AI transcript
0:00:11 pushkin you’re listening to an iheart podcast
0:00:18 run a business and not thinking about podcasting think again more americans listen to podcasts
0:00:23 then add supported streaming music from spotify and pandora and as the number one podcaster
0:00:27 iheart’s twice as large as the next two combined learn how podcasting can help your business call
0:00:35 844-844-iheart can we invest our way out of the climate crisis five years ago it seemed like
0:00:42 wall street was working on it until a backlash upended everything so there’s a lot of alignment
0:00:49 between the dark money right and the oil industry on this effort i’m amy scott host of how we survive
0:00:56 a podcast from marketplace and this season we investigate the rise fall and reincarnation
0:01:02 of climate conscious investing listen to how we survive wherever you get your podcasts
0:01:12 tell me what we don’t know about how smell works oh geez uh be shorter to tell you what we do
0:01:20 this is alex wilchko he’s the co-founder and ceo of a company called osmo and despite his protests there
0:01:29 he did tell me some of the things nobody knows about how smell works why do things smell the way that they do
0:01:37 why can we smell certain things and not other things what is the logic of how molecules are combined to create
0:01:45 beautiful smells why do some smells create incredibly powerful emotional associations instantly and others seem neutral
0:01:51 right why do something smell different to people right i think we we have a hints in all these directions but we
0:01:59 we have nothing like musical scales where we have nothing like a periodic table we don’t know any structure to why
0:02:05 things are the way that they are it’s a ton of mystery um and that’s what makes it so exciting to work on this
0:02:13 topic it’s like there’s so much we don’t know and to be clear like with light we just know whatever
0:02:19 if you tell me the the frequency the wavelength i can know exactly what color you’re talking about
0:02:25 or the same thing with the with a waveform of sound right and so but if i give you some random molecule
0:02:31 and say what does it smell like do you know so that’s what i’ve spent a lot of my professional
0:02:35 life working on it’s exactly that question yeah which is draw a structure of a molecule on a white
0:02:41 board point at it and say hey is this what does it smell like wood or flowers or fruits or whatever
0:02:49 and so there is no way to know that for sure at all but there’s no good way even statistically
0:02:56 to predict that without using large data sets and at least in our hands i you need neural networks you
0:03:06 need deep learning yeah in order to do that i’m jacob goldstein and this is what’s your problem
0:03:12 the show where i talk to people who are trying to make technological progress alex wiltschko’s problem
0:03:19 is this can you use ai to teach computers to smell and once you’ve figured out how to do that
0:03:26 can you build a profitable business around it osmo spun out of google in 2023 the company
0:03:31 recently launched a fragrance house to develop new perfumes they’ve also done some work using scent
0:03:38 to detect counterfeit shoes and in the long run they plan to use scent to diagnose disease
0:03:46 before he started osmo alex worked at google as an ai researcher before that he got a phd at harvard
0:03:52 studying how mice respond to scent but maybe the most important part of his bio came even earlier in
0:04:00 his life specifically when he was 12 years old and went off to summer camp in his home state texas i was
0:04:06 from a small town college station and then most of the kids were from big towns like houston and dallas and
0:04:15 austin san antonio and i hadn’t really been exposed to like i don’t know fashion trends or you know what
0:04:21 was cool or popular but everybody’s all lumped together in summer camp and then there was this
0:04:28 thing called perfume that some of the richer frankly richer and more popular kids had hello and it was just
0:04:38 amazing to me that these boys could spray themselves with this invisible mist a clear mist and then for
0:04:44 the next four to six hours people around them would treat them differently and that just blew my mind
0:04:50 right like there’s no i can’t i can see the clothes yeah i can see how they act and walk and talk and
0:04:57 how they you know posture and all that but i cannot see the fragrance yeah but yet it is obviously doing
0:05:05 something magical it’s like an axe body spray ad uh what what does that cause you to do when i get home
0:05:16 and fail um i uh we shopped at tj maxx and i started to really look out for fragrances there
0:05:20 and then it just it kind of snowballs from there which i just realized there’s like a whole lot of
0:05:26 these things and guess what you can just try them and some of them are actually way better and more
0:05:31 opinionated and more beautiful i don’t i didn’t have the vocabulary then but like it just it was clear
0:05:36 to me early on that like i never really thought about who made the clothes but i started to think
0:05:41 about who made these perfumes uh-huh because it was clear that there were choices that were being made
0:05:48 and like i just remember trying and this was years later trying bulgari black which really kind of
0:05:54 clued me into this world that bulgari black is not necessarily a great fragrance but you can
0:05:59 experience the top middle and base notes in like 40 minutes 45 minutes it’s pretty short and so like
0:06:05 a bigger fragrance like a creed eventus will last on your skin for a day and so you the the whole
0:06:12 fragrance unfolds i mean top notes will last max 15 30 minutes but the heart might last for several
0:06:17 hours and the base note might last for 10 hours right so it so it smells different you can still
0:06:21 smell it but it smells different whatever an hour after you put it on and four hours after that
0:06:26 because a great fragrance is actually many different fragrances within it yeah right there’s the first one
0:06:30 which peels off really quickly burns off quickly there’s the second fragrance which is the heart
0:06:34 note which will last for you know sometimes hours but in this case like 20 another 20 minutes yeah and
0:06:39 then the the base note which is a third fragrance which is what’s left after those two burn off
0:06:44 um and it’s that’s it’s like three acts of a movie right it’s i think it’s quite beautiful so how do we
0:06:55 get from you being a teenager preoccupied with fragrance to you using ai to predict how molecules
0:07:00 will smell yeah like the computer part was always different from the fragrance part i just i love
0:07:05 computers uh we always had computers at home i started programming when i was around i don’t know
0:07:10 eight years old that was my life like my entire life was computers for a long time still is in a way
0:07:17 yeah um and fragrance was not a part of it i got into you know statistics which became machine
0:07:20 learning around the same time again for totally independent reason there’s this thing called the
0:07:26 netflix prize it was like one of the first competitions to build great ml algorithms i competed
0:07:30 in that i mean that’s basically to tell me what else i’ll like on netflix right that’s what that
0:07:37 contest was like if i’ve watched whatever if i watch succession and the sopranos what should i watch next
0:07:42 then you’re gonna like another kind of dark but you know funny kind of uh soap opera type of a thing
0:07:47 exactly and so netflix did a really bold thing which is they released a data set and said here’s what
0:07:52 good looks like here’s how we measure it how about it and they paid a million dollars to the winner um
0:07:58 which was a combination of a few teams but what they really did is they brought a particular kind of
0:08:04 machine learning into the forefront called collaborative filtering really showed that this stuff worked and by
0:08:10 the way other companies were already racing to use this so like this recommender systems was a big thing
0:08:14 um but netflix was putting it out into the public and allowed a kid like me at i think i was 18 or 19
0:08:20 years old to actually compete and do pretty well in that and so i just got exposed to this world through
0:08:26 that and it was super fun i mean they gamified it and i had a i had blast so that was my first exposure
0:08:31 to machine learning turned out to be a good time to start working on machine learning yeah totally because if i
0:08:36 started now they wouldn’t let me in because everybody’s so smart that was probably 10 years ago yeah 10 years
0:08:44 ago yeah yeah and um then you know i was doing my undergraduate training in neuroscience and i was
0:08:50 studying more behavior uh than olfaction because it actually turned out that olfaction was a hyper
0:08:55 specialized subfield of neuroscience i didn’t realize how niche it was i loved smell and i was doing
0:09:00 neuroscience and i knew i wanted to do smell neuroscience the fancy word for that is olfactory
0:09:06 neuroscience and so there’s really two universities in the world that like have a critical mass of
0:09:12 these researchers it’s columbia and it’s harvard and i applied to both i went to harvard and i realized
0:09:19 nobody cares about this problem nobody cares about why molecules smell the way that they do there’s a much
0:09:23 longer conversation as to why that’s the case and why that’s still persisting how that’s changing
0:09:28 well let me ask you this let me ask you this at that time i mean i i get it as a basic research
0:09:32 question i mean i’ll tell you i we i was talking with the producer and editor of this show and we
0:09:36 were getting ready for this interview and we had this interesting conversation talking about scent and
0:09:41 what you’re working on whatever and i went down uh and i saw my daughter and she said what are you
0:09:45 working on i said this guy who’s trying to figure out scent and teach computers to smell and she said
0:09:53 why i said i don’t know i should ask him that so uh why why was it compelling to you i get it as a basic
0:09:58 research question but at that time was it like was there were there applications that came to your mind
0:10:09 look the the the steps of um development of this thing that’s now osmo went through those different
0:10:15 iterations of you know i started as an academic scientist and i was trained in that world and then
0:10:20 i left and i did some entrepreneurship yeah i ended up in industrial research and they’re like being
0:10:26 curious frankly was enough and the idea this is this is at google this is now a google brain yeah and
0:10:31 there’s a few steps in between but basically as a you’re an ai researcher at google at this moment when
0:10:35 you’re doing industrial research right yeah exactly in google brain at the time now it’s google deep
0:10:40 mind very much had like a thousand flowers bloom mentality and so people were working on crazy
0:10:44 stuff including me working on smell bell labs it’s basically like bell labs of the 21st century right
0:10:52 you have it exactly bell labs xerox park yeah kind of it truly was dreamy sounds dreamy it was awesome
0:10:57 right and it’s also a moment in time and now i think that moment’s gone for better or for worse um
0:11:05 the idea was pretty straightforward for google which was are the products that google know what the world
0:11:09 looks like and know what the world sounds like and that’s useful right that’s information that google’s
0:11:15 organizing if we knew what things smelled and tasted like that would be useful right uh-huh the original
0:11:20 mission of google is organize the world’s information right exactly and make it universally accessible and
0:11:26 useful and there was a whole slice of reality uh-huh the the chemical slice of reality that was
0:11:32 invisible right to not just to google but to computers yeah yeah and that felt really important and we had
0:11:36 agreement and buy-in all the way up to the executive level they’re like yeah let’s go let’s go look at
0:11:43 that so you’re doing a like basic ai research at google and you decide to see if you can basically use ai
0:11:49 to figure out scent to say here is a molecule what does it smell like right that’s the basic
0:11:52 endeavor how do you do that what is it that you actually do
0:11:56 yeah so first it starts the motivation which is like let’s figure out smell
0:12:02 um but it actually was a lot more natural than i think it sounds yeah which is
0:12:09 scent is just chemistry yeah it’s molecules and we got to do ai for molecules right if we’re going
0:12:14 to do ai for scent yeah and the thing that had happened in between you know it was a five-year
0:12:21 period between my academic life and my industrial life yeah and what had happened in those five years
0:12:25 is actually some of the some of the people i did my phd with and then some of the people i ended up
0:12:32 working with at google brain really cracked machine learning or ai for molecules but they didn’t do it for
0:12:37 scent they did it for a few other things they did for drug discovery and they did it for um like
0:12:45 materials discovery so like new materials for leds right so you happen to be doing essentially basic
0:12:52 research at google at this moment when there is this new way to use ai that is well suited to
0:12:59 molecules and you say we can do scent do it yeah let’s do it yeah we can do it the other pieces are
0:13:05 great you got the algorithm where’s the data classic that’s the classic ai question right like exactly
0:13:10 where’s the data what i did know just from being obsessed and in this world for a long time prior
0:13:17 that there were these collections of data sets that were honestly really more like magazine catalogs
0:13:23 for fragrance ingredients and so there were these catalogs basically saying this is the ingredient this
0:13:27 is the molecular structure of this ingredient and here’s what it smells like and by the way
0:13:33 the rating of what it smells like was done by a professional by a perfumer and so the special sauce
0:13:39 that we added is we we went and we got that data and we we fused a few data sets together and we cleaned
0:13:44 it very carefully and that that hadn’t been done and it’s something it’s like 5 000 dish right it’s
0:13:51 5 000 or so different molecules okay yep exactly and here is this the one with the list i love the list
0:13:59 here i have it this the one sweet fruity vanilla powdery floral berry fermented nutty ozone buttery musk
0:14:04 it’s that it’s that list right those are they and there’s 138 of those descriptors i think that we
0:14:10 used in that data set sometimes we use smaller subsets but the full set originally is about 140
0:14:17 so okay so you have your whatever your 5 000 molecules labeled with 140 different cents
0:14:24 you train your ai model on this data set and then you want to find out does the model work does the ai
0:14:31 work right if i give the model some new molecule molecule that wasn’t in the training data will it
0:14:37 know what that molecule smells like and and to test that to answer that question you actually do this study
0:14:42 so you get a bunch of people to smell these molecules that are not that your model was not
0:14:47 trained on essentially right and say what it smells like it’s weird like what you you don’t actually
0:14:52 care what it fundamentally smells like you just care what everybody on average thinks it smells like
0:14:56 because guess what that’s what it’s that’s what smell is yeah that’s the meaning of smell yeah
0:15:03 so you ask this panel to what do all what do all these uh molecules smell like and then you ask the
0:15:10 model what do they smell like and you compare the results and how does the model do um that was really the
0:15:17 threshold of breakthrough in my mind was like are you worse than a person yeah or are you slightly better
0:15:23 than a person and we got slightly better than a person which was a breakthrough in my view right and so
0:15:29 yes so that paper you published in science and and you started osmo kind of around the same time right
0:15:34 you started that study at at google is that right and then by the time it was published you had spun
0:15:40 osmo out of google right that’s right so you you have this map you have this model that can basically
0:15:46 given a molecule predict pretty well what the average person thinks that molecule will smell like
0:15:53 but there is still a second problem right which is in the world in the wild you don’t know
0:15:58 what molecules are in the air you don’t know what molecule somebody’s smelling and so for that
0:16:04 second problem you need to try and build some kind of automated system for figuring out what molecules
0:16:11 are in the air at a given that’s correct getting to one molecule structure is actually not trivial
0:16:16 so to go from a physical thing and know all the molecular structures like not a solid problem
0:16:22 so there’s a lot of ways to do that there’s a lot of chemical sensors out there none of them will just
0:16:30 tell you the formula right so that’s hard really hard so there’s like a chemistry problem of like
0:16:37 isolating the molecule basically and and deriving the chemical formula exactly taking a real smell
0:16:40 and it’s composed of a bunch of different molecules with different structures and there’s different
0:16:47 amounts there’s ratios yeah you got to get that recipe out of the air so that’s on that’s hard
0:16:53 that was unsolved at the time to do in an automated way and by the way if we’re following this story
0:16:58 chronologically we hadn’t done this yet at google yeah but we knew we had to do that yeah right so we
0:17:04 knew that okay if we wanted to actually digitize the world of scent and and have a record of what the
0:17:08 world smelled like and maybe even replay it yeah we’re gonna have to do this we needed to automate
0:17:14 that and have it be automatic and that’s what we did so basically you can put any smell into the
0:17:19 machine and it’ll tell you what it’s made of at this point oh yeah so you’re setting out to start
0:17:24 osmo like what are you thinking of in terms of the set of potential commercial applications
0:17:34 so we we really had uh we had three in mind okay and there’s still very much in present in mind the
0:17:40 the focus is has become a lot crisper though in terms of what we’re concentrating on we know the
0:17:47 fragrance industry is huge yeah and very profitable and it’s also something i personally love
0:17:54 that’s a thing we want to automate and understand and then we know that dogs can detect things
0:18:00 right and so we know dogs can detect harmful substances like drugs or bombs or things that
0:18:05 just shouldn’t be there like produce where it shouldn’t be being shipped and then we also know
0:18:11 that dogs and even in some cases people can detect health or disease states right right we know that
0:18:17 mrs milna a nurse in the uk was able to smell parkinson’s disease huh uh
0:18:22 and she’s since been able to teach that skill to other people which is really amazing and then we
0:18:28 figured out all the chemistry of what’s actually being smelled we know that there’s many many instances
0:18:34 where there is a scent signature to a disease or to a wellness or to a health state that hasn’t yet
0:18:40 been fully figured out right but we know that they exist those are the three right so fragrance industry
0:18:49 really security and supply chain and health and wellness and i view them in that order because
0:18:56 that’s like the the order in which i think we can be useful to the world right so designing fragrances
0:19:01 is something that’s much more attainable technically and frankly it’s just a great much faster sales
0:19:07 cycle to be yeah business to be in yeah then ultimately diagnostics which are so hard right i mean it is my
0:19:12 north star it’s like where i want to take the company but i’m i also have no illusions about how hard that is
0:19:17 and i’ve just i’ve seen all the failures of the companies that have attempted it and i think i’ve learned from
0:19:23 what hasn’t worked uh and so i’m incorporating those learnings into how i want to build the company which is
0:19:27 build a great business in fragrance build beautiful fragrances for the world
0:19:32 and then strike out from that position of strengthened even more ambitious uh frontiers
0:19:37 we’ll be back in just a minute
0:19:50 run a business and not thinking about podcasting think again more americans listen to podcasts than
0:19:54 ad-supported streaming music from spotify and pandora and as the number one podcaster
0:19:59 iheart’s twice as large as the next two combined so whatever your customers listen to they’ll hear
0:20:05 your message plus only iheart can extend your message to audiences across broadcast radio think
0:20:12 podcasting can help your business think iheart streaming radio and podcasting call 844-844-iheart
0:20:19 to get started that’s 844-844-iheart can we invest our way out of the climate crisis
0:20:26 five years ago it seemed like wall street was working on it until a backlash upended everything
0:20:32 so there’s a lot of alignment between the dark money right and the oil industry on this effort
0:20:40 i’m amy scott host of how we survive a podcast from marketplace in this season we investigate the rise
0:20:48 fall and reincarnation of climate conscious investing listen to how we survive wherever you get your podcasts
0:20:56 when i first heard about the work alex was doing at osmo i understood why it would be useful for
0:21:04 sensing basically you might be able to build automated sniffing machines that could say detect cancer in a
0:21:12 person or detect a bomb in a suitcase but i couldn’t figure out truly what the business case was for
0:21:20 and in fact osmo has recently launched a perfume business it’s called generation so i asked alex
0:21:27 why is using ai and fancy machines why is that better than just designing perfume in the traditional way
0:21:34 we can go from the first kind of client demand so hey i want to create a fragrance and here’s who my
0:21:39 brand is here’s what i want to do yeah so just that description yeah to a starting place of a fragrance
0:21:46 in a minute or two what happens at a traditional perfumer when somebody comes in with that request
0:21:49 well so let’s say you’re an emerging let’s let’s say you’re an emerging brand right so you’re starting
0:21:54 out or you have your first product and you want to add a second one but you’re small right you’re not
0:21:59 making a billion dollars in revenue you’re making less than that so if you want to make a new custom
0:22:05 fragrance good luck right you’re not going to be able to get the attention of the big fragrance houses
0:22:12 because they want to service business that’s like millions and millions of dollars and you’re not big
0:22:17 enough yet so if you want a great custom fragrance that your consumers are going to love yeah and you
0:22:24 want to do quickly so you’re responding to trends yeah um you aren’t going to be able to get it done
0:22:28 so you have to make compromises right so if you want to move fast you’re going to have to use a
0:22:33 regurgitated fragrance it’s also called a library fragrance which means somebody else on the market
0:22:37 has your smell i’m imagining that people who sell it call it a library fragrance rather than a
0:22:42 regurgitated fragrance they do they don’t say regurgitated but that’s what it effectively yes
0:22:48 right fair regurgitated does have a particular olfactory connotation so it’s a clever word for you
0:22:54 it’s visceral yes it’s visceral it sticks in the mind what like and i’m not i just genuinely don’t
0:22:58 understand like why can’t somebody just have a company with a bunch like who knows the molecules
0:23:04 you know who knows what the 5 000 molecules in the book smell like because they’ve got the book and they
0:23:07 can just use the book and be like oh you want this let’s try that do you know what i mean like
0:23:15 i’m not trying to be difficult but i genuinely don’t understand why you need the technology to do that
0:23:23 yeah i genuinely didn’t understand this either and there’s there’s a a class of professional
0:23:27 called a perfumer yeah and their job is to do what you’re describing which is hey i know all these
0:23:32 ingredients and i’m going to mix them in order to create your fragrance so they typically there’s no
0:23:38 perfumer that knows 5 000 ingredients but the best perfumers know a thousand or two thousand ingredients
0:23:47 most perfumers work with 200 100 or 200 ingredients so already there like where we’re there is very few
0:23:52 people in the world that can do what you’re saying yeah they can do and then um how what are they going
0:23:59 to work on right so it might take them uh weeks or months to create a fragrance they’re working on a
0:24:03 few at a time why would they work on an emerging brands fragrance when they can go work on a much
0:24:09 larger account so there’s just very very limited number of people who can engage in the fragrance
0:24:15 creation process because it is difficult it’s not so much identifying hey all these molecules smell this
0:24:19 particular way and therefore i should be able to mix them like what ratios do you mix them in like what
0:24:24 are the rules right and now you’re actually getting into designing a system which understands scent well
0:24:28 enough to create new fragrance formulas as starting places and then of course a perfumer finishes them
0:24:33 um but uh you’re right it’s like huh why shouldn’t that exist and then when you actually start to peel
0:24:37 back the layers one by one you realize oh you actually have to build what we built it’s actually
0:24:41 in order to answer that question so it’s hard so presumably now your model can not only predict what
0:24:46 one molecule is going to smell like but it can predict the combination of molecules i mean is it predicting
0:24:54 does it know concentration like does it know oh it has to yeah yeah how good is it i mean you have a
0:25:05 perfumer on on staff why um well i think the goal of tools yeah is to have them in the hands of creatives
0:25:11 and there’s many steps to perfumery but i think there’s three that are relevant for for what we’re
0:25:15 talking about the first is a perfumer when they’re when they’re starting on a project they have to have
0:25:21 a starting place they have a starting formula and then they do their creative work step two to evolve
0:25:27 that formula to exactly what the customer wants to a creative expression that that delights the
0:25:31 consumer as well that’s the funnest part perfumers love that that’s like that is actual creation and
0:25:37 the creative part number three is then it has to be the right price it has to be compliant with with
0:25:42 regulatory compliance there cannot be allergens all that stuff that’s more like sound engineering than
0:25:47 it is composition or being a rock star steps one the starting place step three all the regulatory
0:25:54 requirements that’s where we spend the most energy in building these tools and then a perfumer is the
0:26:00 person that is taking the formulas from starting place to creative endpoint and then handing it off
0:26:07 for like regulatory finishing and they’re just way more effective with these tools uh-huh um at least for
0:26:13 now right like that’s the way i feel using an llm like i feel like i have a window when me plus the llm is
0:26:19 better than the llm alone and we haven’t that window hasn’t closed yet but i’m not optimistic about my
0:26:26 long-term prospects um we’ll see though i mean but yeah we’ll see yeah i i honest belief here like
0:26:33 the tools will get better but the drive to create will never go away and i think people will always
0:26:39 want to know like about the person behind the creation in a way and it’s not uniform so like i don’t
0:26:43 think people want to know the perfumer behind the hand soap in the gas station yeah they just don’t
0:26:53 right it but there i think will always be room for craft and creative use of tools and the profession
0:26:58 that uses those tools might change radically yeah and the industry in which those tools are used might
0:27:04 change radically but the tools will always be wielded by people but the work that’s being done might be
0:27:11 unrecognizable so you know we’ll see how the world evolves but like i just like ai is like an engine
0:27:18 it’s just a technology it’s just a tool so what’s the what’s the business model just briefly for
0:27:26 generation like how what you know what’s the model the business model really simply is we’ll design the
0:27:31 fragrance for you and then you’ll buy that fragrance to put in your products or we’ll even actually create
0:27:35 the full finished product we’ll put it in a bottle for you if you if you want uh we are behind the
0:27:41 scenes we’re an engine supporting brands we’re not a brand ourself and we’re here to make beautiful
0:27:49 fragrance products for for brands um so what’s the frontier like you have you know on the business
0:27:55 side the generation is kind of the the central thing you’re working on now but on the more on the on
0:28:00 research side like what are you trying to figure out now what are you working on now so there’s
0:28:07 there’s our starting place which is why does this molecule smell the way that it does and we can
0:28:11 never stop getting better at that then there’s the next question of why does this mixture of molecules
0:28:16 smell the way that it does and we can never stop getting better at that and then there’s do you like
0:28:22 it which is maybe the most important question from a business perspective or who likes it and in what
0:28:29 context yeah exactly exactly which is it’s not just the formula as the input to this model but there’s
0:28:33 also who you are right what are your experiences where are you from what are the other things in
0:28:39 your life that you that actually goes back to your netflix collaborative filters like it does if i watch
0:28:48 succession and the sopranos and i’m 50 and then what’s the cologne for me yeah exactly and so uh i was very
0:28:54 fortunate to be able to start this company with a guy work with at twitter name his name is rich
0:28:59 whitcomb he he’s the he’s our chief technology officer his whole professional life has been
0:29:05 recommender system so he he was the a lead on spotify’s song recommender system so if you like
0:29:10 your wrapped playlist or recommended playlist like that’s his code and then he also worked on self-driving
0:29:16 cars at nvidia but he’s he’s been in this world of like hey you like these things what about this thing
0:29:21 or here’s the inputs that the system’s getting what do i do now so really really deep into that world
0:29:27 and we’re kind of bringing that spirit um that that that mindset to to send into fragrance and then
0:29:35 what about beyond you know for for the parts of your work that are the next steps that you alluded to
0:29:40 farther in the distance the essentially sensing right sensing for security sensing for health like
0:29:47 what what work are you doing now toward that end yeah so we’re we’re incubating this right now um so
0:29:52 i’ll tell you two things one is we we have a partner we’ve deployed sensors out in the field we’re
0:29:58 detecting uh inauthentic or counterfeit goods it’s working what’s really the second thing i’ll say is
0:30:04 it’s we’ve learned something really interesting which is the molecules that smell really good in fruits
0:30:11 and flowers and vegetables that we have to understand to create fragrance are the same molecules in counterfeit
0:30:18 luxury goods and the same molecules in our scent and by getting really good at understanding and
0:30:23 designing fragrance in one domain in the fragrance industry we’re actually strengthening this platform
0:30:28 that we’re building to get really good at the next frontiers right of security detection and then
0:30:35 ultimately what we care about is health yeah so that that’s what really surprised us is as i thought
0:30:42 that by working in fragrance we’re making a trade-off which is we’re here to build a great business to
0:30:47 make ourselves resilient so that we can work on the much longer haul problems but in reality we’re making
0:30:52 progress on those problems by teaching our platform about what the world smells like and it’s all one
0:30:57 it’s just scent it’s just molecules in the air and so the more we learn about really any piece of what
0:31:02 the world smells like the better we get at all of it i think i’ll tell you what i think the big like
0:31:10 technical frontier is is predicting emotion ah that’s interesting uh-huh so when when you smell
0:31:15 something you obviously perceive something like the first thought or first perception is whatever fresh
0:31:22 cut grass or grapefruit but then there’s another thing that happens almost at the same time which is
0:31:30 i remember or i feel a particular thing and predicting that is something i don’t think anybody’s really
0:31:37 figured out but is a beautiful frontier well how do you get the data you got to ask a lot of people
0:31:41 how they feel when they smell a lot of things and they have to be able to articulate it right part of
0:31:47 the thing with scent is it’s so primordial that like you might not even be able to say how you feel
0:31:55 so you need a brain computer interface you might you might but turns out we have voices and faces that
0:32:01 are effectively bci’s there’s a lot of information that leaks out of us all the time and that was what
0:32:07 my phd was in is how do you interpret body language in a way that makes sense and by the way the body
0:32:13 language i worked on most closely was body language driven by odors right things that make i studied this
0:32:20 in animals but makes animals happy or sad or afraid or or calm and you can read that out i mean we’re
0:32:25 our behaviors are meant to communicate to other animals right we’re very social we’re a social species
0:32:31 so i think there’s more fundamentals that we have to figure out but um uh this is i think there’s some
0:32:37 really fundamental stuff that’s that’s still unknown here i heard you say in another interview that you
0:32:44 worry sometimes that you’ll hit some barrier in nature to your work uh and you said it in passing
0:32:49 but i was very curious about that what is what does that mean yeah i i always think about that which is
0:32:56 like what what day will it be when mother nature says you can’t figure the next hard thing out uh-huh
0:33:04 and i just look at this from the history of science that you know how if somebody cared about how the
0:33:09 planets were moving in in 1200 well good luck like you don’t have the right telescopes you don’t have
0:33:16 tycho brahe there’s a bunch of stuff you’re gonna need right and so in a way it’s like mother nature and
0:33:21 what our society and species knows conspiring together that basically says progress will have
0:33:27 to wait and so i think about that i worry about that all the time and and so my mental framework
0:33:32 keeps me super humble is like i’m just thankful for all the progress we’ve been able to make that
0:33:38 the tools were around yeah right so i didn’t invent graph neural networks i didn’t even invent the
0:33:42 data sets like we are piecing together and curating and cobbling together all these we’re standing
0:33:49 on the shoulders of so many people and it’s just always been the case and i don’t know it just it
0:33:57 it makes maybe maybe this is too philosophical but for me when i’ve been up close and personal with
0:34:02 scientific progress either that i’ve had a part in or i’ve observed other people do it all feels so
0:34:08 tenuous it feels so lucky because once you really dig into the details you realize oh my gosh they had to
0:34:14 be right there at that time and have known about that thing yeah it’s amazing that anything happens
0:34:18 when you think of how contingent everything is amazing that anything happens and you know when
0:34:25 you really dig in you’re like wow how does anything could happen at all um but nonetheless you persist and
0:34:30 also i think you can create the conditions where it’s more likely than not to happen and so that’s
0:34:35 what osmo is and that’s why osmo birth generation is like let’s create an environment where we’re much
0:34:40 more likely than not to make both the scientific progress we need to make but also like really
0:34:46 help and change the fragrance industry uh which by the way will teach us the things we need to know to
0:34:52 get to the next thing so i think there’s so much beauty to create in the fragrance industry that i’m
0:34:56 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
0:35:01 things that will allow us to do even more audacious work in the future
0:35:06 we’ll be back in a minute with the lightning round
0:35:20 run a business and not thinking about podcasting think again more americans listen to podcasts than
0:35:24 ad-supported streaming music from spotify and pandora and as the number one podcaster
0:35:29 iheart’s twice as large as the next two combined so whatever your customers listen to they’ll hear
0:35:34 your message plus only iheart can extend your message to audiences across broadcast radio think
0:35:40 podcasting can help your business think iheart streaming radio and podcasting let us show you
0:35:48 at iheartadvertising.com that’s iheartadvertising.com can we invest our way out of the climate crisis
0:35:55 five years ago it seemed like wall street was working on it until a backlash upended everything
0:36:01 so there’s a lot of alignment between the dark money right and the oil industry on this effort
0:36:09 i’m amy scott host of how we survive a podcast from marketplace and this season we investigate the rise
0:36:17 fall and reincarnation of climate conscious investing listen to how we survive wherever you get your podcasts
0:36:25 let’s finish with the lightning round i’m gonna ask you a bunch of questions now
0:36:32 um what seemingly pleasant scent do you never want to smell again seemingly pleasant scent i never want
0:36:38 to smell again uh artificial cherry it was the cough syrup that i was forced to drink as a kid
0:36:43 and i’m super sensitive to it the molecules ethyl molatol do not like okay are you wearing fragrance
0:36:48 right now and if so what is it i am not i stopped wearing as soon as i started the company because i
0:36:56 oh of course of course um what like what’s your what’s your give me give me a pick name some
0:37:04 fragrance that’s that you love for some reason so i really like this is kind of a a basic choice from
0:37:10 folks inside the industry i love terre d’hermès it’s like the hermès flagship men’s fragrance it’s
0:37:15 by a perfumer jean-claude elena i really love his work basic is that like basic in the way of saying
0:37:18 it’s like if i asked you for a watch and you said a rolex submariner or something it’s just like
0:37:23 exactly or saying like what pop music do you like you said taylor swift people like it because it’s
0:37:29 great uh-huh like taylor swift is great a rolex watch is a great watch terre d’hermès is a great
0:37:34 fragrance but it’s very popular what is it about it that you love i love its minimalism and i have i
0:37:39 just happen to like the notes right so it’s really heavy on a molecule i like iso b super i think it’s a
0:37:44 great highlight of that ingredient and it just wears really well on my skin so that was what i used to
0:37:52 wear almost every day before i stopped what’s your second favorite sense my second favorite sense is
0:37:59 probably gonna be it’s a hard between vision and hearing because i love music uh-huh but i like
0:38:06 looking at stuff too like the world world of beautiful um are more expensive perfumes actually
0:38:13 better sometimes right so i think there’s just like anything like bicycles or art as you start to pay
0:38:17 more everything gets better and then it plateaus right what’s the worst thing you ever smelled
0:38:26 i i have a memory i picked a mushroom that i thought looked cool and wanted to show it to my dad when i
0:38:33 was young and i forgot about it and it was just turned completely gross i had a version of that of bringing
0:38:39 shells home from the beach that were alive it turned out i found out when they were dead it’s like great
0:38:43 intentions but didn’t really have wherewithal to think that through or understand the consequences
0:38:58 alex wilchko is the co-founder and ceo of osmo today’s show was produced by gabriel hunter chang it was
0:39:05 edited by lydia jean kott and engineered by sarah bougier you can email us at problem at pushkin.fm
0:39:10 i’m jacob goldstein and we’ll be back next week with another episode of what’s your problem
0:39:29 can we invest our way out of the climate crisis five years ago it seemed like wall street was working
0:39:36 on it until a backlash upended everything so there’s a lot of alignment between the dark money
0:39:43 right and the oil industry on this effort i’m amy scott host of how we survive a podcast from marketplace
0:39:51 and this season we investigate the rise fall and reincarnation of climate conscious investing
0:39:59 listen to how we survive wherever you get your podcasts you’re listening to an iheart podcast
0:00:18 run a business and not thinking about podcasting think again more americans listen to podcasts
0:00:23 then add supported streaming music from spotify and pandora and as the number one podcaster
0:00:27 iheart’s twice as large as the next two combined learn how podcasting can help your business call
0:00:35 844-844-iheart can we invest our way out of the climate crisis five years ago it seemed like
0:00:42 wall street was working on it until a backlash upended everything so there’s a lot of alignment
0:00:49 between the dark money right and the oil industry on this effort i’m amy scott host of how we survive
0:00:56 a podcast from marketplace and this season we investigate the rise fall and reincarnation
0:01:02 of climate conscious investing listen to how we survive wherever you get your podcasts
0:01:12 tell me what we don’t know about how smell works oh geez uh be shorter to tell you what we do
0:01:20 this is alex wilchko he’s the co-founder and ceo of a company called osmo and despite his protests there
0:01:29 he did tell me some of the things nobody knows about how smell works why do things smell the way that they do
0:01:37 why can we smell certain things and not other things what is the logic of how molecules are combined to create
0:01:45 beautiful smells why do some smells create incredibly powerful emotional associations instantly and others seem neutral
0:01:51 right why do something smell different to people right i think we we have a hints in all these directions but we
0:01:59 we have nothing like musical scales where we have nothing like a periodic table we don’t know any structure to why
0:02:05 things are the way that they are it’s a ton of mystery um and that’s what makes it so exciting to work on this
0:02:13 topic it’s like there’s so much we don’t know and to be clear like with light we just know whatever
0:02:19 if you tell me the the frequency the wavelength i can know exactly what color you’re talking about
0:02:25 or the same thing with the with a waveform of sound right and so but if i give you some random molecule
0:02:31 and say what does it smell like do you know so that’s what i’ve spent a lot of my professional
0:02:35 life working on it’s exactly that question yeah which is draw a structure of a molecule on a white
0:02:41 board point at it and say hey is this what does it smell like wood or flowers or fruits or whatever
0:02:49 and so there is no way to know that for sure at all but there’s no good way even statistically
0:02:56 to predict that without using large data sets and at least in our hands i you need neural networks you
0:03:06 need deep learning yeah in order to do that i’m jacob goldstein and this is what’s your problem
0:03:12 the show where i talk to people who are trying to make technological progress alex wiltschko’s problem
0:03:19 is this can you use ai to teach computers to smell and once you’ve figured out how to do that
0:03:26 can you build a profitable business around it osmo spun out of google in 2023 the company
0:03:31 recently launched a fragrance house to develop new perfumes they’ve also done some work using scent
0:03:38 to detect counterfeit shoes and in the long run they plan to use scent to diagnose disease
0:03:46 before he started osmo alex worked at google as an ai researcher before that he got a phd at harvard
0:03:52 studying how mice respond to scent but maybe the most important part of his bio came even earlier in
0:04:00 his life specifically when he was 12 years old and went off to summer camp in his home state texas i was
0:04:06 from a small town college station and then most of the kids were from big towns like houston and dallas and
0:04:15 austin san antonio and i hadn’t really been exposed to like i don’t know fashion trends or you know what
0:04:21 was cool or popular but everybody’s all lumped together in summer camp and then there was this
0:04:28 thing called perfume that some of the richer frankly richer and more popular kids had hello and it was just
0:04:38 amazing to me that these boys could spray themselves with this invisible mist a clear mist and then for
0:04:44 the next four to six hours people around them would treat them differently and that just blew my mind
0:04:50 right like there’s no i can’t i can see the clothes yeah i can see how they act and walk and talk and
0:04:57 how they you know posture and all that but i cannot see the fragrance yeah but yet it is obviously doing
0:05:05 something magical it’s like an axe body spray ad uh what what does that cause you to do when i get home
0:05:16 and fail um i uh we shopped at tj maxx and i started to really look out for fragrances there
0:05:20 and then it just it kind of snowballs from there which i just realized there’s like a whole lot of
0:05:26 these things and guess what you can just try them and some of them are actually way better and more
0:05:31 opinionated and more beautiful i don’t i didn’t have the vocabulary then but like it just it was clear
0:05:36 to me early on that like i never really thought about who made the clothes but i started to think
0:05:41 about who made these perfumes uh-huh because it was clear that there were choices that were being made
0:05:48 and like i just remember trying and this was years later trying bulgari black which really kind of
0:05:54 clued me into this world that bulgari black is not necessarily a great fragrance but you can
0:05:59 experience the top middle and base notes in like 40 minutes 45 minutes it’s pretty short and so like
0:06:05 a bigger fragrance like a creed eventus will last on your skin for a day and so you the the whole
0:06:12 fragrance unfolds i mean top notes will last max 15 30 minutes but the heart might last for several
0:06:17 hours and the base note might last for 10 hours right so it so it smells different you can still
0:06:21 smell it but it smells different whatever an hour after you put it on and four hours after that
0:06:26 because a great fragrance is actually many different fragrances within it yeah right there’s the first one
0:06:30 which peels off really quickly burns off quickly there’s the second fragrance which is the heart
0:06:34 note which will last for you know sometimes hours but in this case like 20 another 20 minutes yeah and
0:06:39 then the the base note which is a third fragrance which is what’s left after those two burn off
0:06:44 um and it’s that’s it’s like three acts of a movie right it’s i think it’s quite beautiful so how do we
0:06:55 get from you being a teenager preoccupied with fragrance to you using ai to predict how molecules
0:07:00 will smell yeah like the computer part was always different from the fragrance part i just i love
0:07:05 computers uh we always had computers at home i started programming when i was around i don’t know
0:07:10 eight years old that was my life like my entire life was computers for a long time still is in a way
0:07:17 yeah um and fragrance was not a part of it i got into you know statistics which became machine
0:07:20 learning around the same time again for totally independent reason there’s this thing called the
0:07:26 netflix prize it was like one of the first competitions to build great ml algorithms i competed
0:07:30 in that i mean that’s basically to tell me what else i’ll like on netflix right that’s what that
0:07:37 contest was like if i’ve watched whatever if i watch succession and the sopranos what should i watch next
0:07:42 then you’re gonna like another kind of dark but you know funny kind of uh soap opera type of a thing
0:07:47 exactly and so netflix did a really bold thing which is they released a data set and said here’s what
0:07:52 good looks like here’s how we measure it how about it and they paid a million dollars to the winner um
0:07:58 which was a combination of a few teams but what they really did is they brought a particular kind of
0:08:04 machine learning into the forefront called collaborative filtering really showed that this stuff worked and by
0:08:10 the way other companies were already racing to use this so like this recommender systems was a big thing
0:08:14 um but netflix was putting it out into the public and allowed a kid like me at i think i was 18 or 19
0:08:20 years old to actually compete and do pretty well in that and so i just got exposed to this world through
0:08:26 that and it was super fun i mean they gamified it and i had a i had blast so that was my first exposure
0:08:31 to machine learning turned out to be a good time to start working on machine learning yeah totally because if i
0:08:36 started now they wouldn’t let me in because everybody’s so smart that was probably 10 years ago yeah 10 years
0:08:44 ago yeah yeah and um then you know i was doing my undergraduate training in neuroscience and i was
0:08:50 studying more behavior uh than olfaction because it actually turned out that olfaction was a hyper
0:08:55 specialized subfield of neuroscience i didn’t realize how niche it was i loved smell and i was doing
0:09:00 neuroscience and i knew i wanted to do smell neuroscience the fancy word for that is olfactory
0:09:06 neuroscience and so there’s really two universities in the world that like have a critical mass of
0:09:12 these researchers it’s columbia and it’s harvard and i applied to both i went to harvard and i realized
0:09:19 nobody cares about this problem nobody cares about why molecules smell the way that they do there’s a much
0:09:23 longer conversation as to why that’s the case and why that’s still persisting how that’s changing
0:09:28 well let me ask you this let me ask you this at that time i mean i i get it as a basic research
0:09:32 question i mean i’ll tell you i we i was talking with the producer and editor of this show and we
0:09:36 were getting ready for this interview and we had this interesting conversation talking about scent and
0:09:41 what you’re working on whatever and i went down uh and i saw my daughter and she said what are you
0:09:45 working on i said this guy who’s trying to figure out scent and teach computers to smell and she said
0:09:53 why i said i don’t know i should ask him that so uh why why was it compelling to you i get it as a basic
0:09:58 research question but at that time was it like was there were there applications that came to your mind
0:10:09 look the the the steps of um development of this thing that’s now osmo went through those different
0:10:15 iterations of you know i started as an academic scientist and i was trained in that world and then
0:10:20 i left and i did some entrepreneurship yeah i ended up in industrial research and they’re like being
0:10:26 curious frankly was enough and the idea this is this is at google this is now a google brain yeah and
0:10:31 there’s a few steps in between but basically as a you’re an ai researcher at google at this moment when
0:10:35 you’re doing industrial research right yeah exactly in google brain at the time now it’s google deep
0:10:40 mind very much had like a thousand flowers bloom mentality and so people were working on crazy
0:10:44 stuff including me working on smell bell labs it’s basically like bell labs of the 21st century right
0:10:52 you have it exactly bell labs xerox park yeah kind of it truly was dreamy sounds dreamy it was awesome
0:10:57 right and it’s also a moment in time and now i think that moment’s gone for better or for worse um
0:11:05 the idea was pretty straightforward for google which was are the products that google know what the world
0:11:09 looks like and know what the world sounds like and that’s useful right that’s information that google’s
0:11:15 organizing if we knew what things smelled and tasted like that would be useful right uh-huh the original
0:11:20 mission of google is organize the world’s information right exactly and make it universally accessible and
0:11:26 useful and there was a whole slice of reality uh-huh the the chemical slice of reality that was
0:11:32 invisible right to not just to google but to computers yeah yeah and that felt really important and we had
0:11:36 agreement and buy-in all the way up to the executive level they’re like yeah let’s go let’s go look at
0:11:43 that so you’re doing a like basic ai research at google and you decide to see if you can basically use ai
0:11:49 to figure out scent to say here is a molecule what does it smell like right that’s the basic
0:11:52 endeavor how do you do that what is it that you actually do
0:11:56 yeah so first it starts the motivation which is like let’s figure out smell
0:12:02 um but it actually was a lot more natural than i think it sounds yeah which is
0:12:09 scent is just chemistry yeah it’s molecules and we got to do ai for molecules right if we’re going
0:12:14 to do ai for scent yeah and the thing that had happened in between you know it was a five-year
0:12:21 period between my academic life and my industrial life yeah and what had happened in those five years
0:12:25 is actually some of the some of the people i did my phd with and then some of the people i ended up
0:12:32 working with at google brain really cracked machine learning or ai for molecules but they didn’t do it for
0:12:37 scent they did it for a few other things they did for drug discovery and they did it for um like
0:12:45 materials discovery so like new materials for leds right so you happen to be doing essentially basic
0:12:52 research at google at this moment when there is this new way to use ai that is well suited to
0:12:59 molecules and you say we can do scent do it yeah let’s do it yeah we can do it the other pieces are
0:13:05 great you got the algorithm where’s the data classic that’s the classic ai question right like exactly
0:13:10 where’s the data what i did know just from being obsessed and in this world for a long time prior
0:13:17 that there were these collections of data sets that were honestly really more like magazine catalogs
0:13:23 for fragrance ingredients and so there were these catalogs basically saying this is the ingredient this
0:13:27 is the molecular structure of this ingredient and here’s what it smells like and by the way
0:13:33 the rating of what it smells like was done by a professional by a perfumer and so the special sauce
0:13:39 that we added is we we went and we got that data and we we fused a few data sets together and we cleaned
0:13:44 it very carefully and that that hadn’t been done and it’s something it’s like 5 000 dish right it’s
0:13:51 5 000 or so different molecules okay yep exactly and here is this the one with the list i love the list
0:13:59 here i have it this the one sweet fruity vanilla powdery floral berry fermented nutty ozone buttery musk
0:14:04 it’s that it’s that list right those are they and there’s 138 of those descriptors i think that we
0:14:10 used in that data set sometimes we use smaller subsets but the full set originally is about 140
0:14:17 so okay so you have your whatever your 5 000 molecules labeled with 140 different cents
0:14:24 you train your ai model on this data set and then you want to find out does the model work does the ai
0:14:31 work right if i give the model some new molecule molecule that wasn’t in the training data will it
0:14:37 know what that molecule smells like and and to test that to answer that question you actually do this study
0:14:42 so you get a bunch of people to smell these molecules that are not that your model was not
0:14:47 trained on essentially right and say what it smells like it’s weird like what you you don’t actually
0:14:52 care what it fundamentally smells like you just care what everybody on average thinks it smells like
0:14:56 because guess what that’s what it’s that’s what smell is yeah that’s the meaning of smell yeah
0:15:03 so you ask this panel to what do all what do all these uh molecules smell like and then you ask the
0:15:10 model what do they smell like and you compare the results and how does the model do um that was really the
0:15:17 threshold of breakthrough in my mind was like are you worse than a person yeah or are you slightly better
0:15:23 than a person and we got slightly better than a person which was a breakthrough in my view right and so
0:15:29 yes so that paper you published in science and and you started osmo kind of around the same time right
0:15:34 you started that study at at google is that right and then by the time it was published you had spun
0:15:40 osmo out of google right that’s right so you you have this map you have this model that can basically
0:15:46 given a molecule predict pretty well what the average person thinks that molecule will smell like
0:15:53 but there is still a second problem right which is in the world in the wild you don’t know
0:15:58 what molecules are in the air you don’t know what molecule somebody’s smelling and so for that
0:16:04 second problem you need to try and build some kind of automated system for figuring out what molecules
0:16:11 are in the air at a given that’s correct getting to one molecule structure is actually not trivial
0:16:16 so to go from a physical thing and know all the molecular structures like not a solid problem
0:16:22 so there’s a lot of ways to do that there’s a lot of chemical sensors out there none of them will just
0:16:30 tell you the formula right so that’s hard really hard so there’s like a chemistry problem of like
0:16:37 isolating the molecule basically and and deriving the chemical formula exactly taking a real smell
0:16:40 and it’s composed of a bunch of different molecules with different structures and there’s different
0:16:47 amounts there’s ratios yeah you got to get that recipe out of the air so that’s on that’s hard
0:16:53 that was unsolved at the time to do in an automated way and by the way if we’re following this story
0:16:58 chronologically we hadn’t done this yet at google yeah but we knew we had to do that yeah right so we
0:17:04 knew that okay if we wanted to actually digitize the world of scent and and have a record of what the
0:17:08 world smelled like and maybe even replay it yeah we’re gonna have to do this we needed to automate
0:17:14 that and have it be automatic and that’s what we did so basically you can put any smell into the
0:17:19 machine and it’ll tell you what it’s made of at this point oh yeah so you’re setting out to start
0:17:24 osmo like what are you thinking of in terms of the set of potential commercial applications
0:17:34 so we we really had uh we had three in mind okay and there’s still very much in present in mind the
0:17:40 the focus is has become a lot crisper though in terms of what we’re concentrating on we know the
0:17:47 fragrance industry is huge yeah and very profitable and it’s also something i personally love
0:17:54 that’s a thing we want to automate and understand and then we know that dogs can detect things
0:18:00 right and so we know dogs can detect harmful substances like drugs or bombs or things that
0:18:05 just shouldn’t be there like produce where it shouldn’t be being shipped and then we also know
0:18:11 that dogs and even in some cases people can detect health or disease states right right we know that
0:18:17 mrs milna a nurse in the uk was able to smell parkinson’s disease huh uh
0:18:22 and she’s since been able to teach that skill to other people which is really amazing and then we
0:18:28 figured out all the chemistry of what’s actually being smelled we know that there’s many many instances
0:18:34 where there is a scent signature to a disease or to a wellness or to a health state that hasn’t yet
0:18:40 been fully figured out right but we know that they exist those are the three right so fragrance industry
0:18:49 really security and supply chain and health and wellness and i view them in that order because
0:18:56 that’s like the the order in which i think we can be useful to the world right so designing fragrances
0:19:01 is something that’s much more attainable technically and frankly it’s just a great much faster sales
0:19:07 cycle to be yeah business to be in yeah then ultimately diagnostics which are so hard right i mean it is my
0:19:12 north star it’s like where i want to take the company but i’m i also have no illusions about how hard that is
0:19:17 and i’ve just i’ve seen all the failures of the companies that have attempted it and i think i’ve learned from
0:19:23 what hasn’t worked uh and so i’m incorporating those learnings into how i want to build the company which is
0:19:27 build a great business in fragrance build beautiful fragrances for the world
0:19:32 and then strike out from that position of strengthened even more ambitious uh frontiers
0:19:37 we’ll be back in just a minute
0:19:50 run a business and not thinking about podcasting think again more americans listen to podcasts than
0:19:54 ad-supported streaming music from spotify and pandora and as the number one podcaster
0:19:59 iheart’s twice as large as the next two combined so whatever your customers listen to they’ll hear
0:20:05 your message plus only iheart can extend your message to audiences across broadcast radio think
0:20:12 podcasting can help your business think iheart streaming radio and podcasting call 844-844-iheart
0:20:19 to get started that’s 844-844-iheart can we invest our way out of the climate crisis
0:20:26 five years ago it seemed like wall street was working on it until a backlash upended everything
0:20:32 so there’s a lot of alignment between the dark money right and the oil industry on this effort
0:20:40 i’m amy scott host of how we survive a podcast from marketplace in this season we investigate the rise
0:20:48 fall and reincarnation of climate conscious investing listen to how we survive wherever you get your podcasts
0:20:56 when i first heard about the work alex was doing at osmo i understood why it would be useful for
0:21:04 sensing basically you might be able to build automated sniffing machines that could say detect cancer in a
0:21:12 person or detect a bomb in a suitcase but i couldn’t figure out truly what the business case was for
0:21:20 and in fact osmo has recently launched a perfume business it’s called generation so i asked alex
0:21:27 why is using ai and fancy machines why is that better than just designing perfume in the traditional way
0:21:34 we can go from the first kind of client demand so hey i want to create a fragrance and here’s who my
0:21:39 brand is here’s what i want to do yeah so just that description yeah to a starting place of a fragrance
0:21:46 in a minute or two what happens at a traditional perfumer when somebody comes in with that request
0:21:49 well so let’s say you’re an emerging let’s let’s say you’re an emerging brand right so you’re starting
0:21:54 out or you have your first product and you want to add a second one but you’re small right you’re not
0:21:59 making a billion dollars in revenue you’re making less than that so if you want to make a new custom
0:22:05 fragrance good luck right you’re not going to be able to get the attention of the big fragrance houses
0:22:12 because they want to service business that’s like millions and millions of dollars and you’re not big
0:22:17 enough yet so if you want a great custom fragrance that your consumers are going to love yeah and you
0:22:24 want to do quickly so you’re responding to trends yeah um you aren’t going to be able to get it done
0:22:28 so you have to make compromises right so if you want to move fast you’re going to have to use a
0:22:33 regurgitated fragrance it’s also called a library fragrance which means somebody else on the market
0:22:37 has your smell i’m imagining that people who sell it call it a library fragrance rather than a
0:22:42 regurgitated fragrance they do they don’t say regurgitated but that’s what it effectively yes
0:22:48 right fair regurgitated does have a particular olfactory connotation so it’s a clever word for you
0:22:54 it’s visceral yes it’s visceral it sticks in the mind what like and i’m not i just genuinely don’t
0:22:58 understand like why can’t somebody just have a company with a bunch like who knows the molecules
0:23:04 you know who knows what the 5 000 molecules in the book smell like because they’ve got the book and they
0:23:07 can just use the book and be like oh you want this let’s try that do you know what i mean like
0:23:15 i’m not trying to be difficult but i genuinely don’t understand why you need the technology to do that
0:23:23 yeah i genuinely didn’t understand this either and there’s there’s a a class of professional
0:23:27 called a perfumer yeah and their job is to do what you’re describing which is hey i know all these
0:23:32 ingredients and i’m going to mix them in order to create your fragrance so they typically there’s no
0:23:38 perfumer that knows 5 000 ingredients but the best perfumers know a thousand or two thousand ingredients
0:23:47 most perfumers work with 200 100 or 200 ingredients so already there like where we’re there is very few
0:23:52 people in the world that can do what you’re saying yeah they can do and then um how what are they going
0:23:59 to work on right so it might take them uh weeks or months to create a fragrance they’re working on a
0:24:03 few at a time why would they work on an emerging brands fragrance when they can go work on a much
0:24:09 larger account so there’s just very very limited number of people who can engage in the fragrance
0:24:15 creation process because it is difficult it’s not so much identifying hey all these molecules smell this
0:24:19 particular way and therefore i should be able to mix them like what ratios do you mix them in like what
0:24:24 are the rules right and now you’re actually getting into designing a system which understands scent well
0:24:28 enough to create new fragrance formulas as starting places and then of course a perfumer finishes them
0:24:33 um but uh you’re right it’s like huh why shouldn’t that exist and then when you actually start to peel
0:24:37 back the layers one by one you realize oh you actually have to build what we built it’s actually
0:24:41 in order to answer that question so it’s hard so presumably now your model can not only predict what
0:24:46 one molecule is going to smell like but it can predict the combination of molecules i mean is it predicting
0:24:54 does it know concentration like does it know oh it has to yeah yeah how good is it i mean you have a
0:25:05 perfumer on on staff why um well i think the goal of tools yeah is to have them in the hands of creatives
0:25:11 and there’s many steps to perfumery but i think there’s three that are relevant for for what we’re
0:25:15 talking about the first is a perfumer when they’re when they’re starting on a project they have to have
0:25:21 a starting place they have a starting formula and then they do their creative work step two to evolve
0:25:27 that formula to exactly what the customer wants to a creative expression that that delights the
0:25:31 consumer as well that’s the funnest part perfumers love that that’s like that is actual creation and
0:25:37 the creative part number three is then it has to be the right price it has to be compliant with with
0:25:42 regulatory compliance there cannot be allergens all that stuff that’s more like sound engineering than
0:25:47 it is composition or being a rock star steps one the starting place step three all the regulatory
0:25:54 requirements that’s where we spend the most energy in building these tools and then a perfumer is the
0:26:00 person that is taking the formulas from starting place to creative endpoint and then handing it off
0:26:07 for like regulatory finishing and they’re just way more effective with these tools uh-huh um at least for
0:26:13 now right like that’s the way i feel using an llm like i feel like i have a window when me plus the llm is
0:26:19 better than the llm alone and we haven’t that window hasn’t closed yet but i’m not optimistic about my
0:26:26 long-term prospects um we’ll see though i mean but yeah we’ll see yeah i i honest belief here like
0:26:33 the tools will get better but the drive to create will never go away and i think people will always
0:26:39 want to know like about the person behind the creation in a way and it’s not uniform so like i don’t
0:26:43 think people want to know the perfumer behind the hand soap in the gas station yeah they just don’t
0:26:53 right it but there i think will always be room for craft and creative use of tools and the profession
0:26:58 that uses those tools might change radically yeah and the industry in which those tools are used might
0:27:04 change radically but the tools will always be wielded by people but the work that’s being done might be
0:27:11 unrecognizable so you know we’ll see how the world evolves but like i just like ai is like an engine
0:27:18 it’s just a technology it’s just a tool so what’s the what’s the business model just briefly for
0:27:26 generation like how what you know what’s the model the business model really simply is we’ll design the
0:27:31 fragrance for you and then you’ll buy that fragrance to put in your products or we’ll even actually create
0:27:35 the full finished product we’ll put it in a bottle for you if you if you want uh we are behind the
0:27:41 scenes we’re an engine supporting brands we’re not a brand ourself and we’re here to make beautiful
0:27:49 fragrance products for for brands um so what’s the frontier like you have you know on the business
0:27:55 side the generation is kind of the the central thing you’re working on now but on the more on the on
0:28:00 research side like what are you trying to figure out now what are you working on now so there’s
0:28:07 there’s our starting place which is why does this molecule smell the way that it does and we can
0:28:11 never stop getting better at that then there’s the next question of why does this mixture of molecules
0:28:16 smell the way that it does and we can never stop getting better at that and then there’s do you like
0:28:22 it which is maybe the most important question from a business perspective or who likes it and in what
0:28:29 context yeah exactly exactly which is it’s not just the formula as the input to this model but there’s
0:28:33 also who you are right what are your experiences where are you from what are the other things in
0:28:39 your life that you that actually goes back to your netflix collaborative filters like it does if i watch
0:28:48 succession and the sopranos and i’m 50 and then what’s the cologne for me yeah exactly and so uh i was very
0:28:54 fortunate to be able to start this company with a guy work with at twitter name his name is rich
0:28:59 whitcomb he he’s the he’s our chief technology officer his whole professional life has been
0:29:05 recommender system so he he was the a lead on spotify’s song recommender system so if you like
0:29:10 your wrapped playlist or recommended playlist like that’s his code and then he also worked on self-driving
0:29:16 cars at nvidia but he’s he’s been in this world of like hey you like these things what about this thing
0:29:21 or here’s the inputs that the system’s getting what do i do now so really really deep into that world
0:29:27 and we’re kind of bringing that spirit um that that that mindset to to send into fragrance and then
0:29:35 what about beyond you know for for the parts of your work that are the next steps that you alluded to
0:29:40 farther in the distance the essentially sensing right sensing for security sensing for health like
0:29:47 what what work are you doing now toward that end yeah so we’re we’re incubating this right now um so
0:29:52 i’ll tell you two things one is we we have a partner we’ve deployed sensors out in the field we’re
0:29:58 detecting uh inauthentic or counterfeit goods it’s working what’s really the second thing i’ll say is
0:30:04 it’s we’ve learned something really interesting which is the molecules that smell really good in fruits
0:30:11 and flowers and vegetables that we have to understand to create fragrance are the same molecules in counterfeit
0:30:18 luxury goods and the same molecules in our scent and by getting really good at understanding and
0:30:23 designing fragrance in one domain in the fragrance industry we’re actually strengthening this platform
0:30:28 that we’re building to get really good at the next frontiers right of security detection and then
0:30:35 ultimately what we care about is health yeah so that that’s what really surprised us is as i thought
0:30:42 that by working in fragrance we’re making a trade-off which is we’re here to build a great business to
0:30:47 make ourselves resilient so that we can work on the much longer haul problems but in reality we’re making
0:30:52 progress on those problems by teaching our platform about what the world smells like and it’s all one
0:30:57 it’s just scent it’s just molecules in the air and so the more we learn about really any piece of what
0:31:02 the world smells like the better we get at all of it i think i’ll tell you what i think the big like
0:31:10 technical frontier is is predicting emotion ah that’s interesting uh-huh so when when you smell
0:31:15 something you obviously perceive something like the first thought or first perception is whatever fresh
0:31:22 cut grass or grapefruit but then there’s another thing that happens almost at the same time which is
0:31:30 i remember or i feel a particular thing and predicting that is something i don’t think anybody’s really
0:31:37 figured out but is a beautiful frontier well how do you get the data you got to ask a lot of people
0:31:41 how they feel when they smell a lot of things and they have to be able to articulate it right part of
0:31:47 the thing with scent is it’s so primordial that like you might not even be able to say how you feel
0:31:55 so you need a brain computer interface you might you might but turns out we have voices and faces that
0:32:01 are effectively bci’s there’s a lot of information that leaks out of us all the time and that was what
0:32:07 my phd was in is how do you interpret body language in a way that makes sense and by the way the body
0:32:13 language i worked on most closely was body language driven by odors right things that make i studied this
0:32:20 in animals but makes animals happy or sad or afraid or or calm and you can read that out i mean we’re
0:32:25 our behaviors are meant to communicate to other animals right we’re very social we’re a social species
0:32:31 so i think there’s more fundamentals that we have to figure out but um uh this is i think there’s some
0:32:37 really fundamental stuff that’s that’s still unknown here i heard you say in another interview that you
0:32:44 worry sometimes that you’ll hit some barrier in nature to your work uh and you said it in passing
0:32:49 but i was very curious about that what is what does that mean yeah i i always think about that which is
0:32:56 like what what day will it be when mother nature says you can’t figure the next hard thing out uh-huh
0:33:04 and i just look at this from the history of science that you know how if somebody cared about how the
0:33:09 planets were moving in in 1200 well good luck like you don’t have the right telescopes you don’t have
0:33:16 tycho brahe there’s a bunch of stuff you’re gonna need right and so in a way it’s like mother nature and
0:33:21 what our society and species knows conspiring together that basically says progress will have
0:33:27 to wait and so i think about that i worry about that all the time and and so my mental framework
0:33:32 keeps me super humble is like i’m just thankful for all the progress we’ve been able to make that
0:33:38 the tools were around yeah right so i didn’t invent graph neural networks i didn’t even invent the
0:33:42 data sets like we are piecing together and curating and cobbling together all these we’re standing
0:33:49 on the shoulders of so many people and it’s just always been the case and i don’t know it just it
0:33:57 it makes maybe maybe this is too philosophical but for me when i’ve been up close and personal with
0:34:02 scientific progress either that i’ve had a part in or i’ve observed other people do it all feels so
0:34:08 tenuous it feels so lucky because once you really dig into the details you realize oh my gosh they had to
0:34:14 be right there at that time and have known about that thing yeah it’s amazing that anything happens
0:34:18 when you think of how contingent everything is amazing that anything happens and you know when
0:34:25 you really dig in you’re like wow how does anything could happen at all um but nonetheless you persist and
0:34:30 also i think you can create the conditions where it’s more likely than not to happen and so that’s
0:34:35 what osmo is and that’s why osmo birth generation is like let’s create an environment where we’re much
0:34:40 more likely than not to make both the scientific progress we need to make but also like really
0:34:46 help and change the fragrance industry uh which by the way will teach us the things we need to know to
0:34:52 get to the next thing so i think there’s so much beauty to create in the fragrance industry that i’m
0:34:56 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
0:35:01 things that will allow us to do even more audacious work in the future
0:35:06 we’ll be back in a minute with the lightning round
0:35:20 run a business and not thinking about podcasting think again more americans listen to podcasts than
0:35:24 ad-supported streaming music from spotify and pandora and as the number one podcaster
0:35:29 iheart’s twice as large as the next two combined so whatever your customers listen to they’ll hear
0:35:34 your message plus only iheart can extend your message to audiences across broadcast radio think
0:35:40 podcasting can help your business think iheart streaming radio and podcasting let us show you
0:35:48 at iheartadvertising.com that’s iheartadvertising.com can we invest our way out of the climate crisis
0:35:55 five years ago it seemed like wall street was working on it until a backlash upended everything
0:36:01 so there’s a lot of alignment between the dark money right and the oil industry on this effort
0:36:09 i’m amy scott host of how we survive a podcast from marketplace and this season we investigate the rise
0:36:17 fall and reincarnation of climate conscious investing listen to how we survive wherever you get your podcasts
0:36:25 let’s finish with the lightning round i’m gonna ask you a bunch of questions now
0:36:32 um what seemingly pleasant scent do you never want to smell again seemingly pleasant scent i never want
0:36:38 to smell again uh artificial cherry it was the cough syrup that i was forced to drink as a kid
0:36:43 and i’m super sensitive to it the molecules ethyl molatol do not like okay are you wearing fragrance
0:36:48 right now and if so what is it i am not i stopped wearing as soon as i started the company because i
0:36:56 oh of course of course um what like what’s your what’s your give me give me a pick name some
0:37:04 fragrance that’s that you love for some reason so i really like this is kind of a a basic choice from
0:37:10 folks inside the industry i love terre d’hermès it’s like the hermès flagship men’s fragrance it’s
0:37:15 by a perfumer jean-claude elena i really love his work basic is that like basic in the way of saying
0:37:18 it’s like if i asked you for a watch and you said a rolex submariner or something it’s just like
0:37:23 exactly or saying like what pop music do you like you said taylor swift people like it because it’s
0:37:29 great uh-huh like taylor swift is great a rolex watch is a great watch terre d’hermès is a great
0:37:34 fragrance but it’s very popular what is it about it that you love i love its minimalism and i have i
0:37:39 just happen to like the notes right so it’s really heavy on a molecule i like iso b super i think it’s a
0:37:44 great highlight of that ingredient and it just wears really well on my skin so that was what i used to
0:37:52 wear almost every day before i stopped what’s your second favorite sense my second favorite sense is
0:37:59 probably gonna be it’s a hard between vision and hearing because i love music uh-huh but i like
0:38:06 looking at stuff too like the world world of beautiful um are more expensive perfumes actually
0:38:13 better sometimes right so i think there’s just like anything like bicycles or art as you start to pay
0:38:17 more everything gets better and then it plateaus right what’s the worst thing you ever smelled
0:38:26 i i have a memory i picked a mushroom that i thought looked cool and wanted to show it to my dad when i
0:38:33 was young and i forgot about it and it was just turned completely gross i had a version of that of bringing
0:38:39 shells home from the beach that were alive it turned out i found out when they were dead it’s like great
0:38:43 intentions but didn’t really have wherewithal to think that through or understand the consequences
0:38:58 alex wilchko is the co-founder and ceo of osmo today’s show was produced by gabriel hunter chang it was
0:39:05 edited by lydia jean kott and engineered by sarah bougier you can email us at problem at pushkin.fm
0:39:10 i’m jacob goldstein and we’ll be back next week with another episode of what’s your problem
0:39:29 can we invest our way out of the climate crisis five years ago it seemed like wall street was working
0:39:36 on it until a backlash upended everything so there’s a lot of alignment between the dark money
0:39:43 right and the oil industry on this effort i’m amy scott host of how we survive a podcast from marketplace
0:39:51 and this season we investigate the rise fall and reincarnation of climate conscious investing
0:39:59 listen to how we survive wherever you get your podcasts you’re listening to an iheart podcast
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.