May 22, 2025 • 36 mins
Alex Wiltschko got obsessed with perfume when he was 12 years old. He grew up to be an AI researcher at Google. Then he started Osmo, a company that fused his job at Google with his childhood obsession: Osmo is using AI to teach computers to smell.
The company is getting into the perfume business, and it plans eventually to use scent to diagnose disease and detect security risks.
Get early, ad-free access to episodes of What's Your Problem? by subscribing to Pushkin+ on Apple Podcasts or Pushkin.fm. Pushkin+ subscribers can access ad-free episodes, full audiobooks, exclusive binges, and bonus content for all Pushkin shows.
Subscribe on Apple: apple.co/pushkin
Subscribe on Pushkin: pushkin.com/plus
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:15):
Pushkin, tell me what we don't know about how smell works?
Oh jeez, be sure to tell you what we do.
This is Alex Wilscoe. He's the co founder and CEO
of a company called Osmo, and despite his protests there
he did tell me some of the things nobody knows
(00:37):
about how smell works.
Speaker 2 (00:40):
Why do things smell the way that they do. Why
can we smell certain things and not other things. What
is the logic of how molecules are combined to create
beautiful smells? Why do some smells create incredibly powerful emotional
associations instantly and others seem neutral? Right? Why do something
(01:01):
smell different to people? I think we have a hints
in all these directions, but we have nothing like musical scale,
where we have nothing like a periodic table. We don't
know any structure to why things are the way that
they are. It's a ton of mystery, and that's what
makes it so exciting to work on this topic, is
like there's so much we don't know, and to.
Speaker 1 (01:22):
Be clear, like with light, we just know whatever. If
you tell me the frequency the wavelength, I can know
exactly what color you're talking about, or the same thing
with a wave form of sound, right, and so but
if I give you some random molecule and say what
does it smell like?
Speaker 2 (01:40):
Do you know? So that's what I've spent a lot
of my professional life working on. It is exactly that question. Yeah,
which is a draw structure of a molecule on a whiteboard,
point out it and say, hey, what does this smell
like wood or flowers or fruits or whatever? And so
there is no way to know that for sure at all.
(02:01):
But there's no good way even statistically to predict that
without using large data sets, and at least in our hands.
You need neural networks, you need deep sorting in order
to do that. I'm Jacob Goldstein, and this is what's
your problem?
Speaker 1 (02:20):
The show where I talk to people who are trying
to make technological progress.
Speaker 2 (02:24):
Alex Wilch.
Speaker 1 (02:25):
Goo's problem is this, Can you use AI to teach
computers to smell? And once you figured out how to
do that, can you build a profitable business around it?
Osmo's fun out of Google in twenty twenty three. The
company recently launched a fragrance house to develop new perfumes.
They've also done some work using scent to detect counterfeit shoes,
(02:47):
and in the long run, they plan to use scent
to diagnose disease. Before he started Osmo, Alex worked at
Google as an AI researcher. Before that, he got a
PhD at Harvard studying how mice respond to scent. But
maybe the most important part of his bio came even
earlier in his life, specifically when he was twelve years
(03:09):
old and went off to summer camp in his home state, Texas.
Speaker 2 (03:13):
I was from a small town college station, and then
most of the kids were from big towns like Houston
and Dallas and Austin and San Antonio, and I hadn't
really been exposed to, like I don't know fashion trends
or you know, what was cool or popular. But everybody's
all lumped together in summer camp. And then there is
(03:34):
this thing called perfume that some of the richer, frankly
richer and more popular kids had, And it was just
amazing to me that these boys could spray themselves with
this invisible mist, a clear mist, and then for the
next four to six hours people around them would treat
them differently, and that just blew my mind right, like
(03:59):
there's no I can't I can see the clothes. Yeah,
I can see how they act and walk and talk
and how they you know, posture and all that, but
I cannot see the fragrance. Yeah, but yet it is
obviously doing something magical. It's like an.
Speaker 1 (04:13):
Axe body spray ad what what does that cause you
to do?
Speaker 2 (04:18):
When I get home, beg my parents to buy and fail.
We shopped at TJ Max and I started to really
look out for fragrances there and then it just kind
of snowballs from there, which I just realized there was
like a whole lot of these things. I guess what,
you can just try them, and some of them are
(04:38):
actually way better and more opinionated and more beautiful. I
don't I didn't have the vocabulary then, but like it
just it was clear to me early on that like
I never really thought about who made the clothes, but
I started to think about who made these perfumes, because
it was clear that there were choices that were being made.
And like I just remember trying, and this is years later,
(04:59):
trying Bulgari black, which really kind of clued me into
this world. Bulgary black is not necessarily a great fragrance,
but you can experience the top, middle and base notes
in like forty minutes. Forty five minutes is pretty short,
and so like a bigger fragrance, like a creed, events
will last on your skin for a day, and so
the whole fragrance unfolds. I mean, top notes will last
(05:22):
max Fifteen thirty minutes, but the heart might last for
several hours, and the base note might last for ten hours. Right,
so it smells different.
Speaker 1 (05:30):
You can still smell it, but it smells different whatever
an hour after.
Speaker 2 (05:33):
You put it on and four hours after that, because
a great fragrance is actually many different fragrances within it. Yeah, Right,
There's the first one which peels off or that quickly
burns off quickly. There's the second fragrance, which is the
heart note, which will last for you know, sometimes hours,
but in this case, like another twenty minutes. And then
the base note, which is a third fragrance, which is
what's left after those two burn off. And it's like
(05:55):
three acts of a movie, right, I think it's quite beautiful.
So how do we.
Speaker 1 (05:58):
Get from you being a teenager preoccupied with fragrance to
you using AI to predict how malls will smell?
Speaker 2 (06:10):
Yeah, Like the computer part was always different from the
fragrance part I just I love computers. We always had
computers at home. I started programming around I don't know,
eight years old. It was my life, Like, my entire
life was computers for a long time, still lose in
a way, and fragrance was not a part of it.
I got into you know, statistics, which became machine learning
(06:31):
around the same time. Again for totally independent reason. There's
this thing called the Netflix Prize. It was like one
of the first competitions to build great mL algorithms. I competed.
Speaker 1 (06:40):
Now, that's basically to tell me what else I'll like
on Netflix, right, That's what that contaest was. Like, if
I've watched whatever, if I watched Succession and the Sopranos,
what should I watch next?
Speaker 2 (06:51):
Then you're gonna like another kind of dark, but you know, funny,
kind of soap opera type of a thing, exactly. And
so Netflix did a really bold thing, which is they
released a data set and said, here's what good looks like,
here's how we measure it. Have at it. And they
paid a million dollars to the Winter, which was a
combination of a few teams. But what they really did
is they brought a particular kind of machine learning into
(07:13):
the forefront called collaborative filtering. Really showed that this stuff worked,
and by the way, other companies were already racing to
use this, So like this recommended systems was a big thing,
but Netflix was putting it out into the public and
allowed a kid like me I think I was eighteen
or nineteen years old to actually compete and pretty well
in that. And so I just got exposed to this
world through that and it was super fun. I mean
(07:36):
they gamified it and had I had a blast. So
that was my first exposure to machine learning.
Speaker 1 (07:40):
Turned out to be a good time to start working
on machine learning, Yeah.
Speaker 2 (07:44):
Totally, because if I had started now, they wouldn't let
me in because that's probably ten years ago. Yeah, ten
years ago. Yeah. And then, you know, I was doing
my undergraduate training in neuroscience, and I was studying more
behavior than old action because it actually turned out that
olf action was a hyper specialized sub field of neuroscience.
(08:05):
I didn't realize how niche it was. I loved smell
and I was doing neuroscience, and I knew I wanted
to do smell neuroscience the fancy word for nactual factory neuroscience.
And so there's really two universities in the world that
like have a critical mass of these researchers. It's Columbia
and it's Harvard, and I applied to both. I went
to Harvard, and I realized nobody cares about this problem.
(08:28):
Nobody cares about why molecules smell the way that they do.
There's a much longer conversation as so why that's the
case and why that's still persistent. Now that's changing. Well,
let me ask you this. Let me ask you this.
Speaker 1 (08:38):
At that time, I mean I get it as a
basic research question.
Speaker 2 (08:42):
I mean, I'll tell you.
Speaker 1 (08:43):
We was talking with the producer and editor of this show,
and we were getting ready for this interview, and we
had this interesting conversation talking about scent and what you're
working on whatever. And then I went down and I
saw my daughter, and so, what are you organized? Said
this guy who's trying to figure out scent and teach
computers to smell?
Speaker 2 (08:58):
And she said why, I said, I don't know.
Speaker 1 (09:01):
I ask, so why was it compelling to I get
it as a basic research question, but at that time
was like, there were there applications that came to your mind.
Speaker 3 (09:12):
Look, the the steps of development of this thing that's
now OSMO went through those different iterations.
Speaker 2 (09:23):
You know, I started as an academic scientist and I
was trained in that world, and then I left and
it did some entrepreneurship, but it ended up in industrial research,
and they're like being curious, frankly was enough, and the
idea this is this is Google, this is now a
Google Brain. Yeah, and there's a few steps in between.
Speaker 1 (09:41):
But basically as you're an AI researcher at Google at.
Speaker 2 (09:44):
This moment when you're doing industrial research, right, yeah, exactly,
and Google Brain at the time now it's Google Deep
Mind very much had like a thousand flowers bloom mentality,
and so people were working on crazy stuff, including me
working on something Bell Labs.
Speaker 1 (09:56):
It's basically like Bell Labs of the twenty first century, right, you.
Speaker 2 (09:59):
Have it, exactly, Bell Labs, Erox Park, that kind of
avid and it truly was dreamy, sounds dreamy. It was awesome, right,
And it was also a moment in time, and now
I think that moment's going for better or for worse.
The idea was pretty straightforward for Google, which was, are
the products at Google know what the world looks like
(10:19):
and know what the world sounds like, and that's useful. Right,
that's information that Google's organizing. If we knew what things
smelled and tasted like, that would be useful, right. Uh huh.
Speaker 1 (10:28):
The original mission of Google is organized the world's information, right.
Speaker 2 (10:32):
Exactly, and make it universally accessible and useful. And there
was a whole slice of reality, huh, the chemical slice
of reality that was invisible, right, not just to Google
but two computers. Yeah. Yeah, and that felt really important,
and we had agreement and buying all the way up
to the executive level. They're like, yeah, let's go, let's
go look at that.
Speaker 1 (10:50):
So you're doing a like basic AI research at Google
and you decide to see if you can basically use
AI to figure out scent, to say, here's a molecule,
what does it smell like?
Speaker 2 (11:02):
Right, that's the basic endeavor. How do you do that?
What is it that you actually do? Yeah, so first
it starts to innovation, which I was like, let's figure
out smell. But it actually was a lot more natural
than I think it sounds, which is scent is just chemistry.
It's molecules, and we got to do AI for molecules. Right.
(11:22):
If we're going to do AI for scent, and the
thing that had happened in between. You know, there's a
five year period between my academic life and my industrial life,
and what had happened in those five years is actually
some of the people I did my PhD with and
then some of the people I ended up working with
at Google Brain really cracked machine learning or AI for molecules.
(11:45):
But they didn't do it for scent. They did it
for a few other things. They did for drug discovery,
and they did it for like materials discovery, so like
new materials for LEDs.
Speaker 1 (11:54):
Right, So you happen to be doing essentially basic research
at Google at this moment when there is this new
way to use AI that is well suited to molecules,
and you say, we can do let's do it, Yeah,
let's do it.
Speaker 2 (12:11):
Yeah, we can do it. The other pieces are great.
You got the algorithm. Where's the data? Classic? That's the
classic AI question, right, like, exactly where's the data? What
I did know just from being obsessed and in this
world for a long time prior to that, there were
these collections of data sets that were honestly really more
like magazine catalogs for fragrance ingredients, and so there were
(12:33):
these catalogs basically saying this is the ingredient, this is
the molecular structure of this ingredient, and here's what it
smells like. And by the way, the rating of what
it smells like was done by a professional, by a perfumer.
And so the special sauce that we added is we
we went and we got that data and we fused
a few data sets together, and we cleaned it very carefully,
(12:54):
and that that hadn't been done.
Speaker 1 (12:55):
And it's something it's like five thousand dish right, it's
five thousand or so different molecules.
Speaker 2 (13:00):
Okay, yep, exactly. And here is this the one with
the list. I love the list. Here I have it.
Speaker 1 (13:06):
This the one sweet fruity, vanilla, powdery, fluoral, barry, fermented,
nutty ozone, buttery musk.
Speaker 2 (13:12):
It's it's that lispright. Those are they? And there's one
hundred and thirty eight of those descriptors I think that
we used in that data set. Sometimes we use smaller subsets,
but the full set originally is about one hundred and forty.
Speaker 1 (13:24):
So okay, so you have your whatever, your five thousand
molecules labeled with one hundred and forty different sets. You
train your AI model on this data set, and then
you want to find out does the model work?
Speaker 2 (13:38):
Does the AI work right?
Speaker 1 (13:39):
If I give the model some new molecule molecule that
wasn't in the training data, will it know what that
molecule smells like? And to test that to answer that question,
you actually.
Speaker 2 (13:50):
Do this study.
Speaker 1 (13:51):
So you get a bunch of people to smell these
molecules that are not that your model was not trained on,
essentially right, and say what it smells like? It's weird
like what? You don't actually care what it fundamentally smells like.
You just care what everybody on average thinks it smells.
Speaker 2 (14:05):
Like, because guess what, that's what it's what smell is? Yeah,
that's what do you think of smell? Yeah? Ye?
Speaker 1 (14:11):
So you asked this panel to what do all these
molecule smell like? And then you ask the model what
do they smell like? And you compare the results and
how does the model do?
Speaker 2 (14:23):
That was really the threshold of breakthrough in my mind
was like are you worse than a person? Yeah? Or
are you slightly better than a person? And we got
slightly better than a person, which was a breakthrough in
my view.
Speaker 1 (14:36):
Right, and so yes, so that paper you published in
Science and you started Ozmo kind of around the same time, Right,
you started that study at at Google, is that right?
And then by the time it was published you had
spun Osmo out of Google, right, that's right. So you
have this map, you have this model that can basically
given a molecule, predict pretty well what the average person
(14:58):
thinks that molecule smell like. But there's still a second problem, right,
which is, in the world, in the wild, you don't
know what molecules are in the air. You don't you
know what molecules somebody's smelling. And so for that second problem,
you need to try and build some kind of automated
system for figuring out what molecules are in the air
(15:18):
at a given that's correct.
Speaker 2 (15:20):
Getting to one molecule structure is actually not trivial. So
to go from a physical thing and know all the
molecular structures like not a solveable. So there's a lot
of ways to do that. There's a lot of chemical
sensors out there, none of them will just tell you
the formula, right, So that's hard, really hard.
Speaker 1 (15:41):
So there's like a chemistry problem of like isolating the
molecule basically and deriving the chemical formula.
Speaker 2 (15:49):
Exactly taking a real smell and it's composed of a
bunch of different molecules with different structures, and there's different amounts,
there's ratios. You got to get that recipe out of
the air. So that's on. That's hard. That was unsolved
at the time to do it in an automated way.
And by the way, if we're following this story chronologically,
we hadn't done this yet, agah, but we knew we
(16:10):
had to do that, right, So we knew that, Okay,
if we wanted to actually digitize the world of scent
and have a record of what the world smelled like
and maybe even replay it, we're going to have to
do this. We needed to automate that and have it
be automatic, and that's what we did.
Speaker 1 (16:24):
So basically, you can put any smell into the machine
and it'll tell you what it's made of. At this point,
oh yeah, so you're setting out to start OSMO, Like,
what are you thinking of in terms of the set
of potential commercial applications.
Speaker 2 (16:39):
So we really had we had three in mind, and
there's still very much present in mind the focuses has
become a lot crisper though in terms of what we're
concentrating on. We know the fragrance industry is huge and
very profitable, and it's also something I personally love. That's
(17:02):
a thing we want to automate and understand. And then
we know that dogs can detect things, right, and so
we know dogs can detect harmful substances like drugs or bombs,
or things that just shouldn't be there, like produce where
it shouldn't be being shipped. And then we also know that
dogs and even in some cases people can detect health
(17:22):
or disease states. Right. We know that missus Milner, a
nurse in the UK, was able to smell Parkinson's disease
and she's since been able to teach that skill to
other people, which is really amazing. And then we figured
out all the chemistry of what's actually being smelled. We
know that there's many many instances where there is a
(17:43):
scent signature to a disease or to a wellness or
to a health state that hasn't yet been fully figured out, right,
but we know that they exist. Those are the three, right,
So fragrance industry really security and supply chain and health
and wellness, and I view them in that order because
(18:03):
that's like the order in which I think we can
be useful to the world. Right, Designing fragrances is something
that's much more attainable technically, and frankly, it's just a great,
much faster sales cycle to be business to be in
than ultimately diagnostics, which are so hard. Right, I mean,
it is my north star. It's like where I want
to take the company. But I also have no illusions
(18:25):
about how hard that is, and I just I've seen
all the failures of the companies that have attempted it,
and I think I've learned from what hasn't worked, and
so I'm incorporating those learnings into how I want to
build the company, which is, build a great business in fragrance,
build beautiful fragrances for the world, and then strike out
from that position of strength and to even more ambitious frontiers.
(18:49):
We'll be back in just a minute.
Speaker 1 (19:00):
When I first heard about the work Alex was doing
at OSMO, I understood why it would be useful for sensing. Basically,
you might be able to build automate sniffing machines that
could say detect cancer in a person or detect a
bomb in a suitcase. But I couldn't figure out truly
what the business case was for perfume. And in fact
(19:22):
Osmo has recently launched a perfume business. It's called Generation.
So I asked Alex, why is using AI and fancy machines?
Why is that better than just designing perfume in the
traditional way.
Speaker 2 (19:36):
We can go from the first kind of client demand. So, hey,
I want to create a fragrance, and here's who my
brand is, here's what I want to do. So just
that description to a starting place of a fragrance in
a minute or two.
Speaker 1 (19:50):
What happens at a traditional perfumer when somebody comes in
with that request.
Speaker 2 (19:54):
Well, so let's say you're an emerging Let's say you're
an emerging brand, right, So you're starting out or you
have your first product and you want to add a
second one. But you're small, right, You're not making a
billion dollars in revenue. You're making less than that. So
if you want to make a new custom fragrance, good luck. Right,
you're not going to be able to get the attention
of the big fragrance houses because they want a service
(20:15):
business that's like millions and millions of dollars and you're
not big enough yet. So if you want a great
custom fragrance that your consumers are going to love, and
you want to do it quickly so you're responding to trends,
you aren't going to be able to get it done.
So you have to make compromises, right, So if you
(20:35):
want to move fast, you're gonna have to use a
regurgitated fragrance. It's also called a library fragrance, which means
somebody else in the market has your spell.
Speaker 1 (20:42):
I'm imagining that people who sell it call it a
library fragrance rather than a regurgitated library fragrance.
Speaker 2 (20:47):
They do, they don't say regurgitated, but that's what it
effectively is, right fair.
Speaker 1 (20:51):
Regurgitated does have a particular old factory connotation, so it's
a level.
Speaker 2 (20:57):
It's vic or all sticks in the mind.
Speaker 1 (20:59):
What like, And I'm not I just genuinely don't understand, Like,
why can't somebody just have a company with a bunch
like who knows the molecules? You know, who knows what
the five thousand milecules in the book smell like? Because
they've got the book and they can just use the
book and be like, oh, you want this, let's try that.
Speaker 2 (21:15):
Do you know what I mean? Like, I'm not trying
to be difficult, but I.
Speaker 1 (21:18):
Genuinely don't understand why you need the technology to do that.
Speaker 2 (21:24):
Yeah, I genuinely didn't understand this either. And there's there's
a class of professional called a perfumer, and their job
is to do what you're describing, which is, Hey, I
know all these ingredients and I'm going to make them
in order to create your fragrance. So they typically there's
no perfumer that knows five thousand ingredients, but the best
perfumers know a thousand or two thousand ingredients. Most perfumers
(21:48):
work with two hundred, one hundred, two hundred ingredients. So
already there, like where we're there is very few people
in the world that can do what you're saying. Yeah,
they can do, and then how what are they going
to work on? Right? So it might take them weeks
or months to create a fragrance. They're working on a
few at a time. Why would they work on an
(22:10):
emerging brands fragrance when they can go work on a
much larger account. So there's just very very limited number
of people who can engage in the fragrance creation process,
because it is difficult. It's not so much identifying, Hey,
all these molecules smell this particular way, and therefore I
should be able to mix them, Like what ratios do
you mix them in? Like? What are the rules? Right?
And now you're actually getting into designing a system which
(22:31):
understands sent well enough to create new fragrance formulas, is
starting places, and then of course a perfumer finished system.
But you're right, it's like, oh, why shouldn't that exist?
And then when you actually start to peel back the
layers one by one, you realize, oh, you actually have
to build what we built. It's actually in order to
answer that question.
Speaker 1 (22:46):
So presumably now your model cannot only predict what one
molecule is going to smell like, but it can predict
the combination of molecules. I mean, is it predicting? Does
it know concentration? Like does it know oh yeah, yeah,
how good is it?
Speaker 2 (23:02):
I mean you have a perfumer on staff? Why? Well,
I think the goal of tools is to have them
in the hands of creatives. And there's many steps to perfumery,
but I think there's three that are relevant for what
we're talking about. The first is a perfumer when they're
(23:22):
when they're starting on a project, they have to have
a starting place, they have a starting formula, and then
they do their creative work step two to evolve that
formula to exactly what the customer wants, to a creative
expression that the lights the consumer as well. That's the
funnest part. Perfumers love that that is actual creation in
the creative part. Number three is then it has to
(23:43):
be the right price. It has to be compliant with
with a regulatory compliance. There cannot be allergens, all that
stuff that's more like sound engineering than it is composition
or being a rock star. Steps one, the starting place,
step three, all the regulatory requirements. That's where we spend
the most energy in building these tools. And then a
perfumer is the person that is taking the formulas from
(24:06):
starting place to creative endpoint and then handing it off
for like regulatory finishing. And they're just way more effective
with these.
Speaker 1 (24:13):
Tools, at least for now, right Like, that's the way
I feel using an LLM, like I feel like I
have a window and me Plus the LLM is better
than the LLM alone and we haven't. That window hasn't
closed yet, but I'm not optimistic about my long term prospects.
Speaker 2 (24:29):
We'll see though, I mean, but like listen, I honest
belief here, like the tools will get better, but the
drive to create will never go away. And I think
people will always want to know about the person behind
the creation in a way, and it's not uniforms. So
I don't think people want to know the perfumer behind
the hand soap in the gas station. They just don't,
(24:52):
right it. But there I think will always be room
for craft and creative use of tools, and the profession
that uses those tools might change radically, in the industry
in which those tools are used might change radically, but
the tools will always be wielded by people, but the
work that's being done might be unrecognizable. So you know,
(25:15):
we'll see how the world evolves. But like I just
like AI is like an engine, it's just a technologist,
just a tool.
Speaker 1 (25:23):
So what's the what's the business model? Just briefly for generation,
like how what you know?
Speaker 2 (25:29):
What's the model? The business model really simply is we
all design the fragrance for you and then you'll buy
that fragrance to put in your products or will even
actually create the full finished product. We'll put in a
bottle for you if you if you want. We are
behind the scenes. We're an engine supporting brands. We're not
a brand ourselves, and we're here to make beautiful fragrance
(25:50):
products for for brands. So what's the frontier like you have?
Speaker 1 (25:56):
You know, on the business side, the generation is kind
of the central thing you're working on now, but on
the more on the on the research side, like what
are you trying to figure out?
Speaker 2 (26:05):
Now? What are you working on now? So there's there's
our starting place, which is why does this molecule smell
the way that it does? And we can never stop
getting better at that. Then there's the next question of
why does this mixture of molecule smell the way that
it does? And we can never stop getting better at that.
And then there's do you like it? Which is maybe
(26:27):
the most important question from a business perspective, or who
likes it? And in what context? Yeah, exactly exactly, which
is it's not just the formula as the input to
this model, but there's also who you are, what are
your experiences, where are you from, what are the other
things in your life that you've got.
Speaker 1 (26:42):
That actually goes back to your Netflix collaborative. It gets
me like if I watched Succession and the Sopranos and
I'm fifty and.
Speaker 2 (26:51):
Then what's the cologne for me? Yeah? Exactly. And so
I was very fortunate to be able to start this
company with a guy work with at Twitter name. His
name is Rich Witcombe. He's our chief technology officer. His
whole professional life has been recommended system. So he was
a lead on Spotify's song recommenders system US. So if
(27:13):
you like your wrapped playlist or recommended playlist, like, that's
his code. And then he also worked on self driving
cars at Nvidia. But he's been in this world of like,
hey you like these things, what about this thing? Or
here's the inputs that the system is getting. What do
I do nowt so really really deep into that world.
Then we're kind of bringing that spirit, that mindset to
(27:34):
sent into fragrance.
Speaker 1 (27:35):
And then what about beyond you know, for the parts
of your work that are the next steps that you
alluded to farther in the distance, the essentially sensing right,
sensing for security, sensing for health, Like what work are
you doing now toward that end?
Speaker 2 (27:53):
Yeah? So we're incubating this right now. So I'll tell
you two things. So one is, we have a partner,
We've deployed sensors out in the field. We're detecting inauthentic
or counterfeit goods. It's working. What's really the second thing
I'll say is it's we've learned something really interesting, which
is the molecules that smell really good and fruits and
(28:13):
flowers and vegetables that we have to understand to create
fragrance are the same molecules in counter for luxury goods
and the same molecules in our scent. And by getting
really good at understanding and designing fragrance in one domain
in the fragrance industry, we're actually strengthening this platform that
we're building to get really good at the next frontiers
(28:35):
of security detection. And then ultimately, what we care about
is healthy. So that's what really surprised us as I
thought that by working in fragrance, we're making a trade off,
which is we're here to build a great business to
make ourselves resilient so that we can work on the
much longer haul problems. But in reality, we're making progress
(28:56):
on those problems by teaching our platform about what the
world smells like. And it's all one it's just scent,
it's just molecules in the air. And so the more
we learn about really any piece of what the world
smells like, the better we get at all of it.
I think I'll tell you what I think the big
technical frontier is is predicting emotion. Ah, that's interesting. Uh huh.
(29:18):
So when you smell something, you obviously perceive something like
the first thought or first perception is whatever, fresh cut
grass or grapefruit. But then there's another thing that happens
almost at the same time, which is I remember or
I feel a particular thing. And predicting that is something
I don't think anybody's really figured out. But is a
(29:41):
beautiful frontier. Well, how do you get the data? You
got to ask a lot of people how they feel,
what they smell a lot of things, and they have
to be able to articulate it. Right.
Speaker 1 (29:49):
Part of the thing with scent is it's so primordial that, like,
you might not even be able to say how you feel,
so you need in the computer interface.
Speaker 2 (30:00):
You might you might, But turns out we have voices
and faces that are effectively BCIs there's a lot of
information that leaks out of us all the time. And
that was what my PhD was in, is how do
you interpret body language in a way that makes sense?
And by the way, the body language I worked on
most closely was body language driven by odors, right, things
(30:20):
that make I studied this in animals, but makes animals
happy or sad or afraid or calm, And you can
read that out. I mean, our behaviors are meant to
communicate to other animals. Right, We're very social, we're social species.
So I think there's more fundamentals that we have to
figure out. But this is I think there's some really
fundamental stuff that's still unknown here.
Speaker 1 (30:43):
I heard you say in another interview that you worry
sometimes that you'll hit some barrier in nature to your work,
and you said it in passing. But I was very
curious about that.
Speaker 2 (30:54):
What does that mean? I always think about that, which
is like, what day will it be when mother nature
says you can't figure the next hard thing out? And
I just look at this from the history of science.
You know, how if somebody cared about how the planets
were moving in twelve hundred, well good luck, Like you
(31:16):
don't have the right telescopes, you don't have tycho brahe.
There's a bunch of stuff you're gonna need, right, And
so in a way, it's like mother nature and what
our society and species knows conspiring together that basically says
progress will have to wait. And so I think about that.
I worry about that all the time. And so my
mental framework that keeps me super humble is like, I'm
(31:38):
just thankful for all the progress we've been able to make.
That the tools were around right. So I didn't invent
graph neural networks. I didn't even invent the data sets
like we are piecing together and curating, cobbling together. All
these were standing on the shoulders of so many people
and it's just always been the case, and I don't know,
it just it makes Maybe this is too philosophical, but
(32:03):
for me, when I've been up close and personal with
scientific progress, either that I've had a part in or
I've observed other people do, it all feels so tenuous.
It feels so lucky because once you really dig into
the details, you realize, oh my gosh, they had to
be right there at that time and have known about
that thing.
Speaker 1 (32:20):
It's amazing that anything happens, whether you think of how
confut is everything.
Speaker 2 (32:24):
The amazing that anything happens, and you know, when you
really dig in, you're like, wow, how does anything good
happen at all? But nonetheless you persist. And also I
think you can create the conditions where it's more likely
than not to happen. And so that's what OZMO is,
and that's why OSMO Birth Generation is, like, let's create
an environment where we're much more likely than not to
(32:45):
make both the scientific progress we need to make, but
also like really help and change the fragrance industry, which,
by the way, will teach us the things we need
to know to get to the next thing. So I
think there's so much beauty to create in the fragrance
industry that I'm going to just enjoy the heck out
of it and do it for the rest of my life.
But I think it's going to teach us things that
(33:05):
will allow us to do even more audacious work in
the future.
Speaker 1 (33:13):
We'll be back in a minute with the lightning round.
M let's finish with the lightning round. I'm gonna ask you.
Speaker 2 (33:27):
A bunch of questions.
Speaker 1 (33:28):
Now, what seemingly pleasant scent do you never want to
smell again.
Speaker 2 (33:34):
Seemingly pleasant scent that I never want to smell again.
Artificial cherry. It was the cough syrup that I was
forced to drink as a kid, and I'm super sensitive
to it. The molecules ethyl moltole do not like, are
you wearing fragrance right now? And if so, what is it?
I am not. I stopped wearing as soon as I
started the company because I needed to smell, of course,
(33:55):
but like, what's your what's your? Well, give me give
me a pick. Name some fragrance that's that you love
for some reason. So I really like this is kind
of a basic choice from folks inside the industry. I
love Terrator Maz. It's like the RMEZ flagship men's fragrance.
It's by a perfumer, Jean Claude Eleena. I really love
his work.
Speaker 1 (34:15):
Basic Is that like basic in the way of saying
it's like if I asked you for a watch and
you said a Rolex Submarine or something.
Speaker 2 (34:21):
It's just like exactly or saying, like what pop music
do you like? You said Taylor Swift. People like it
because it's great, Huh, Taylor Swift is great? A Rolex
watch is a great watch. Terrtormnz is a great fragrance,
but it's very popular. What is it about it that
you love? I love it's minimalism and I just happen
to like the notes, right, So it's really heavy on
a molecule. I like iso be super. I think it's
(34:42):
a great highlight of that ingredient and it just wears
really well on my skin. So that was what I
used to wear almost every day before I stopped. What's
your second favorite sense? My second favorite sense is probably
gonna be It's a hard between vision and hearing because
I love music, but I like looking at stuff too,
(35:03):
Like the World World of Beautiful are more expensive. Perfumes
actually better sometimes, right, So I think there's just like
anything like bicycles or art, as you start to pay more,
everything gets better.
Speaker 1 (35:18):
And then at Plato's right, what's the worst thing you
ever smelled?
Speaker 2 (35:22):
I have a memory. I picked a mushroom that I
thought looked cool and wanted to show it to my
dad when I was young, and I forgot about it
and it was just turned completely gross.
Speaker 1 (35:35):
I had a version of that, of bringing shells home
from the beach that were alive.
Speaker 2 (35:39):
It turned out I found out when they were dead.
It's like great intentions, but didn't really have wherewithal to
thick that through or understand the consequences.
Speaker 1 (35:54):
Alex Wiltscow is the co founder and CEO of OSMO.
Today's show was produced by Gabriel Hunter Chang. It was
edited by Lyddya jen Kott and engineered by Sarah Bruguer.
You can email us at problem at Pushkin dot FM.
I'm Jacob Goldstein, and we'll be back next week with
another episode of What's Your Problem.
Pushkin, tell me what we don't know about how smell works?
Oh jeez, be sure to tell you what we do.
This is Alex Wilscoe. He's the co founder and CEO
of a company called Osmo, and despite his protests there
he did tell me some of the things nobody knows
(00:37):
about how smell works.
Speaker 2 (00:40):
Why do things smell the way that they do. Why
can we smell certain things and not other things. What
is the logic of how molecules are combined to create
beautiful smells? Why do some smells create incredibly powerful emotional
associations instantly and others seem neutral? Right? Why do something
(01:01):
smell different to people? I think we have a hints
in all these directions, but we have nothing like musical scale,
where we have nothing like a periodic table. We don't
know any structure to why things are the way that
they are. It's a ton of mystery, and that's what
makes it so exciting to work on this topic, is
like there's so much we don't know, and to.
Speaker 1 (01:22):
Be clear, like with light, we just know whatever. If
you tell me the frequency the wavelength, I can know
exactly what color you're talking about, or the same thing
with a wave form of sound, right, and so but
if I give you some random molecule and say what
does it smell like?
Speaker 2 (01:40):
Do you know? So that's what I've spent a lot
of my professional life working on. It is exactly that question. Yeah,
which is a draw structure of a molecule on a whiteboard,
point out it and say, hey, what does this smell
like wood or flowers or fruits or whatever? And so
there is no way to know that for sure at all.
(02:01):
But there's no good way even statistically to predict that
without using large data sets, and at least in our hands.
You need neural networks, you need deep sorting in order
to do that. I'm Jacob Goldstein, and this is what's
your problem?
Speaker 1 (02:20):
The show where I talk to people who are trying
to make technological progress.
Speaker 2 (02:24):
Alex Wilch.
Speaker 1 (02:25):
Goo's problem is this, Can you use AI to teach
computers to smell? And once you figured out how to
do that, can you build a profitable business around it?
Osmo's fun out of Google in twenty twenty three. The
company recently launched a fragrance house to develop new perfumes.
They've also done some work using scent to detect counterfeit shoes,
(02:47):
and in the long run, they plan to use scent
to diagnose disease. Before he started Osmo, Alex worked at
Google as an AI researcher. Before that, he got a
PhD at Harvard studying how mice respond to scent. But
maybe the most important part of his bio came even
earlier in his life, specifically when he was twelve years
(03:09):
old and went off to summer camp in his home state, Texas.
Speaker 2 (03:13):
I was from a small town college station, and then
most of the kids were from big towns like Houston
and Dallas and Austin and San Antonio, and I hadn't
really been exposed to, like I don't know fashion trends
or you know, what was cool or popular. But everybody's
all lumped together in summer camp. And then there is
(03:34):
this thing called perfume that some of the richer, frankly
richer and more popular kids had, And it was just
amazing to me that these boys could spray themselves with
this invisible mist, a clear mist, and then for the
next four to six hours people around them would treat
them differently, and that just blew my mind right, like
(03:59):
there's no I can't I can see the clothes. Yeah,
I can see how they act and walk and talk
and how they you know, posture and all that, but
I cannot see the fragrance. Yeah, but yet it is
obviously doing something magical. It's like an.
Speaker 1 (04:13):
Axe body spray ad what what does that cause you
to do?
Speaker 2 (04:18):
When I get home, beg my parents to buy and fail.
We shopped at TJ Max and I started to really
look out for fragrances there and then it just kind
of snowballs from there, which I just realized there was
like a whole lot of these things. I guess what,
you can just try them, and some of them are
(04:38):
actually way better and more opinionated and more beautiful. I
don't I didn't have the vocabulary then, but like it
just it was clear to me early on that like
I never really thought about who made the clothes, but
I started to think about who made these perfumes, because
it was clear that there were choices that were being made.
And like I just remember trying, and this is years later,
(04:59):
trying Bulgari black, which really kind of clued me into
this world. Bulgary black is not necessarily a great fragrance,
but you can experience the top, middle and base notes
in like forty minutes. Forty five minutes is pretty short,
and so like a bigger fragrance, like a creed, events
will last on your skin for a day, and so
the whole fragrance unfolds. I mean, top notes will last
(05:22):
max Fifteen thirty minutes, but the heart might last for
several hours, and the base note might last for ten hours. Right,
so it smells different.
Speaker 1 (05:30):
You can still smell it, but it smells different whatever
an hour after.
Speaker 2 (05:33):
You put it on and four hours after that, because
a great fragrance is actually many different fragrances within it. Yeah, Right,
There's the first one which peels off or that quickly
burns off quickly. There's the second fragrance, which is the
heart note, which will last for you know, sometimes hours,
but in this case, like another twenty minutes. And then
the base note, which is a third fragrance, which is
what's left after those two burn off. And it's like
(05:55):
three acts of a movie, right, I think it's quite beautiful.
So how do we.
Speaker 1 (05:58):
Get from you being a teenager preoccupied with fragrance to
you using AI to predict how malls will smell?
Speaker 2 (06:10):
Yeah, Like the computer part was always different from the
fragrance part I just I love computers. We always had
computers at home. I started programming around I don't know,
eight years old. It was my life, Like, my entire
life was computers for a long time, still lose in
a way, and fragrance was not a part of it.
I got into you know, statistics, which became machine learning
(06:31):
around the same time. Again for totally independent reason. There's
this thing called the Netflix Prize. It was like one
of the first competitions to build great mL algorithms. I competed.
Speaker 1 (06:40):
Now, that's basically to tell me what else I'll like
on Netflix, right, That's what that contaest was. Like, if
I've watched whatever, if I watched Succession and the Sopranos,
what should I watch next?
Speaker 2 (06:51):
Then you're gonna like another kind of dark, but you know, funny,
kind of soap opera type of a thing, exactly. And
so Netflix did a really bold thing, which is they
released a data set and said, here's what good looks like,
here's how we measure it. Have at it. And they
paid a million dollars to the Winter, which was a
combination of a few teams. But what they really did
is they brought a particular kind of machine learning into
(07:13):
the forefront called collaborative filtering. Really showed that this stuff worked,
and by the way, other companies were already racing to
use this, So like this recommended systems was a big thing,
but Netflix was putting it out into the public and
allowed a kid like me I think I was eighteen
or nineteen years old to actually compete and pretty well
in that. And so I just got exposed to this
world through that and it was super fun. I mean
(07:36):
they gamified it and had I had a blast. So
that was my first exposure to machine learning.
Speaker 1 (07:40):
Turned out to be a good time to start working
on machine learning, Yeah.
Speaker 2 (07:44):
Totally, because if I had started now, they wouldn't let
me in because that's probably ten years ago. Yeah, ten
years ago. Yeah. And then, you know, I was doing
my undergraduate training in neuroscience, and I was studying more
behavior than old action because it actually turned out that
olf action was a hyper specialized sub field of neuroscience.
(08:05):
I didn't realize how niche it was. I loved smell
and I was doing neuroscience, and I knew I wanted
to do smell neuroscience the fancy word for nactual factory neuroscience.
And so there's really two universities in the world that
like have a critical mass of these researchers. It's Columbia
and it's Harvard, and I applied to both. I went
to Harvard, and I realized nobody cares about this problem.
(08:28):
Nobody cares about why molecules smell the way that they do.
There's a much longer conversation as so why that's the
case and why that's still persistent. Now that's changing. Well,
let me ask you this. Let me ask you this.
Speaker 1 (08:38):
At that time, I mean I get it as a
basic research question.
Speaker 2 (08:42):
I mean, I'll tell you.
Speaker 1 (08:43):
We was talking with the producer and editor of this show,
and we were getting ready for this interview, and we
had this interesting conversation talking about scent and what you're
working on whatever. And then I went down and I
saw my daughter, and so, what are you organized? Said
this guy who's trying to figure out scent and teach
computers to smell?
Speaker 2 (08:58):
And she said why, I said, I don't know.
Speaker 1 (09:01):
I ask, so why was it compelling to I get
it as a basic research question, but at that time
was like, there were there applications that came to your mind.
Speaker 3 (09:12):
Look, the the steps of development of this thing that's
now OSMO went through those different iterations.
Speaker 2 (09:23):
You know, I started as an academic scientist and I
was trained in that world, and then I left and
it did some entrepreneurship, but it ended up in industrial research,
and they're like being curious, frankly was enough, and the
idea this is this is Google, this is now a
Google Brain. Yeah, and there's a few steps in between.
Speaker 1 (09:41):
But basically as you're an AI researcher at Google at.
Speaker 2 (09:44):
This moment when you're doing industrial research, right, yeah, exactly,
and Google Brain at the time now it's Google Deep
Mind very much had like a thousand flowers bloom mentality,
and so people were working on crazy stuff, including me
working on something Bell Labs.
Speaker 1 (09:56):
It's basically like Bell Labs of the twenty first century, right, you.
Speaker 2 (09:59):
Have it, exactly, Bell Labs, Erox Park, that kind of
avid and it truly was dreamy, sounds dreamy. It was awesome, right,
And it was also a moment in time, and now
I think that moment's going for better or for worse.
The idea was pretty straightforward for Google, which was, are
the products at Google know what the world looks like
(10:19):
and know what the world sounds like, and that's useful. Right,
that's information that Google's organizing. If we knew what things
smelled and tasted like, that would be useful, right. Uh huh.
Speaker 1 (10:28):
The original mission of Google is organized the world's information, right.
Speaker 2 (10:32):
Exactly, and make it universally accessible and useful. And there
was a whole slice of reality, huh, the chemical slice
of reality that was invisible, right, not just to Google
but two computers. Yeah. Yeah, and that felt really important,
and we had agreement and buying all the way up
to the executive level. They're like, yeah, let's go, let's
go look at that.
Speaker 1 (10:50):
So you're doing a like basic AI research at Google
and you decide to see if you can basically use
AI to figure out scent, to say, here's a molecule,
what does it smell like?
Speaker 2 (11:02):
Right, that's the basic endeavor. How do you do that?
What is it that you actually do? Yeah, so first
it starts to innovation, which I was like, let's figure
out smell. But it actually was a lot more natural
than I think it sounds, which is scent is just chemistry.
It's molecules, and we got to do AI for molecules. Right.
(11:22):
If we're going to do AI for scent, and the
thing that had happened in between. You know, there's a
five year period between my academic life and my industrial life,
and what had happened in those five years is actually
some of the people I did my PhD with and
then some of the people I ended up working with
at Google Brain really cracked machine learning or AI for molecules.
(11:45):
But they didn't do it for scent. They did it
for a few other things. They did for drug discovery,
and they did it for like materials discovery, so like
new materials for LEDs.
Speaker 1 (11:54):
Right, So you happen to be doing essentially basic research
at Google at this moment when there is this new
way to use AI that is well suited to molecules,
and you say, we can do let's do it, Yeah,
let's do it.
Speaker 2 (12:11):
Yeah, we can do it. The other pieces are great.
You got the algorithm. Where's the data? Classic? That's the
classic AI question, right, like, exactly where's the data? What
I did know just from being obsessed and in this
world for a long time prior to that, there were
these collections of data sets that were honestly really more
like magazine catalogs for fragrance ingredients, and so there were
(12:33):
these catalogs basically saying this is the ingredient, this is
the molecular structure of this ingredient, and here's what it
smells like. And by the way, the rating of what
it smells like was done by a professional, by a perfumer.
And so the special sauce that we added is we
we went and we got that data and we fused
a few data sets together, and we cleaned it very carefully,
(12:54):
and that that hadn't been done.
Speaker 1 (12:55):
And it's something it's like five thousand dish right, it's
five thousand or so different molecules.
Speaker 2 (13:00):
Okay, yep, exactly. And here is this the one with
the list. I love the list. Here I have it.
Speaker 1 (13:06):
This the one sweet fruity, vanilla, powdery, fluoral, barry, fermented,
nutty ozone, buttery musk.
Speaker 2 (13:12):
It's it's that lispright. Those are they? And there's one
hundred and thirty eight of those descriptors I think that
we used in that data set. Sometimes we use smaller subsets,
but the full set originally is about one hundred and forty.
Speaker 1 (13:24):
So okay, so you have your whatever, your five thousand
molecules labeled with one hundred and forty different sets. You
train your AI model on this data set, and then
you want to find out does the model work?
Speaker 2 (13:38):
Does the AI work right?
Speaker 1 (13:39):
If I give the model some new molecule molecule that
wasn't in the training data, will it know what that
molecule smells like? And to test that to answer that question,
you actually.
Speaker 2 (13:50):
Do this study.
Speaker 1 (13:51):
So you get a bunch of people to smell these
molecules that are not that your model was not trained on,
essentially right, and say what it smells like? It's weird
like what? You don't actually care what it fundamentally smells like.
You just care what everybody on average thinks it smells.
Speaker 2 (14:05):
Like, because guess what, that's what it's what smell is? Yeah,
that's what do you think of smell? Yeah? Ye?
Speaker 1 (14:11):
So you asked this panel to what do all these
molecule smell like? And then you ask the model what
do they smell like? And you compare the results and
how does the model do?
Speaker 2 (14:23):
That was really the threshold of breakthrough in my mind
was like are you worse than a person? Yeah? Or
are you slightly better than a person? And we got
slightly better than a person, which was a breakthrough in
my view.
Speaker 1 (14:36):
Right, and so yes, so that paper you published in
Science and you started Ozmo kind of around the same time, Right,
you started that study at at Google, is that right?
And then by the time it was published you had
spun Osmo out of Google, right, that's right. So you
have this map, you have this model that can basically
given a molecule, predict pretty well what the average person
(14:58):
thinks that molecule smell like. But there's still a second problem, right,
which is, in the world, in the wild, you don't
know what molecules are in the air. You don't you
know what molecules somebody's smelling. And so for that second problem,
you need to try and build some kind of automated
system for figuring out what molecules are in the air
(15:18):
at a given that's correct.
Speaker 2 (15:20):
Getting to one molecule structure is actually not trivial. So
to go from a physical thing and know all the
molecular structures like not a solveable. So there's a lot
of ways to do that. There's a lot of chemical
sensors out there, none of them will just tell you
the formula, right, So that's hard, really hard.
Speaker 1 (15:41):
So there's like a chemistry problem of like isolating the
molecule basically and deriving the chemical formula.
Speaker 2 (15:49):
Exactly taking a real smell and it's composed of a
bunch of different molecules with different structures, and there's different amounts,
there's ratios. You got to get that recipe out of
the air. So that's on. That's hard. That was unsolved
at the time to do it in an automated way.
And by the way, if we're following this story chronologically,
we hadn't done this yet, agah, but we knew we
(16:10):
had to do that, right, So we knew that, Okay,
if we wanted to actually digitize the world of scent
and have a record of what the world smelled like
and maybe even replay it, we're going to have to
do this. We needed to automate that and have it
be automatic, and that's what we did.
Speaker 1 (16:24):
So basically, you can put any smell into the machine
and it'll tell you what it's made of. At this point,
oh yeah, so you're setting out to start OSMO, Like,
what are you thinking of in terms of the set
of potential commercial applications.
Speaker 2 (16:39):
So we really had we had three in mind, and
there's still very much present in mind the focuses has
become a lot crisper though in terms of what we're
concentrating on. We know the fragrance industry is huge and
very profitable, and it's also something I personally love. That's
(17:02):
a thing we want to automate and understand. And then
we know that dogs can detect things, right, and so
we know dogs can detect harmful substances like drugs or bombs,
or things that just shouldn't be there, like produce where
it shouldn't be being shipped. And then we also know that
dogs and even in some cases people can detect health
(17:22):
or disease states. Right. We know that missus Milner, a
nurse in the UK, was able to smell Parkinson's disease
and she's since been able to teach that skill to
other people, which is really amazing. And then we figured
out all the chemistry of what's actually being smelled. We
know that there's many many instances where there is a
(17:43):
scent signature to a disease or to a wellness or
to a health state that hasn't yet been fully figured out, right,
but we know that they exist. Those are the three, right,
So fragrance industry really security and supply chain and health
and wellness, and I view them in that order because
(18:03):
that's like the order in which I think we can
be useful to the world. Right, Designing fragrances is something
that's much more attainable technically, and frankly, it's just a great,
much faster sales cycle to be business to be in
than ultimately diagnostics, which are so hard. Right, I mean,
it is my north star. It's like where I want
to take the company. But I also have no illusions
(18:25):
about how hard that is, and I just I've seen
all the failures of the companies that have attempted it,
and I think I've learned from what hasn't worked, and
so I'm incorporating those learnings into how I want to
build the company, which is, build a great business in fragrance,
build beautiful fragrances for the world, and then strike out
from that position of strength and to even more ambitious frontiers.
(18:49):
We'll be back in just a minute.
Speaker 1 (19:00):
When I first heard about the work Alex was doing
at OSMO, I understood why it would be useful for sensing. Basically,
you might be able to build automate sniffing machines that
could say detect cancer in a person or detect a
bomb in a suitcase. But I couldn't figure out truly
what the business case was for perfume. And in fact
(19:22):
Osmo has recently launched a perfume business. It's called Generation.
So I asked Alex, why is using AI and fancy machines?
Why is that better than just designing perfume in the
traditional way.
Speaker 2 (19:36):
We can go from the first kind of client demand. So, hey,
I want to create a fragrance, and here's who my
brand is, here's what I want to do. So just
that description to a starting place of a fragrance in
a minute or two.
Speaker 1 (19:50):
What happens at a traditional perfumer when somebody comes in
with that request.
Speaker 2 (19:54):
Well, so let's say you're an emerging Let's say you're
an emerging brand, right, So you're starting out or you
have your first product and you want to add a
second one. But you're small, right, You're not making a
billion dollars in revenue. You're making less than that. So
if you want to make a new custom fragrance, good luck. Right,
you're not going to be able to get the attention
of the big fragrance houses because they want a service
(20:15):
business that's like millions and millions of dollars and you're
not big enough yet. So if you want a great
custom fragrance that your consumers are going to love, and
you want to do it quickly so you're responding to trends,
you aren't going to be able to get it done.
So you have to make compromises, right, So if you
(20:35):
want to move fast, you're gonna have to use a
regurgitated fragrance. It's also called a library fragrance, which means
somebody else in the market has your spell.
Speaker 1 (20:42):
I'm imagining that people who sell it call it a
library fragrance rather than a regurgitated library fragrance.
Speaker 2 (20:47):
They do, they don't say regurgitated, but that's what it
effectively is, right fair.
Speaker 1 (20:51):
Regurgitated does have a particular old factory connotation, so it's
a level.
Speaker 2 (20:57):
It's vic or all sticks in the mind.
Speaker 1 (20:59):
What like, And I'm not I just genuinely don't understand, Like,
why can't somebody just have a company with a bunch
like who knows the molecules? You know, who knows what
the five thousand milecules in the book smell like? Because
they've got the book and they can just use the
book and be like, oh, you want this, let's try that.
Speaker 2 (21:15):
Do you know what I mean? Like, I'm not trying
to be difficult, but I.
Speaker 1 (21:18):
Genuinely don't understand why you need the technology to do that.
Speaker 2 (21:24):
Yeah, I genuinely didn't understand this either. And there's there's
a class of professional called a perfumer, and their job
is to do what you're describing, which is, Hey, I
know all these ingredients and I'm going to make them
in order to create your fragrance. So they typically there's
no perfumer that knows five thousand ingredients, but the best
perfumers know a thousand or two thousand ingredients. Most perfumers
(21:48):
work with two hundred, one hundred, two hundred ingredients. So
already there, like where we're there is very few people
in the world that can do what you're saying. Yeah,
they can do, and then how what are they going
to work on? Right? So it might take them weeks
or months to create a fragrance. They're working on a
few at a time. Why would they work on an
(22:10):
emerging brands fragrance when they can go work on a
much larger account. So there's just very very limited number
of people who can engage in the fragrance creation process,
because it is difficult. It's not so much identifying, Hey,
all these molecules smell this particular way, and therefore I
should be able to mix them, Like what ratios do
you mix them in? Like? What are the rules? Right?
And now you're actually getting into designing a system which
(22:31):
understands sent well enough to create new fragrance formulas, is
starting places, and then of course a perfumer finished system.
But you're right, it's like, oh, why shouldn't that exist?
And then when you actually start to peel back the
layers one by one, you realize, oh, you actually have
to build what we built. It's actually in order to
answer that question.
Speaker 1 (22:46):
So presumably now your model cannot only predict what one
molecule is going to smell like, but it can predict
the combination of molecules. I mean, is it predicting? Does
it know concentration? Like does it know oh yeah, yeah,
how good is it?
Speaker 2 (23:02):
I mean you have a perfumer on staff? Why? Well,
I think the goal of tools is to have them
in the hands of creatives. And there's many steps to perfumery,
but I think there's three that are relevant for what
we're talking about. The first is a perfumer when they're
(23:22):
when they're starting on a project, they have to have
a starting place, they have a starting formula, and then
they do their creative work step two to evolve that
formula to exactly what the customer wants, to a creative
expression that the lights the consumer as well. That's the
funnest part. Perfumers love that that is actual creation in
the creative part. Number three is then it has to
(23:43):
be the right price. It has to be compliant with
with a regulatory compliance. There cannot be allergens, all that
stuff that's more like sound engineering than it is composition
or being a rock star. Steps one, the starting place,
step three, all the regulatory requirements. That's where we spend
the most energy in building these tools. And then a
perfumer is the person that is taking the formulas from
(24:06):
starting place to creative endpoint and then handing it off
for like regulatory finishing. And they're just way more effective
with these.
Speaker 1 (24:13):
Tools, at least for now, right Like, that's the way
I feel using an LLM, like I feel like I
have a window and me Plus the LLM is better
than the LLM alone and we haven't. That window hasn't
closed yet, but I'm not optimistic about my long term prospects.
Speaker 2 (24:29):
We'll see though, I mean, but like listen, I honest
belief here, like the tools will get better, but the
drive to create will never go away. And I think
people will always want to know about the person behind
the creation in a way, and it's not uniforms. So
I don't think people want to know the perfumer behind
the hand soap in the gas station. They just don't,
(24:52):
right it. But there I think will always be room
for craft and creative use of tools, and the profession
that uses those tools might change radically, in the industry
in which those tools are used might change radically, but
the tools will always be wielded by people, but the
work that's being done might be unrecognizable. So you know,
(25:15):
we'll see how the world evolves. But like I just
like AI is like an engine, it's just a technologist,
just a tool.
Speaker 1 (25:23):
So what's the what's the business model? Just briefly for generation,
like how what you know?
Speaker 2 (25:29):
What's the model? The business model really simply is we
all design the fragrance for you and then you'll buy
that fragrance to put in your products or will even
actually create the full finished product. We'll put in a
bottle for you if you if you want. We are
behind the scenes. We're an engine supporting brands. We're not
a brand ourselves, and we're here to make beautiful fragrance
(25:50):
products for for brands. So what's the frontier like you have?
Speaker 1 (25:56):
You know, on the business side, the generation is kind
of the central thing you're working on now, but on
the more on the on the research side, like what
are you trying to figure out?
Speaker 2 (26:05):
Now? What are you working on now? So there's there's
our starting place, which is why does this molecule smell
the way that it does? And we can never stop
getting better at that. Then there's the next question of
why does this mixture of molecule smell the way that
it does? And we can never stop getting better at that.
And then there's do you like it? Which is maybe
(26:27):
the most important question from a business perspective, or who
likes it? And in what context? Yeah, exactly exactly, which
is it's not just the formula as the input to
this model, but there's also who you are, what are
your experiences, where are you from, what are the other
things in your life that you've got.
Speaker 1 (26:42):
That actually goes back to your Netflix collaborative. It gets
me like if I watched Succession and the Sopranos and
I'm fifty and.
Speaker 2 (26:51):
Then what's the cologne for me? Yeah? Exactly. And so
I was very fortunate to be able to start this
company with a guy work with at Twitter name. His
name is Rich Witcombe. He's our chief technology officer. His
whole professional life has been recommended system. So he was
a lead on Spotify's song recommenders system US. So if
(27:13):
you like your wrapped playlist or recommended playlist, like, that's
his code. And then he also worked on self driving
cars at Nvidia. But he's been in this world of like,
hey you like these things, what about this thing? Or
here's the inputs that the system is getting. What do
I do nowt so really really deep into that world.
Then we're kind of bringing that spirit, that mindset to
(27:34):
sent into fragrance.
Speaker 1 (27:35):
And then what about beyond you know, for the parts
of your work that are the next steps that you
alluded to farther in the distance, the essentially sensing right,
sensing for security, sensing for health, Like what work are
you doing now toward that end?
Speaker 2 (27:53):
Yeah? So we're incubating this right now. So I'll tell
you two things. So one is, we have a partner,
We've deployed sensors out in the field. We're detecting inauthentic
or counterfeit goods. It's working. What's really the second thing
I'll say is it's we've learned something really interesting, which
is the molecules that smell really good and fruits and
(28:13):
flowers and vegetables that we have to understand to create
fragrance are the same molecules in counter for luxury goods
and the same molecules in our scent. And by getting
really good at understanding and designing fragrance in one domain
in the fragrance industry, we're actually strengthening this platform that
we're building to get really good at the next frontiers
(28:35):
of security detection. And then ultimately, what we care about
is healthy. So that's what really surprised us as I
thought that by working in fragrance, we're making a trade off,
which is we're here to build a great business to
make ourselves resilient so that we can work on the
much longer haul problems. But in reality, we're making progress
(28:56):
on those problems by teaching our platform about what the
world smells like. And it's all one it's just scent,
it's just molecules in the air. And so the more
we learn about really any piece of what the world
smells like, the better we get at all of it.
I think I'll tell you what I think the big
technical frontier is is predicting emotion. Ah, that's interesting. Uh huh.
(29:18):
So when you smell something, you obviously perceive something like
the first thought or first perception is whatever, fresh cut
grass or grapefruit. But then there's another thing that happens
almost at the same time, which is I remember or
I feel a particular thing. And predicting that is something
I don't think anybody's really figured out. But is a
(29:41):
beautiful frontier. Well, how do you get the data? You
got to ask a lot of people how they feel,
what they smell a lot of things, and they have
to be able to articulate it. Right.
Speaker 1 (29:49):
Part of the thing with scent is it's so primordial that, like,
you might not even be able to say how you feel,
so you need in the computer interface.
Speaker 2 (30:00):
You might you might, But turns out we have voices
and faces that are effectively BCIs there's a lot of
information that leaks out of us all the time. And
that was what my PhD was in, is how do
you interpret body language in a way that makes sense?
And by the way, the body language I worked on
most closely was body language driven by odors, right, things
(30:20):
that make I studied this in animals, but makes animals
happy or sad or afraid or calm, And you can
read that out. I mean, our behaviors are meant to
communicate to other animals. Right, We're very social, we're social species.
So I think there's more fundamentals that we have to
figure out. But this is I think there's some really
fundamental stuff that's still unknown here.
Speaker 1 (30:43):
I heard you say in another interview that you worry
sometimes that you'll hit some barrier in nature to your work,
and you said it in passing. But I was very
curious about that.
Speaker 2 (30:54):
What does that mean? I always think about that, which
is like, what day will it be when mother nature
says you can't figure the next hard thing out? And
I just look at this from the history of science.
You know, how if somebody cared about how the planets
were moving in twelve hundred, well good luck, Like you
(31:16):
don't have the right telescopes, you don't have tycho brahe.
There's a bunch of stuff you're gonna need, right, And
so in a way, it's like mother nature and what
our society and species knows conspiring together that basically says
progress will have to wait. And so I think about that.
I worry about that all the time. And so my
mental framework that keeps me super humble is like, I'm
(31:38):
just thankful for all the progress we've been able to make.
That the tools were around right. So I didn't invent
graph neural networks. I didn't even invent the data sets
like we are piecing together and curating, cobbling together. All
these were standing on the shoulders of so many people
and it's just always been the case, and I don't know,
it just it makes Maybe this is too philosophical, but
(32:03):
for me, when I've been up close and personal with
scientific progress, either that I've had a part in or
I've observed other people do, it all feels so tenuous.
It feels so lucky because once you really dig into
the details, you realize, oh my gosh, they had to
be right there at that time and have known about
that thing.
Speaker 1 (32:20):
It's amazing that anything happens, whether you think of how
confut is everything.
Speaker 2 (32:24):
The amazing that anything happens, and you know, when you
really dig in, you're like, wow, how does anything good
happen at all? But nonetheless you persist. And also I
think you can create the conditions where it's more likely
than not to happen. And so that's what OZMO is,
and that's why OSMO Birth Generation is, like, let's create
an environment where we're much more likely than not to
(32:45):
make both the scientific progress we need to make, but
also like really help and change the fragrance industry, which,
by the way, will teach us the things we need
to know to get to the next thing. So I
think there's so much beauty to create in the fragrance
industry that I'm going to just enjoy the heck out
of it and do it for the rest of my life.
But I think it's going to teach us things that
(33:05):
will allow us to do even more audacious work in
the future.
Speaker 1 (33:13):
We'll be back in a minute with the lightning round.
M let's finish with the lightning round. I'm gonna ask you.
Speaker 2 (33:27):
A bunch of questions.
Speaker 1 (33:28):
Now, what seemingly pleasant scent do you never want to
smell again.
Speaker 2 (33:34):
Seemingly pleasant scent that I never want to smell again.
Artificial cherry. It was the cough syrup that I was
forced to drink as a kid, and I'm super sensitive
to it. The molecules ethyl moltole do not like, are
you wearing fragrance right now? And if so, what is it?
I am not. I stopped wearing as soon as I
started the company because I needed to smell, of course,
(33:55):
but like, what's your what's your? Well, give me give
me a pick. Name some fragrance that's that you love
for some reason. So I really like this is kind
of a basic choice from folks inside the industry. I
love Terrator Maz. It's like the RMEZ flagship men's fragrance.
It's by a perfumer, Jean Claude Eleena. I really love
his work.
Speaker 1 (34:15):
Basic Is that like basic in the way of saying
it's like if I asked you for a watch and
you said a Rolex Submarine or something.
Speaker 2 (34:21):
It's just like exactly or saying, like what pop music
do you like? You said Taylor Swift. People like it
because it's great, Huh, Taylor Swift is great? A Rolex
watch is a great watch. Terrtormnz is a great fragrance,
but it's very popular. What is it about it that
you love? I love it's minimalism and I just happen
to like the notes, right, So it's really heavy on
a molecule. I like iso be super. I think it's
(34:42):
a great highlight of that ingredient and it just wears
really well on my skin. So that was what I
used to wear almost every day before I stopped. What's
your second favorite sense? My second favorite sense is probably
gonna be It's a hard between vision and hearing because
I love music, but I like looking at stuff too,
(35:03):
Like the World World of Beautiful are more expensive. Perfumes
actually better sometimes, right, So I think there's just like
anything like bicycles or art, as you start to pay more,
everything gets better.
Speaker 1 (35:18):
And then at Plato's right, what's the worst thing you
ever smelled?
Speaker 2 (35:22):
I have a memory. I picked a mushroom that I
thought looked cool and wanted to show it to my
dad when I was young, and I forgot about it
and it was just turned completely gross.
Speaker 1 (35:35):
I had a version of that, of bringing shells home
from the beach that were alive.
Speaker 2 (35:39):
It turned out I found out when they were dead.
It's like great intentions, but didn't really have wherewithal to
thick that through or understand the consequences.
Speaker 1 (35:54):
Alex Wiltscow is the co founder and CEO of OSMO.
Today's show was produced by Gabriel Hunter Chang. It was
edited by Lyddya jen Kott and engineered by Sarah Bruguer.
You can email us at problem at Pushkin dot FM.
I'm Jacob Goldstein, and we'll be back next week with
another episode of What's Your Problem.
Popular Podcasts
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
True Crime Tonight
If you eat, sleep, and breathe true crime, TRUE CRIME TONIGHT is serving up your nightly fix. Five nights a week, KT STUDIOS & iHEART RADIO invite listeners to pull up a seat for an unfiltered look at the biggest cases making headlines, celebrity scandals, and the trials everyone is watching. With a mix of expert analysis, hot takes, and listener call-ins, TRUE CRIME TONIGHT goes beyond the headlines to uncover the twists, turns, and unanswered questions that keep us all obsessed—because, at TRUE CRIME TONIGHT, there’s a seat for everyone. Whether breaking down crime scene forensics, scrutinizing serial killers, or debating the most binge-worthy true crime docs, True Crime Tonight is the fresh, fast-paced, and slightly addictive home for true crime lovers.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com