI, you mean humanlike
intelligence. Yes, that's a blocker to
AGI. We are embodied creatures. We have
we learn our intelligence through
interactions with the real world and
intervention into the real world.
There's lots of interesting
psychological work that suggests
learning and interaction are super
related. So, interaction is super
important. Um, is that a blocker to AGI?
Like, yeah, definitely. But I don't
there's a whole bunch of blockers to AGI
and that's just that's just one of them.
>> Um and and the technology as as it
exists today is massively impactful,
massively useful, absolutely
transformative on the nature of
computers and subsequently the nature of
work, massively transformative on the
economy in general, then I don't think
it's AGI and nor do I think the
transformer alone will get us to AGI.
Nor do I care. I don't really I don't
really look out in the world and say,
"Oh, gez, I wish my computer was a
person,
>> right?
>> I look out in the world and I say, "Oh
man, there's so much stuff that a a
computer should be doing and not me. My
time should be free to to to think
strategically, to think creatively. Um
there's so much work that a large
language model when connected into the
things that I am using can do for me and
and and subsequently allow me to do the
interesting in the human and like that's
what that's what I want to make. I want
to make a technology that does that as
good as possible." Do you think that AI
the the people building AI the the
leaders of the AI industry Sam Alman
probably being the the high priest right
now at least do you think that there's
not enough appreciation of that? Do you
think that people are too obsessed with
we need AGI, we need humanlike
intelligence? I just look at the
contract between Microsoft and OpenAI
which basically one of the stipulations
in the contract is you know the terms
will change once we achieve AGI. I mean
there are many questions like what does
that even mean? But the fact that AGI is
sort of the benchmark for everyone and
I'm even asking you like how do we get
to it? Do you think there's too much
obsession with this concept of AGI in
the AI industry right now? Yeah.
Yeah. I mean, you said, yeah, look, high
priest is a good term. A lot of the
thought around AGI and discussion around
AGI feels religious to me. It's calmed
down a little bit, right? Like if we
back in 2023, 20 like 2024,
my views on this were a little
heretical. People would disagree. If I
said, "Hey, AGI's probably not around
the corner." People would disagree and
say, "Why do you think that?" Like I I
would get a lot of push back. I don't
get much push back these days. I'm like,
"Yeah, guys, guys, guys, we know
Transformers. Incredible. Super awesome.
Super good. Can definitely be way better
than they are. Need to be deployed
correctly. Need to have lots of stuff,
you know, to get them into production.
That's what we focus on, but
AGI like no." And everybody's like,
"Yeah, yeah, yeah, totally. I get it."
And and if you use a large language
model, which as everybody does these
days, you'll feel that pretty soon.
You'll be like, "Yeah, they're amazing
at these things." And then I ask them
some other things and they don't
understand at all. completely different
and they have a completely different
it's very different talking to a
language model as it is chatting to a
person and people you know kind of know
that when they're grounded in an
environment the focus on it is I think
you know a narrative device more so than
it is a scientific belief
we'll be right back
for the show comes from LinkedIn in.
It's a shame when the best B2B marketing
gets wasted on the wrong audience. Like,
imagine running an ad for cataract
surgery on Saturday morning cartoons or
running a promo for this show on a video
about Roblox or something. No offense to
our Gen Alpha listeners, but that would
be a waste of anyone's ad budget. So,
when you want to reach the right
professionals, you can use LinkedIn ads.
LinkedIn has grown to a network of over
1 billion professionals and 130 million
decision makers according to their data.
That's where it stands apart from other
ad buys. You can target your buyers by
job title, industry, company role,
seniority, skills, company revenue. All
so you can stop wasting budget on the
wrong audience. That's why LinkedIn ads
boasts one of the highest B2B return on
ad spend of all online ad networks.
Seriously, all of them. Spend $250 on
your first campaign on LinkedIn ads and
get a free $250 credit for the next one.
Just go to linkedin.com/scott.
That's linkedin.com/scott.
Terms and conditions apply.
We're back with first time founders. I'm
going to ask you the question that you
said everyone asks about Coher, which is
um what is the difference between Coher
versus the other foundation model
companies? What is the difference
between Coher and OpenAI? Between Coher
and Anthropic, those are the two big
ones in my head. Uh what is the
difference? The big difference is we are
not a consumer company and we're only an
enterprise company. So we don't have we
can't pay $20 a month to get access to
our tech. We're not trying to build a a
a product that people use in their
personal lives. We are instead selling
only to large and medium enterprise
companies and we create language models
and search models and an agentic
framework for using them. that is
tailored to the needs of those
companies.
That strategic difference comes from a
philosophical difference which is a
different view on the technology, right?
Like I don't think the technology is
going to get us to hi and I don't think
the biggest utility of the models is in
people's personal lives. I think the
biggest utility of these models is in
work. I think that like their ability to
augment and automate work at a desk
behind a computer is I think what they
are the best at. And so we have that
different view of the technology that
leads us to think differently about
where we can add the most value to the
world and that leads us to being an
enterprise a singularly enterprise
focused company and that is as as
mentioned unique amongst the
foundational model players.
>> Just so we can picture like what kinds
of work is being done. What is an
example of a use case that an enterprise
is is adopting um because of using
cohhere as a as a foundation model? Lots
of people will go into work, open up
North, which is the name of our agentic
platform. So, it's like a it's like a
chat app with automations and you can
make custom agents and you can share
those across people like but it's a it's
a chat app um that on its surface you
would be familiar with. So, they'll go
into work, they'll open that up and they
might open up our model and say, you
know, hey, you know, somebody emailed me
yesterday uh about a brief for a
meeting. uh read that email then cross
reference that with our Salesforce data
and then make a table sharing like
telling me the the state of that
customer right that's something that the
model can do for it or they might say
hey you know I just got this data room
from this company I'm trying to evaluate
read through the data room do some
analysis come up with a cited and
detailed document on how you think that
company looks and then send a Slack
message to my co-workers with that PDF
>> just looking at where you are in the AI
world. You are automating
uh tasks that are done at businesses and
enterprises
uh that as we are all talking about
would otherwise be done by humans which
introduces the question of is AI going
to replace people and this has been a
large debate. We're obviously seeing a
lot of layoffs in tech right now. Um a
very charged debate.
How do you think about all of this?
>> How do I think about it? Frequently.
>> Uh yeah, so there's a lot. So I think
this technology is, you know, there is a
huge amount of stuff that people do that
large language models should be doing
for them. Large language models will do
a better job of them. The work itself is
not very enjoyable. Humans are really
good at a lot of stuff that large
language models are very bad at. And
largely they enjoy the stuff that
they're good at and don't enjoy the
stuff that large language models are
good at. So I think, you know, in the
same way that we've had previous
industrial revolutions that augmented
and automated a huge amount of stuff
that people generally didn't really like
doing. And we look back on those periods
of time as kind of chaotic, but largely
a good idea, right? I don't no one's
running around saying, "Hey, the steam
engine was a step in the wrong
direction." Or, "Hey, the industrial
revolution was bad. We, you know, we
should all still be farmers." I think
there's something similar going on with
this. Now, I do think this technology is
fundamentally augmentative, right? I
think this technology, anybody working
behind a computer, I think this
technology can automate, I don't know,
20 or 20, 30% of their work.
I don't think it can automate 100% of
pretty much anybody's
>> huge amounts of the work that we do is
not just text on a computer or images on
a computer. It's it's personal. It's
understanding the cultural context. It's
talking to people and coordinating and
aligning. It's thinking strategically.
It's doing all of this stuff. And that's
true at every level of an organization.
So, I think it's there's a lot of people
say, "Oh, this is just going to take out
the bottom bit of an organization."
Like, no. What this is going to do is
make augment and improve and increase
efficiency and productivity across the
entire organization. Is that going to
have consequences on the labor market?
Yes, absolutely it is. Just as the
industrial revolution had huge
consequences on the labor market, just
as you know the widespread adoption of
computers had huge influences on the
labor market. Like in our lifetime, your
in my lifetime, we have seen wild
changes in the way that work is done as
a result of technology. It was not so
long ago that every organization had a
huge number of people working as typists
to type stuff up because mo because
people didn't have computers and that
was that needed to get done. that like
that that doesn't exist anymore. But the
labor market evolved, the labor market
figured it out. All those people are,
you know, still were doing good work. Um
just doing different work. So I think
that this will have similar effects to
the computer, to the internet, to the
industrial revolution, on the labor
market. And I think governments and
organizations and unions and businesses
should be thinking about how to make
sure that that goes well. Mhm.
>> How to make sure that that is largely
that that uplifts people and that builds
a resilient economy and that allows
people to do things they like to do. And
I really like that's the conversation
I'm encouraging everybody to have. Like
what are the what are the policy
decisions that can be made in order to
make sure that that is good for all
people. But I I think recently like all
the talk of you know there's been a lot
of tech layoffs and I know that that's
kind of tried to be tied towards AI. I I
don't I think that's a lot more related
to the overhiring that happened during
the pandemic uh than I think it's
related to to having those people
suddenly like an AI is doing that job
for them.
>> Um yeah, I think that's kind of borne
out if you look at the look at the data.
>> Yeah. So I do think it's going to have
consequences on the labor market. I
think in in when history looks back at
this, we'll largely say that it was a
good idea the same way people say the
computer was a good idea, the same way
people say the industrial revolution was
a good idea. But it is going to be a
chaotic moment and it does require
attending.
>> Do you have concerns about what this
will do in terms of inequality? I mean I
I think about the downsides. I think
longterm
um it's you know value accretive which
means that's a good thing for society in
the same way that the steam engine was
and the internet was. But I think the
the the the big concern that seems super
likely to me is um that the value is
acred to the people who own the AI and
that yes maybe some of us might might be
getting some value out of using AI but
we won't be the ones who own it and it
will only make wealth inequality
even worse which could have all of its
own impacts. Do you worry about that?
>> I do worry about that. Yeah. Um income
wealth inequality is the thing I is one
of the things I think is the most
pressing issue. I think yeah I think
it's one of the most pressing issues for
the world right now and I do worry that
this technology similar to other
technologies stands to exacerbate the in
the wealth inequality that was already
rising
>> over the past you know few decades. Um I
think the correct solution to that is
policy. oftentimes when people thinking
about the economy that they they kind of
forget that this is a system we create
and it's a system that can be subtly
pushed in one direction or another
direction and you can add policies you
can change things in order to make sure
that this works for everybody for all
people in in your country or in you know
in whatever organization you're within.
Um, I I think that's the conversation I
want the world to be having. And one of
the reasons why I'm very vocal about
saying, "Hey, I don't think we're
getting to AGI," is that the AGI
conversation often distracts from that
conversation. Cuz if you're talking
about, oh no, what if we make a digital
god and it kills all people, it's very
difficult to have the conversation, hey,
like, you know, do do we have the right
policies in place to encourage better
income distribution such that we don't
we don't end up in a bifurcated society,
which which I don't think anybody wants.
>> What kinds of policies or what do you
think that would look like? This is my
first question and then my second
question is as someone who cares about
that are you a pariah in the tech
industry because from what from my
understanding
there is a feeling of if you're talking
about policy and regulation then you're
a lite and you're just trying to hold AI
back and you're just scared. So
I guess how do you think about those two
questions? Am I a pariah for talking
about that stuff?
Um, no. No. But I also don't live in
Silicon Valley,
right? Like I don't I live in Toronto.
I'm certainly in the tech scene, you
know? I talk to I talk to VCs all the
time. I talk to, you know, other tech
people. Um, I talk to, you know, lots of
people thinking about this.
But
um would I you know would I be a pariah
in I'd certainly have I certainly have
lots of different views than people
would have hanging out in the the
remnants of the effective altruist
parties in Silicon Valley right I
certainly have very different views than
than the culture that developed there I
I mean am I like I'm certainly not a
lite right like I'm certainly not
against the creation of technology but
you know having been to where was that
town I went
Uh
there's a town in England that's that
was kind of the epicenter of that and I
went to a museum there on Leites. That
was very interesting.
>> I wish I had it. I but I know I've
studied it. I know what you're talking
about. I wish I had the know.
>> Yeah. You know, and a lot of the people
at the time were were you know what they
were frustrated at the loom for making
their economic situation worse. right
>> now. I again, we all look back at the
automated loom and we think that was a
good idea and the economic situation
that people live in now is better than
the economic situation that they were
living in during that time. Um, but I'm
empathetic. I'm empathetic to saying,
hey, like my economic situation is
shitty and it's shitty at a systematic
level, at a at a population system
level. Let's figure out how to how to
make that better. Right? So, I am
empathetic with that. Um, so would I be
a pariah? I don't think so. I I think
actually a lot of people know this. I
think if you go to Silicon Valley and
you tell somebody, hey, you know, income
inequality is bad.
It's hard to live in that city and not
think that. Just looking at uh the
future of Coher, uh reportedly Coher is
looking to go public. I don't know if
you can talk about that, but that's what
we've been reading. Can you tell us
about those plans? Aiden and Ivan and my
goal in creating this company was to
create something that outlasted us was
to create a generational company. Like
that's motivating, that's exciting.
That's a fun thing to be part of. That's
what we're excited about. Um I think the
right way to do that is to become a
public company. I think that's how you
can build I think that's like I I like
those mechanisms. I think that's how you
can build a company that is bigger and
longer than you. And that's exciting. I
think the tech we're building
is speaks for itself and is getting
there. Um, I think the the customers
that we've closed and and the
relationships we've built and the value
we've been able to add to our customers
um is is something I'm enormously proud
of and something I want to keep doing
and something I think would be best done
through a public company. We'll be right
back.
Support for this podcast and the
following message is brought to you by E
Trade from Morgan Stanley. Simplify your
finances and discover the convenience of
investing and banking all in one place
with E Trade from Morgan Stanley. Choose
from a wide range of investment choices
and award-winning banking solutions
together in one platform. Plus, get up
to $1,500 when you open a brokerage
account with a qualifying deposit today.
Learn more at errade.com/offer.
Banking products and services are
provided by Morgan Stanley Private Bank,
National Association, Member FDIC. Terms
and other fees apply. Investing involves
risks. Morgan Stanley Smith Barney LLC,
member SIPC.
We're back with Firsttime Founders.
We've discussed this sort of the decline
of the IPO on on our markets podcast a
little bit. Just the fact that there are
there are fewer public companies in
America than ever. Um, and also the fact
that so many of these massively
transformative companies are taking so
long to go public. The idea that OpenAI
is only now, I mean, they had the
nonprofit issues, but the reality is
this company is supposedly going to go
public at a trillion dollar valuation,
that's crazy. Uh, so how did you balance
it? It seems as though the startup world
is more interested in staying private as
long as possible uh for or at least
that's what the data would tell us um
versus going public. So, how did you how
do you think about going public? What
are the the pros? What are the cons? And
and why nowish?
>> Well, I've made no promises to timing.
>> Yeah, fair enough.
>> We've made no promises to timing. Um,
but I do think, look, I I don't Yeah,
again, I don't know what OpenAI is
doing. Um, I I think there are very I I
look forward to the interesting books
that will be written about that company
over the next 10 years. I don't know
what's going on and I don't know when
they're going to go public. I know that
they've announced stuff, but I I don't
know. It's it's a very it's it's its own
beast. It's its own unique and
interesting um company and I'm sure
people lots will be written to describe
that story.
For us, our business looks a lot
different because we don't have an
enterp we don't have a consumer
offering. We don't have the same losses
that they have. Um, we're not losing
money on every customer. Our margins are
actually, you know, they look a lot more
like SAS margins. When we work with a
customer, we deploy our models into
their environment, right? And that that
allows that model to access their
private data securely so that we can't
see it. It also makes the nature of our
margins completely different.
Um, you know, I think that puts us in a
very different position than the
consumer companies that are out there.
And I think that puts us in a position
that makes it a lot more resonant with a
public market.
A lot more understandable, a a lot more
uh yeah, it looks a lot better. Um, so I
I do think the right way for us to make
sure that this company outlasts us and
continues to deliver is to eventually go
public. Um, when Yeah, I don't know. And
there's an interesting thing you're
pointing out about there being a smaller
number of public companies. There's an
interesting thing you're pointing out uh
about companies staying private for
longer. I don't think those are
unrelated to the income inequality,
wealth inequality stuff that we were
talking about earlier,
>> right?
>> There's an interesting dynamic in the
economy going on right now. And like all
of those things are kind of related.
>> Um but for us like yeah, no, that's
that's what we're gunning for. Um and I
think that's the path that we're on. You
are a Canadian company. You're based in
Canada. Uh
how do you think about AI as a sort of
international geopolitical race? Um
we've got some big AI companies in
America, some big AI companies in China.
I guess Mrol is another one that's in
France.
>> And there's us in Canada. And that's it,
>> right?
>> Yeah. There are four countries in the
world that can make this technology.
tell us more about what that means for
for society.
>> Yeah, it's a that's a strange one. Um I
think that's when you think about how
difficult it is to make this technology
and how resource intensive it is. It's
not super surprising that there aren't
that there like just in it's not
surprising that there aren't that many
companies doing it. It's not surprising
that there aren't that many countries.
So, but it is it is a strange reality
that yeah, there are four countries in
the world that can build this tech. When
I think about the what that means for
geopolitics,
I think this technology is best in the
same way, you know, I use this analogy
of like rockets. It's like building
rockets. Um, another analogy is to say
it's like building power plants. It's
like building infrastructure. The
technology is is a lot more like
infrastructure
than previous computer science efforts.
Um, and so I think it's a good idea for
countries to have the ability to build
infrastructure themselves. Like it's a
good idea for countries to be able to
build their own nuclear power plants.
Like that's useful. That sets up the
country for success. Um, it's a good
idea for them to be able to build their
own roads, right? Like like
infrastructure for people is good and
it's good for countries to be able to do
that from a strategic perspective, from
a security perspective, like from an
economic perspective. It's generally a
good idea. So I think this technology is
important for countries to to be able to
build. I think there's ways that that
countries can work with the providers in
order to give that ability to their
country. Um and I think that's something
that a lot of the world is seeing right
now. You know, we had two decades of the
history of tech really being American,
really being centered on America. And
America is that dynamic, fastmoving,
ingenious place that will is going to
continue to be defining on the
technology and technology in general.
But I do think it's good for the world
to have tech that comes from other
places, you know, to have a more
distributed um view on on how technology
is developed and what it can do for
people. So that's one of the reasons why
I'm happy to be building out of Canada.
>> Is that the the thing that changes the
trajectory of geopolitics? Like for
example, you made the comparisons to to
to other technologies. I think some
people would also make the comparison to
the nuclear arms race. Not to say that
AI is like a nuke, but to say that it
was the belief of nations that this is
what will tilt the balance of power
across the world. We have to build these
things because if we don't and if Russia
or anyone else gets their hands on this
transformative technology, it will
completely upend uh the geopolitical
structure of Earth. Do you view AI in
the same way? Not in the sense that it
would be I'm not making a nukes and it's
going to be destructive point. I'm
making a point of the power of it. Is it
a is it is it a question of whoever
builds the the the AI first and whoever
builds the best AI they will be the most
powerful force in the world. Do you see
it that way?
>> No, that's a little extreme.
>> Okay. Yeah. I see it as like a a
strategic and a imperative for countries
to have this technology to facilitate
economic growth to do stuff. In the same
way I see it's imperative for them to
build, you know, roads, really great
healthcare, build, you know, nuclear
power plants, build wind, build other
pieces of of infrastructure. I don't
think I would go so far as to say it
will be the defining thing. Um,
certainly, yeah, the the nuclear bomb
analogy I disagree with. Um, and I know
that's often used when people are
talking about AI as an existential
threat, but again, because I don't think
transformers are going to get us to AGI,
I don't think they pose an existential
threat. And so I don't I don't think
that analogy serves us. Um, in
conversation, I do think it's important
to think about the technology as
infrastructure and infrastructure that's
good to build, but one piece of
infrastructure amongst many pieces of
infrastructure that are good to build
and important to build in this moment.
And I do, we're certainly in a dynamic
and changing geopolitical time, right?
like this you know these these are
unprecedented times as has been said for
the past decade
but uh yeah and I and I think the
technology will have an impact on that
but I don't think it will be the
defining thing
>> you are one of the leaders in AI which
is the most important and transformative
technology
it's certainly of my time I'm Gen Z I'm
I I was not there to see the internet be
created and built. So I think you know
this is an extremely important moment
not just for America or Canada but for
the world.
Does that weigh on you? What is it like
to be a founder who is at the forefront
of this world changing technology?
>> Yeah, it weighs on me. Yeah, it's
totally a strange place to end up in and
not a place I thought I would. It
excites me. I love working with Coher,
working at Coher. I love working with
all the people at Coher and I get really
excited and I'm enormously proud and
occasionally deeply moved by the work
that we're doing and the group of people
that I get to spend time with working on
it. It is a complicated emotional
experience to think, hey, this
technology is, you know,
the defining narrative and we are one of
10 companies in the world,
>> four countries in the world that are
building it, you know. Um, I still love
I mentioned earlier like I still I love
the tech and I'm moved by it and that's
that influences how I think about this.
Like I think we're building something
beautiful and cool and can be useful.
Um, and it's it's very meaningful to the
world and to the the you know to the
people around me and that's interesting.
Um, there is a subsequently a pressure
and an intensity
that I did not anticipate when we
started this company. I I don't think
any anybody did. Um,
I think I solve that by staying grounded
in things that have nothing to do with
tech sometimes, right? I think that's an
important part of the way that I'm that
I live my life is by doing stuff
occasionally that is completely
unrelated to to AI, to Transformers, to
to tech itself. And I think that might
be why I have pretty different views
than the rest of the people in similar
positions to me. A lot of young people
uh watch this show. What would be your
advice to young people, not necessarily
just founders, but I think young people
in general uh perhaps even young people
who are concerned about their job
prospects, their career prospects,
people who believe that AI could be
taking their jobs. I mean, from the guy
who's building the AI, what would your
advice be to young people? terms of
jobs, my advice is to it has been the
same for young people for a while, which
is that I know I meet a lot of people,
young people who are like anxious about
making the right decision. They're like,
I got to work on this cuz that's going
to be the right thing or that's going to
be the right thing. And my my advice has
often been, look, the world's too
chaotic for you to predict what's right.
Like you can't like every every year you
could read an article of somebody saying
the next big job is this and you got to
go into this and they're almost always
wrong. And so it's just too chaotic. You
can't predict it. What you should
instead do is focus on what you're
interested in. And what you can optimize
for is your own excitement, your own
curiosity, your own interest. And when
you're thinking about what career you
want to pick or something, you should
first and foremost be like, well, what
am I excited about? What am I interested
in? And conditioned on that, your
ability to be successful is much higher
than conditioned on you choosing, you
know, what you think is the optimal
decision at that moment. So I would
really encourage young people to like
follow their their curiosity, follow
their passion more than they think,
follow what's optimal just cuz you you
can't predict it. It's really hard. Um
my other advice is to when the central
when the narrative around the world
these days is one of like it's an
unprecedented chaotic absolutely crazy
time. I definitely encourage people to
learn about history. just read just
whatever history from whatever time like
whatever you find ancient history
prehistoric pre prehistory um you know
enlightenment history modern history
like wherever literally wherever yes we
live in unprecedented times yes stuff is
chaotic and weird right now and I think
when the history of this moment I think
people are going to read the history of
these few decades with curiosity in the
future um
but there's been a whole lot of crazy
times
there's been a whole lot of absolutely
nut stuff that has happened in the
history of humanity. Um, and and it is
calming sometimes to read about those
and understand the good things that
happened, the bad things that happened,
the way stuff continued in the face of
it. I find that grounding and that
grounding is helpful for keeping you
focused on like what you're interested
in, what you're passionate about, what
you're curious about. Nick Frost is the
co-founder of Coher. Nick, this was
great. We really appreciate your time.
>> You as well. Thanks for the
conversation.
Thank you for listening to First-Time
Founders from Profit Media. We'll see
you next month with another founder
story.