Alex Wiltschko - Giving Computers A Sense Of Smell - [Invest Like the Best, EP.415]

0:00 I know firsthand how complex the tech stack is for asset management firms. And seemingly every new tool and data source makes the problem even worse, adding more complexity, more headcount, and more risk. Ridge line offers a better way forward, one unified platform that automates away the complexity across portfolio accounting. Reconciliation, reporting, trading, compliance, and more, all at scale. Ridge line is revolutionizing investment management, helping ambitious firms scale faster.

0:25 Operate smarter and stay ahead of the curve. See what Ridgeline can unlock for your firm. Schedule a demo at ridgeline.ai. Hello and welcome, everyone. I'm Patrick O'Shaughnessy and this is Invest Like the Best. This show is an open-ended exploration of markets, ideas, stories, and strategies that will help you better invest both your time and your money. If you enjoy these conversations and want to go deeper, check out Colossus Review, our quarterly publication with in depth profiles of the people shaping business and investing. You can find Colossus Review along with all of our podcasts at Join Colossus dot com.

1:00 Mm. Patrick O'Shaughnessy is the CEO of Positive Sun. All opinions expressed by Patrick and podcast guests are solely their own opinions and do not reflect the opinion of Positive Some. This podcast is for informational purposes only and should not be relied upon as a basis for investment decisions. Clients of Positive Sum may maintain positions in the securities discussed in this podcast.

1:23 To learn more, visit psum.vc. My guest today is Alex Wilchko. Alex is the founder and CEO of Osmo, a science and technology company giving computers a sense of smell. He set out on a mission to digitize our sense of smell, and he describes how Osmo is teaching computers to both read and write scent. Alex was kind enough to walk me through the laboratory, which you could watch in the video version of this interview on YouTube and Spotify, where he demonstrates the method to their madness.

1:52 We discussed their first commercial application, a new brand called Generation. Which is revolutionizing the fragrance industry by dramatically accelerating the typically years long process. of custom scent creation. We discuss all the potential business implications this technology unlocks. applications ranging from counterfeit detection to health monitoring.

2:10 And creating a cutting edge proprietary platform in a historically routine industry. Please enjoy my conversation with Alex. So we smell a lot of stuff. Twice a day we run sensory panels where we just sniff stuff and label stuff. So it's just like scale AI has people labeling images all over the world. Turns out like we couldn't just buy that service from anybody. We had to build it from the ground up. So this is where things get tested.

2:35 And low latency work happens. And then when we scale it, that happens elsewhere. So they're literally going through Smelling stuff. Are they like notably talented smellers? We we actually have literal rankings. And so like when we need really accurate data of a certain kind, we will call on our like top dogs, you know? So in there is synthetic chemistry. Just like a drug company to create a new molecule to don't affect human health, we create new molecules that affect human perception.

3:03 And we design those on spec for customers that like a large customer might say All right, we're having this problem making our detergent smell this way, or the regulatory landscape is changing. We can't use this molecule. Can you please help us? So we take all those requirements, we go back into the lab. And we use AI but when it's applied to olfaction we call it OI or olfactory intelligence. So we use OI to design new molecules, which we then synthesize with our synthetic chemistry team, smell'em, they work, we launch them. The smell is so like remarkable and it's like w it wafs and change. Yeah. It's so cool. So it's like a little bit maple syrupy today. Um so I don't know what they're making. I guess they're they're going through a lot of commercial fragrances right now. So

3:42 You're smelling kind of the symptom of our work. So behind you is what's called the perfumer's organ. Each bottle there is like kind of a key on a piano. And instead of eighty eight keys, it's about twelve hundred keys. So a perfumer Will be able to pull any of these ingredients together and mix them in the right ratios to recreate your scent memory. So like a smell of fresh laundry was made by a person. Right, from some of these ingredients.

4:08 А the smell of You know, a clean kitchen. uh was created. From some of these ingredients, so ninety percent of the products in your home. Have a fragrance.

4:18 And every single one of those fragrances was crafted by an individual. And they were crafted by combining these ingredients together. So We're doing at Osmo is teaching AI. about these ingredients and how to work with them in a safe way.

4:31 Do super fast, do super affordable. And um just be able to launch new beautiful scents that were possible. What is the first set of building blocks for doing that? So you've got like the individual I don't know, isolated smells or whatever. How do you create the digital footprint for each one of those things.

4:48 This is A machine called a GCMS. This is basically a camera for the molecular world. I'll just show you kinda how it works. So This is a a robotic autoloader, and so each one of these has a smell that we want to analyze at the molecular level. So this thing can run twenty four seven. So we load this thing up, we just let it run. What happens is you suck up a little bit of the smell as a liquid.

5:10 When you inject it, it goes into this half of the device. Which is basically an oven. With a fifty meter long Very thin cable. And you're shoving the smell through that cable. And what you're trying to make the smell do is

5:24 Um like runners in a marathon, so every molecule. In that scent is all clumped together and you experience that as one kind of unified sensation as a smell. You gotta separate them to analyze them. And so what you do first is you run them through a race and the light molecules make it through the race first.

5:42 And so they can be analyzed one by one here. And heavy molecules come out later and later and later. So this basically separates the scent into each individual molecule that's in the smell. And then this side weighs them. So the molecules enter the mass spectrometer after being separated. And you basically hit it with a

5:59 Yeah. An electron gun. And it shatter the molecule to pieces and you very carefully weigh those pieces and then you play kind of like a sudoku puzzle. To figure out okay, given the weights of these fragments, given how long it took to run this race.

6:12 What was that molecule? And typically this interpretation is done Part by software, part by people, and what we've done at Osmo is make that happen entirely by software. So that's a part of our OI. system. So how much does the individual atomic unit of smell differ from the combinations like if I think about

6:32 Like the col like primary colours or something. These are primary smells, is how I'm thinking about it. Is it pretty reliable, like how you can combine those things? Into some new set of things like what is the periodic table equivalent or something. But we're teaching machines to figure that out.

6:48 Right. That has been the Core core. Issue of why Scent hasn't been digitized is because exactly what you're saying. People have been analyzing the molecular content of these smells for a long time. So like you can go

7:03 Look up in some textbook what the molecules of evil syrup are. Um But the ability to say, Okay, I want maple syrup with a little bit more cherry. Or I want maple syrup, but don't use that molecule because we know it's not safe. Use this other molecule. That requires tons of tradecraft. That is what we're automating. How much will this machine change in the next five years if you're successful? Like will you be building your own version of this? I see Gen Tech on there, it's not, you know, an Osmo machine. Yeah. So these machines are great.

7:31 And what were Not gonna do is we're not gonna change the hardware. 'Cause there's a about a there's twelve Nobel prizes worth of Uh advances inside of these machines. They're fantastic.

7:42 What we've done is rip out the braids, and we've replaced it with our own braids. So a lot of what we've noticed is the hardware actually is already pretty good in the realm of scent and chemistry. But the software or the maps that link the different pieces of hardware. has been completely missing. That's what we built. This is kind of like an inner sanctum here. Um

8:00 So This is where we keep every AI design molecule that we've made, which is probably a significant fraction of all AI design molecules ever. Um so this is uh But just one slice of it. So in this room is, you know, ten, twenty thousand molecules.

8:17 That have all been designed by AI, and we have a digital twin of each. So if we need to go back and access it, we know it's you know fridge one, shelf three, row column two, four. In the sum total of it's kinda smells like a bready radish or something like that. Do you yourself have an abnormally good sense of smell? We've brought a lot of people into Osmo that have like truly world class noses. And so I can say definitively that like I'm not world class. So this is kind of the Rolls-Royce machine. Um it does the same thing as the other one, except you there's two more things that are interesting. One is you don't have to inject a liquid into this. You can put like

8:54 Anything into these vials and it will suck the smell from out of what you put in the vials and will analyze it. By turning it into a liquid first or just directly? Directly. So what it does is it is it pumps air into these vials with a needle syringe, so it'll get dropped in here, a needle will be pushed into it. So basically we'll suck the air and we'll concentrate it onto the You know like Kodak film absorbs light? Do we have film that absorbs scent?

9:18 And then basically concentrate the smell on that thin piece of film. And then you move that needle and you inject it into the spectrometer and it uses a flash of heat to remove all those molecules, it kind of develops the film. And then the normal machine runs, we analyze the data with AI. And then we can pull back out what the scent actually was.

9:37 So this means that we can analyze flowers and vegetables and people and fruits. So what we did the first sent that we fully teleported digitally. Was a a fresh summer plum. So it was like kind of the purple plum

9:53 You know, like the really good ones had like a snap when you bite into it. It was like one of those fresh ones, so we sliced it. We put it into one of these vials, we analyze the smell. And then we actually reprinted the smell on the other side of the lab, which I'll show you. The other thing you can do with this machine, which is really cool. Is you can pause

10:10 the smell at any point in time. You can just sniff molecule by molecule. So A cent will be like thirty molecules, a hundred molecules all blended together, different types. You can spell them one by one by putting your nose on here. So it's kind of like a debugger for software. Wow. So this is called a GCO or gas chromatograph olfactometer. But when we really want to understand the smell and kind of like build our intuition when we're building new protocols, we'll actually sit here and sniff stuff that comes off of the sheet. And so I understand like the strategy behind all this. So

10:39 You've got The ability to read, and then you've got the ability to write. Yep. And that In so doing. Those are the that's just the first step to giving computers a sense of smell. We'll talk more later about all the applications that you could then build on top of that capability.

10:54 But is that how you thought about it that to give computers the capability in the first place. It's read. And right. And write is especially important because it confirms whether or not it's being read correctly. Exactly. And if you can read and write, then you can create this virtuous cycle.

11:08 where you're creating data at every run of the loop. Right. So if you actually can Create new spells. And then you can turn those smells into data readings of some kind. You're training AI. Right. And then if you can tilt that process, so the next smells that you create the next day.

11:24 Teach the system even more. That's when you're doing what's called active learning and that's how you get AI systems to get smart really fast. And that's what we do. Can you talk about the measurement of the fidelity gap between read and write? Like if I give you uh if I give you the the plum. And you stick it through your machine and you read it in and then and then you give you the oil, essential oil on the other side. How how you measure the gap between the smell of one versus the other and how close you are? Yeah, it's like

11:51 I'm simplifying, but I'm gonna hand you The real and the recreated and you're gonna tell me We there's a more nuance to how we do that to Create data that can actually be fed into a machine learning system, but that's effectively it, which is like, do these things match? And there's a few tricks that you use to help de bias people, get reliable data, but like

12:11 You're the arbiter, right? If it's like a smell you're familiar with and I'm trying to recreate a memory that you have, like We either did it or we didn't. We were just with a very famous uh Hollywood person who said in the eighties they tried to do this in theaters where they would have something that like puffed out smells. Roma Rama. Yeah, and it just didn't work. They only had certain smells that were like whole scenes. And so they didn't have primary odors. They didn't have the ability to create any smell. And so here on this robot, you're obviously not putting this behind a couch cushion. Yet. But we're gonna make this smaller.

12:41 But the idea here is you need to have all the ingredients together that can be mixed on the fly to create any experience. Not just like eight pre-programmed experiences. That's like a slideshow, right? We want An actual display that can show anything. What is the most surprising thing about

12:56 Primary smells. We're kind of at the scientific frontier. And so like everything that we discover every week, every month, like pushes back what's known about smell and how to construct it. Um I think Yeah. One thing that I've found in doing

13:11 science and machine learning and combining these things. Is like Problems that people sometimes think are totally intractable. Once you just get started, you're like, Oh, we're actually making progress. And so the idea of like

13:23 creating sense with AI and creating those sense in partnership with people and like fusing human and machine to like work in this very emotional world of scent like People like you just don't think of it. It's like start making progress. Then you start making progress bit by bit. And the first sense are like dumb. They don't work or they don't smell right. And then like you come back four weeks later, you're like That one was really good. Holy crap, I think it's working. And then they all start to work. And then you start to go talk to customers and you start to accept some of your Sense for products.

13:54 And then it's like really starts to roll. And you know, bit by bit you just Yeah. Keep going. You know StockX is they have a problem with fakes. From Timu. I want you to hold this in your right hand and smell inside of it. And I want you to hold this.

14:09 In your left hand and smell inside of it. And I want you to look at them. Can you tell the difference between these? I mean there are Not really. Yeah, they're the same, right? Like they're constructed to be perfectly the same. This one smells like new shoe. That's

14:24 The fake. You can't work a stock exactly. So the difference is The counterfeiters are really good at visual identity. But The smell of the shoe is basically the fingerprint of everything that ever happened to make this. What we've been able to do with Stock X is show that we can take s those really big sensors that are in the lab.

14:46 Cut the right corners, make them smaller, so it's actually the sensors are about the size of These two shoe boxes together, that's one right there. And what we do is we can take the thumb hole of the shoe box and basically insert it into a a sniffer. And it will within twenty seconds it'll tell you a real thing.

15:03 And we just have to show it. А філс, а фі фейкс. Сем тайс ісмор, сам та с. But then whenever a new ski shows up. We grab some of the fakes, we grab some of the reels, we train it on the new skew. And then now there's a device that can tell them apart. It's funny to imagine a future where there's an arms race and the and the counterfeiters are injecting their own

15:21 Yep. There there already is an arms race. And so this I'm sure is gonna be the next frontier because we're about to really stem the tide. The first investor prospectus that we made the Double edged sword was like okay, so like we have this entire space to ourselves. Now the risk is how do you focus? This is kinda like a piece of history here on Plum one point oh. Um this was the first sent that was teleported. So the first scent to be digitized and then reconstituted in another place.

15:53 And what we've done with this and I will Show you the vial here. Is this is the essence of a fresh summer plum. And in this vial, this kind of clear liquid

16:06 Is probably Thousands or tens of thousands. of sniff's worth of this one moment of biting into a fresh summer plum. And what we're showing here is literally everything the source code. Of that.

16:19 memory of that sent experience. So I want you to smell it. So I've already dipped some. in the vial. So just put the blodder into the vial. But close your eyes and think of biting into a fresh summer plum. Mm.

16:34 Crazy. That's wild. Right, did we get it? Yeah. So this is I I think a piece of history and like we we've we've only made a hundred of these, but I'd like you to have one. Oh wow, amazing. It's important to me that what we've done and um I think uh you know It's just a plum today, but it's a lot more that we're doing in the future. So one thing that we're doing.

16:56 is we're launching a new kind of fragrance house uh called generation And the idea is to take all the technology that we've built, but also all the humanity, the people, the perfumers, the noses, and to combine that. in a new way of designing scents for people who might not have been able to have access to designing a new scent. or haven't been able to do it fast enough. And so that's called generation, and that's something that we're just now launching. What we're doing

17:22 With this scent is this is a scent that we custom designed for a creator. Це саму аудієнс о інстаграм. has like a really great rapport and a brand, frankly. But What she doesn't have access to is a way of creating her own product that is resonant with her values, but also just is straight up beautiful. And so what we're able to do is take all of our technology, take our perfumery.

17:47 And what we've done in very short order is design her a fragrance and we're gonna help her launch it. And so Part of the value proposition of generation is Do you want to launch a fragrance? Do you have a place to put it in front of people, but you're missing all the pieces? Because we can now automate lots of this and we can bring humanity to the rest of this.

18:06 Like let's work together, let's build you a fragrance. So if I wanted to go through that process and said I want to create my own. building blocks of that process. Like I could I guess I could start to describe I like plums and I like fennel and I like this. Yep. It can be. So like let me tell you how it's done today and then how we're changing it. So the way that you get a fragrance design today, and I'm not even talking about launching the full product, like literally just the smell. There's more that you have to do to launch the full fragrance. You submit what's called a brief.

18:35 So briefly. Um in other industries would be called a request for proposal or an RFP. And It can be anything. So it's actually very, very freeform in this industry. That itself could be revolutionized. So let's say you write a word document, you describe your brand.

18:49 Um You're gonna describe what you want it to smell like, who you are, what you want the brand to be resonant of. And then at the bottom, you'll usually specify two pretty important things. What's the benchmark? So is there a scent that you want to beat? And usually that means I kinda want my thing to smell like this, but make these changes. And then what's the price you want to pay? So how many dollars per kilo? Which can be as you know low as like you know, ten bucks per kilo, it can go for fine fragrance can go up into the many hundreds of dollars per kilo.

19:18 Just depends on what you want to launch. You submit that brief to a fragrance house. And uh and again this is the traditional way. Somebody receives that and in their kind of weekly meeting, they read the brief. They read the volume that you want to make and the price and they decide whether or not they want to work on it. And here you haven't paid the fragrance house anything.

19:36 Uh they are look at the brief and then they say okay we want to work on this. Usually ninety percent of the time they go to their library and they say oh we've already made something for somebody else. Let's send Patrick, this scent that is from our library. And then you'll get that and maybe you like it, maybe you're done.

19:54 But ninety percent of the time, sometimes more. Um You aren't getting a new custom sent, and by the way, that process itself. That may take. Weeks, months.

20:04 Um, but let's say you push back and you say, actually, I want something that's really custom. So you know what you you're a sophisticated buyer. You said don't give me a library sample. Now you're looking at like a 12 to 18 month process of going back and forth. Every time you submit notes, it might take three months for them to get back to you. Super long process. And by the way, once you get the fragrance, you still don't know if it actually works in the application that you're going to put it in. Right? So you have to do application testing. So let's say you want to launch a

20:34 Uh A skin cream. And you want it to be slightly fragranced. Well the fragrance can't make the skin cream turn a wrong color.

20:41 And it can't make uh the scent change too much. So you have to do application testing as well. Now you've added more time. So that whole process is handled In most fragrance houses, something that looks very close to Pencil and paper.

20:55 And a lot of guesswork. And so what we're doing with generation is taking each one of those pieces and applying modern methodology. In some cases AI, in some cases just You know, efficiencies. To make that

21:07 faster and to make sure that you get something custom every time that's actually tailored to your brand. Right. So when you submit the brief That should be a chat GPT interface. You should have a conversation, right? That should be available to start instantly. So we have a a tool. So let's make but you want to make a Colossus fragrance? Sure. Great. Um a fragrants For Colossus.

21:28 What do you think A vocal I'll describe what we what what our mission is and and maybe it'll come out of that. So Our hope would be That by finding

21:39 I frankly, people like you. That А і персут. of what we would call their life's work. for you it might be giving computers a sense of smell and all the all the applications that are born from that.

21:52 That we By showing people These great examples of people really doggedly on the hunt to build their thing. They'll wonder what their thing is and start building it. So I would say

22:04 Uh very focused on business and investing. Those are the those are the forms of art that I like. But I would hope that our work encourages more people to chase their thing. because we're showing them such great examples of people like you chasing theirs. And really just give them permission to do so. So what is that what do I hope evokes possibility. Like I think of open air, maybe would be like a smell that I think about. The Sequoia Parks or the Redwood Parks in San Francisco.

22:29 When you get to the top there's like a very specific, like crisp smell. So What's happening behind the scenes is we're tapping into all the tools that we've built, the olfactory intelligence that we've built over time. We will take what we've put in and we'll embed it into our map of scent. Mm.

22:47 So our map is not three dimensional like R G B it's about three hundred dimensional, which is I think why Scent had to wait for artificial intelligence to be digitized. It's because it's just more complicated. And will then decode that. coordinate into Ascent profile.

23:03 We'll show you where the scent lives in the map of like the hundred Top mass market hits. And then I'll show you the source code of the fragrance. And If we have Something similar to it, we can actually go grab it and smell it.

23:14 So And this is by the way, this is a multi month process that we're condensing down into You know, minutes. How long do you think it will be until There's literally something sitting here.

23:26 That the the feedback loop is more or less instantaneous. The mountain peak is something you can hold in your hand, like your AirPods and your phone. They can read. The chemical slice of reality. Capture sent moments.

23:40 Tell you if you need to go to the doctor, tell you what vitamins to take. To really read scents. And then another device that can recreate it. So that we can fill this room with It can smell like walking through the redwoods, which

23:52 Like that forest bathing smell is one of my favorite smells of all time. Um so we're getting there. So you saw the big reader. You saw the one that we made smaller. We have to take that down by a factor of like four to eight in order for this to be something you would really say is portable. Right now, we can move it around, we could deploy it. I don't think you can say it's portable today.

24:12 And then you saw the sent printer, which is half the size of this table, right? We've got a lot of work to do to make that thing smaller. But if we've known one thing about technology is making things smaller is like a thing we can do. Um So let's take a look. So we've got an image. Ambition trail. Ambition trail. All right. The marketing copy is a fragrance inspired by timeless determination in the pursuit of one's mission, capturing the essence of ambition and discovery. It seems based on all this that it's not that long from now that you will enable things that traditionally have sense, candles. Detergent products.

24:43 Whatever. Uh infusers. Because they're all based on the essential oil that you're delivering. Yep. That like People will be able to design sense very soon.

24:52 Absolutely. So that We are rolling that up. Our Business To

25:02 open up scent design to more people, right? And to do that faster. If this company is gonna be the thing that sort of has your name written all over it, and that you're the most proud of having created. Yeah. I mean Because You know

25:19 Um David Senra's in your world, uh he has like a few phrases that are like really resonate, which is like the exit strategy is death, right? Like this is the last thing I wanna do. I grew up in um A town called College Station in Texas. Um Not too big, not too small. I got bit by the computer bug pretty early, so started programming computers when I was like eight or nine.

25:41 Um full on computer nerd by twelve. when I started collecting perfume. And I started collecting perfume because I noticed That People

25:52 when they put this invisible thing on them. Would all of a sudden be treated differently by everybody around them. But within this little radius, right? So like It just

26:02 It was this magic Potion spell combination that When you just say it like that almost is unbelievable, which is Can you Spray an invisible thing on you.

26:13 That changes how people see you and treat you. For the better or for the worse? I I just couldn't understand it. I like I had already felt a little bit like a social outsider. And I was trying to decode

26:25 That like why are they popular and I feel like I'm on the outside. And so I looked at the clothes, but it was the fragrance that really like But It confused me to no end at first, but then it fascinated me. And so I started looking into fragrance and I found out what these you know these kids my age were buying. It was Polo Blue, it was Abercrombie and Fitch Fierce.

26:47 Um both fragrances that our master perfumer that we just passed designed. Right, many years ago. So it's completely full circle now. Um But then I realized there's not just two or three fragrances, there's Tens of thousands.

27:00 And it was like Discovering that movies exist, and you can go to the movie theater and there's nobody watching the good films. Everybody's watching the popular films. And for me, this whole world opened up a fragrants where I mean I remember the first fragrance that Really

27:17 taught me that this isn't art. It was Bulgari Black. Which is frankly not like a very long lasting or particularly performant fragrance. It's like um it comes in a bottle shaped like a hockey puck. And you spray it on it lasts forty five minutes. But would it does is it unfolds in three acts.

27:34 Right, so the first smell is like the smell of screeching tires and rubber. And then within five or ten minutes it cools down to like of this vanilla rubbed on a leather chair. And then all of a sudden, after like another fifteen, twenty minutes, there's like this smoky tobacco leather chair smoking room vibe. And

27:52 I m remember the first time I experienced Who This fragrance changed, has it gone bad? And then I just sprayed it again and again and again and watched this movie play out for like An afternoon I was like No no somebody made this And it this is this is the whole thing. The whole fragrance unfolds over time.

28:08 And that kind of was the end of it for me. I got just completely hooked. And You know, I the way that my brain works, I wanted to understand like where it came from, how it was made, how the brain processes it the way that it does. I think if I was born in southern France, maybe I'd be a perfumer.

28:25 But I was born to two academics, and so I became a scientist. And um I went to school for neuroscience um at the university misigan. And then realize that there is a sub specialty of neuroscience called olfactory neuroscience. So people who want to figure out how the brain processes smell. And the most people who study that are at Harvard, so I went to Harvard.

28:47 And realized after you know many years of doing science there that like Actually we don't really know how smell works at all. Like we're making progress, we're learning things, but like A simple question like Let me draw a uh a molecule on the whiteboard, like we're in chemistry class.

29:03 Can you look at that molecule and tell me what it's gonna smell like? Will it smell like apple or cinnamon or You know, Anis or what? So it turns out that's like a hundred year old problem nobody'd been able to solve. That really stuck in my crown like why don't we know how to do this? Um

29:19 I ended up leaving academia. I started and sold two AI companies, um, one in the biotech space, one uh in the kind of pure ML as a service space. Um That ML company was bought by Twitter. I helped to start their deep learning team with um my co founders and with another company that we were combined with. That's where I really learned like internet scale artificial intelligence applications. So we

29:42 applied AI to their ads platforms and made them a lot of money and uh applied AI to their data centers and save them a lot of money. Um And then I I was recruited away to Google Brain, which is now called Google Deep Mind. Which is

29:56 Kind of their Xerox Park or Bell Labs. And um I worked on some internal projects for a bit, but after a year or so I said you know what Like let's take a crack at the smell problem again. And it turned out that in the time between when I left academia and got into tech and entrepreneurship and when I arrived at Google Brain, some breakthroughs happened.

30:15 And what happened is AI researchers figured out how to make Artificial intelligence work on chemistry. And that maybe doesn't sound to crazy

30:26 But Up until then AI systems really liked their inputs to be rectangles. Right, like images are like grids of pixels and text is like a long thin string of words. But molecules can have any shape. There can be any number of atoms and the bonds can be all rearranged.

30:44 And There had been a technique that had been really improved and figured out made to work better called graph neural networks. And that turned out to be like chocolate and peanut butter. for AI and chemistry. So we didn't figure that out, but a lot of my colleagues, so I was very fortunate to work with at Brain, they figure that out.

31:02 for the world of drug discovery. So the intellectual arbitrage that we did was we said Let's take those techniques. And let's apply it to the realm of scent. And I had spent a long time thinking about scent and traveling in that world, and so I knew where to get the data sets, where to buy them, where to license them, how to treat them. And we fuse those two things together. And I was fortunate enough to work with an incredibly talented team of folks at Google Brain, and we made this happen together.

31:27 And what we're able to do is solve this hundred year old problem. It sounds so simple, but like why does this molecule with this shape Smell the way that it does. And we validated it in a really stringent way. We basically did a double blind trial. Where

31:43 We predicted the smell of Hundreds of thousands of molecules. We picked four hundred. that were very different looking from anything we'd seen before. We kept our prediction secret. We bought or made the molecules. So some of these had never been made before.

31:58 We sent them to our collaborator at Monell, um Professor Mainland was running this. And She trained a panel of people, and this is kind of like what we do now, but just um initially it was at a smaller scale. Train people to smell something and say, Okay

32:13 This smells fruity. And mineral. And that's it. So I'll give it a a three out of five fruit, I'll give it a one out of five mineral, the rest zeroes. And that's called rate all that apply. It's just like that's how we label data. And then what we did is we said, Okay.

32:28 We have our predictions. People have their double blind ratings. Where do our predictions fit within the people? 'Cause the best is the average of the panel. That's how you get really high quality data for AI. So were our predictions worse than the worst person?

32:42 Or were they in the pack somehow? It turned out that our AI predictions of what these smells We're going to be. were better than the average panelist. Meaning, if you're gonna add one more person to this panel, you'd actually prefer to ask our software what it smells like that doesn't have access to the physical molecule.

33:00 than to train up another person to physically spell it. Which is kind of like passing an odor touring test. When that happened It was very clear that mother nature was not gonna stand in the way. of continuing on this journey of actually digitizing the sense.

33:15 So if you can solve that one problem. It means you can start to ask, okay, great. Now what happens If instead of feeding this AI algorithm a pre-digitized Molecule.

33:27 What if I feed it? A reading from a sensor. like the data off of a camera, right? If we were talking about images. And then What if I then ask it to recreate that smell?

33:37 Right, with uh the ability to mix together different molecules to create a new scent. If you can actually round trip. A smell. So take a physical smell, put it in one system. And then round trip through the reader this map that we built, this graph neural network based map. And then write it back out again and then compare it.

33:55 And it actually smells like The thing that you Put in. It means that you have actually digitized A human sense.

34:04 We hit all of our scientific milestones at Google and we asked ourselves, What's the right way to scale this idea? And that's where Uh Josh Wolf comes in. So we were thinking internally at Google maybe this should be a company. And I was working with Krishna Yeshwant at G V who's a a very close friend. We'd worked together for five years.

34:24 And someone at G V uh another investor named Izzy Rosen. was getting lunch with Josh, uh who's the founding and managing partner at at Lux Capital. And apparently Josh had this ten year long thesis about digitizing all faction. And we had never met. And so Izzy was listening to Josh give this pitch again, he said, Hey, have you talked to this guy, Alex? He's kind of in the smell.

34:45 And then Josh and I met and it just was an instant connection. And so Josh was integral. In pulling this IP out of Google Brain and building it into a completely new company. So Josh led the round. Uh Krishna at G V Coled and we

35:02 Build Osmo. Um and We're on our way. When you think about building Osma the business. How do you think about the trade off between A P creating a pure play platform.

35:13 Defined as You enable develop I'll call them developers. to build any sorts of application they want on top of Osmo's root level capabilities, all the things we've talked about. And You know, you charge them a platform fee and

35:28 You know, lots of platforms out there that are are wonderful businesses. Versus like Okay, we have the platform, but we're also gonna create the generations, the vertical application companies on top of our own

35:40 Raw tech capabilities. What are what are the trade offs of One approach versus the other. Are they mutually exclusive? I I think in the limit we're gonna be able to explore that.

35:51 design space more fully. But it it really depends on what's the market you're entering with that platform capability. And so a lot of successful platforms are just are entering markets where there's already a ton of vibrant activity and they're helping to You know? grease business and make that um happen more fluidly. There's not a lot of fragrance companies out there.

36:12 Right, so there's not that many buyers. Um The question is like Do you become a software provider for the incumbents? Or

36:22 Do you take your capabilities And Do you compete in that market? And I think there's been examples on both sides of this. There's plenty where you are an input or a service provider. I think a recent example where folks decided to just enter into compete would be like Metropolis, if you've heard of that example.

36:41 Um I can they're making software for managing parking lots, um, the parking lot industry But it actually worked. It made parking lots more efficient. So they became a parking lot company, right?

36:56 We went through a similar journey where we I mean If I could have sold the software here, and believe me, we've tried. Like yeah we

37:06 Um But there's not that many fragrance houses, period. And I think that we have the opportunity not to just sell into an industry that is like look very large, it is very secretive. But I think with this software and these tools, we have the chance to really transform it for the better. Like the way that business is done here hasn't transformed for three hundred years.

37:26 In a lot of the way that things are done. should stay the same, right? Like they've stood the test of time. But Like the world's getting faster. People are asking for more transparency.

37:36 Um People want to make sure that the fragrances that they're using are safe. And also there's a lot of people who still don't even know how to get a fragrance made. Right. And like look. Every company has visual branding.

37:50 They have A feeling that they're trying to create. for the people that interact with the company, for the people that are in the company. But there is no modality, there's no sense that is more emotional. And that has deeper ties and associations it can build than scent.

38:05 So there's a lot of businesses out there that Need to have a smell. It it's already happening, right? Like the Ritz Carlton has a cent. I remember the Gramercy Park Hotel so so distinctly. Right, and you walk in and what do you feel when you walk in? Yeah, familiarity. Yeah, familiarity, it's elegant. Right. It's like if you smell it anywhere else, you're gonna think of exactly that.

38:25 Hotel. And so I think one thing that we're realizing is there is an appetite. To add scent to more layers of our economy, to more businesses, to more markets. To more products. It's just inaccessible.

38:38 And so what we're trying to do is bring more people to scent and create more sense for people. So that's what generation is all about is If you want to make a sent. And if you want to do it quickly. And if you want to do it safely, like

38:52 We're here, like we figured out how to fuse AI with The human aspect to create really beautiful smell. Teach us just a little bit about smell itself. W what's its history? Why is it so important? Why is it so emotional? Why are scents so memorable? It this sounds like hyperbolic. It sounds extreme, but It's the first sense.

39:13 Right. So if you think of us as little microbes a billion years ago. We survive by eating things. And we got better at surviving by eating things by detecting if the thing we want to eat's nearby. That's what smell is, right? Smell is like sipping little amounts of the chemical environment. Around us to figure out

39:30 Where is there more of that thing or less of that thing? So It's a super old smell. You can even see it in the brain. So If you smell something, first of all.

39:41 That is the physical world touching your brain. Right. Sends neurons out of your skull. into the top of your nose and your brain is literally touching the world when you smell something.

39:55 And the number of steps it takes for that information to get to your centers of memory. And your centers of emotion. Існе, со ви анатомили Wired. To have sent project directly. to our memory and to our uh centres of emotion.

40:11 Um Those areas are called the hippocampus and the amygdala. So we're wired. To associate smell with emotion. So it's it's like evolutionarily super

40:22 Old. Um There's still a lot of mysteries that remain about smell. There's amazing researchers that are pushing back the frontiers of what we know. figuring out why things smell the way that they do, of engineering better smells. So like

40:35 Look, it's still also the most mysterious sense. Um We were talking a little bit earlier. It feels weird that we haven't figured out. Sent.

40:45 Right. It's like computers can see Computers can hear, they can touch, right? We have touch screens and we have um the work that uh formerly control labs is doing with these sure wristbands exact haptics. But computers can't smell. And

41:01 That's weird because it's such a fundamental thing. It feels like almost free or like easy for us to smell things. Like why can't we teach computers to do this? There's this concept called Morovek's paradox. Have you heard of it? Yeah. Yeah, so like just really briefly The idea is like the easier it is for a person to do, the harder it is for a computer.

41:19 Because if it's easy for a person, it means evolution has spent a ton of time making it easy for us. But if something is really hard, like proving a math theorem or something. Turns out we've taught computers to do that stuff. Right, and it was weird that those were the first problems to fall.

41:35 Um but like walking. has been hard and we're just now kind of getting good at that. By making robots walk. And smelling has been extremely hard and we're just now kind of cracking the code there.

41:46 So if I think of generation as Um The creation of and manufacture of any smell that I want. You know, with with sort of no limits on the possibilities of how I could use that scent in a product in a space in a showroom in a whatever.

42:03 What are the other like next two, three, four things. Where there's a big stack of potential utility that would be unlocked. because computers can smell. Yeah. I think of like dogs in an airport or something like that. Totally. So that's what we're doing with StockX. Then again, like Our main thing is generation. We think that that's gonna

42:24 really have a massive impact. That's creating smell. But there's all these applications for detecting smell. So I don't know if you've ever experienced this, but

42:33 A lot of people Can sometimes smell if they're a loved one or their partner. Is getting sick. Or like something's a little bit off. And then two days later they actually get sick.

42:43 That's real. Right? Like what's on the inside of us. gets to the outside. What's in our blood and in our organs is eventually exhausted through our breath, our sweat. Whatever. And we know that dogs can pick up on that stuff.

42:57 So We we've proved out that We can use the sent of a product. To tell whether it's a real or a fake, basically tell its provenance. And that's something we've got deployed at StockX and that's going quite well.

43:10 We think that there's other counterfeit detection and and kind of truth and Um And safety applications for scent. But I think it goes deeper than that, right? So I think, you know, we could be detecting harmful substances at the border and stopping them. Whatever sniffer dogs are doing, I think eventually a computer will be able to uh either help with or or do entirely.

43:31 But A holy grail for us is um human health and wellness. Right. the signal that is inside of the scent that we emit. It's completely untapped.

43:41 Um, it what's weird is It turns out that by getting really good at designing The sense of fruits and flowers and vegetables. You actually

43:52 For free get good at these Other scent problems like with human scent or with product scent because The overlap. of the actual molecules that you see is actually pretty high.

44:02 So there's not an infinite number of molecules out there. There's a lot. But If you get really good at one domain of scent, it turns out to help you in other adjacent domains. Last time you and I spoke, we were looking up together on our computers the market cap of the fragrance houses. Uh they're quite huge companies. No, they're big. The sort of like margin profile of these things. Like what what what is it that makes uh fragrants just in the publicly traded ones that you can go check out. Relatively big

44:32 Good businesses. a a first uh brush are recession proof. So if people aren't buying luxury fragrances they're buying hand soap.

44:43 Right. So Fragrances in ninety percent of the products in your household. And there's a very small number of companies that provide all of that. And so if one category's going down, another is typically going up. So really, really great.

44:58 long term profile. Um The Margin profiles are also very great. So fundamentally these are manufacturing businesses. Um with

45:08 non manufacturing margins. Because there's a ton of know how that goes into producing the finished blended product, so although it's just ingredients mixed together in a jar that's then sent to a customer who then puts it in their packaging. how you get the exact right blend of those molecules.

45:26 Is Uh typically a deeply held secret. And what we've done is studied the industry very, very deeply and figured out okay No. a lot of what is being provided in the industry we think can be augmented.

45:40 um by artificial intelligence and we can do this faster. Um The Other piece here is customers are typically quite sticky. So

45:49 If you're running a beauty or a CPG business. And you Run out of stock. Your first inclination is to reorder from your past supplier not to bid out again.

46:01 Um and so typically if you've won the business um industry um Standards for repurchasing are well above fifty percent. So if you build this very wide book of business um that has different parts that fluctuate based on the macro. Um, you have a really resilient

46:19 business there. And the margin profile is typically quite good. If you think about the Things that could go wrong on this on this journey of the Which I always do, every day. What do you think gets in the way? So I'm always paranoid. When

46:35 Mother nature is gonna show up and say, You're done. Right. No in twenty twenty five or twenty twenty six or twenty twenty seven. This is not the year for you to. Peel back another mystery of how this human sense works.

46:50 And so you're blocked. for taking the next step. And that could manifest in any number of places. We could fail to make these sensors small enough to be Uh held in your hand at an appropriate. Price.

47:02 Um, there could be something fundamental we don't understand about the world. This is an existential kind of this is an existential risk that we can never really remove. But We continue into the darkness and into the fog regardless. What we're trying to do. is never lose sight of the mountain top. Which is we fully digitize the human sense and it's personal, it's portable, it's affordable. And

47:25 Our philosoph for doing that is not to climb up the sheer face of the mountain to that single goal. But to find a route up that mountain with a shallow enough grade. Where А сам поиске бізнес. Філософі хіра вже. I've seen other startups can fail to do this is

47:44 Build along a responsible path. No. makes you harder to kill over time as opposed to makes your likelihood of success even riskier over time. This is just 'Cause I wanna do this for my whole life, I don't wanna just flip this company and sell it. Like I really want this

48:00 To survive. It has to survive. And so we're we're building in And our strategy Uh

48:07 A way to make that. Much more likely than not. One of the things that's so interesting to me about Osmo is It isn't AI company in a very strict sense. Oh yeah. If you look at the org chart, you're like, oh, it's an AI company that married a chemistry company. But but it's quite distinctive in the sense that most AI companies, especially building

48:26 I'll call these applications. are remixes of a lot of the same stuff. There's a lot of code and traditional code involved, there's software involved, there's Um Yeah, putting things around the incredibly powerful reasoning models that now exist and the whole world is kind of

48:42 I saw the other day that uh that ChatGPT now has 400 million monthly active users, like five percent of the world's population is using ChatGPT. So people are now familiar with these like language models and these generative models, but this seems like You're using the power. Much different.

48:58 And I'd love to just kinda describe some of the ins and outs of that. As we think about other problems where we can apply AI that aren't just Text generation, image generation, video generation. Um that aren't pure generative in that sense.

49:12 and pure software and get into some other world. So maybe here it's chemistry and an AI. But maybe describe like how how you're using the tools and how you think about the growth of the capability of those tools and how it will impact what you do. Technology usually proceeds on an S curve. Right. It sucks. It sucks. It's getting better. Oh my gosh, it's getting better super fast.

49:33 And we're pretty much done. And it levels off. And I think we're pretty close to the right side of that S curve with text. Like

49:42 We kinda blew past it, but yeah, we passed the Turing test. Right? Like I regularly am fooled and curious whether or not this was written by Chat GPT or not. So text works and then Ilya Setkover who is really

49:56 progenitors of modern AI for text. got up at the main AI conference in Europe and said look there's one Internet and we've trained on it.

50:07 Right. There's no more data. Right. So yes, there'll be some remaining tricks. We'll make it cheaper, we'll make it better, we'll add reasoning, but like we're out of the raw fuel that drove a ton of the innovation in text. And I think also similarly for images, right? Like we've downloaded all the world's images and all the image models are trained on all those images.

50:27 Video we're not done yet'cause it's super expensive. So like we're not quite at the end of the curve. Those are just three modalities. Right. There's drug discovery. Right, there's chemistry, design of chemistry to treat diseases.

50:40 There's you know materials design. There's all kinds of things. And what I'm concerned with at Osmo is marching up the S curve of a human sense. Right, so we're on text. We're on vision and images. Those are handled by really brilliant people.

50:57 But As far as I can tell We're the ones who are driving AI up the S curve. For cent. And

51:05 We're really at the far left. So we're just starting to take off right now. And the thing that is the fuel here is data. And so

51:15 Yes, we use specific kinds of AI models. We even use LLMs in some of the work that we do. They're super useful. Um our philosophy is the right tool for the job. And so we're not gonna take a dogmatic approach and try to shove everything into an LLM, although LLMs are extremely useful. for this and I think actually they will get more capable over time.

51:33 uh for even what we do. Um But most of what's under the water line, most of the iceberg as as it were, is just creating the data. Right, like having the infrastructure for accepting that data, having the operations to

51:47 created you know the physical and the digital all linked together. Um You have to have the data. Like that's the fuel that drives you up that S curve. And so you know We've just realized this step by step.

51:58 So we started out saying, hey, we want to digitize smell. Great. Where's the data? We tried to do some licensing deals, we were successful. The data wasn't kind of AI compatible. Uh, so we said, Okay, guess I o I guess we have to create it ourselves. And so then bit by bit we begin to build basically the entire AI ecosystem That exists for images, we built it internal and proprietary for the sense of smell.

52:20 So we have A building of people that label spells all day every day. Right, we have a laboratory full of sensors the twenty four seven are dissecting sent down to the molecular level.

52:32 We have robots that are creating smell. We're gonna get even bigger robots to create even more smells. So we've created the entire virtuous cycle here. That allows us to um you know build the data that allows us to train the models to bring these capabilities to the world. How much of the tooling do you use? Like are you tapping models from the major model providers or or anyone else, and if so, how? It's suffused into everything that we do. And again, right tool for the job. So like I don't think there's any code that we write that isn't at least touched by AI in some way, right? It's it's just like that's the new auto complete.

53:05 Right. It a ton of boilerplate stuff is just no longer relevant. It's amazing, right? So productivity is higher. So like Osmo's probably a smaller company as a result of all these AI tools. Um

53:17 There's a bunch of stuff that we just need to get up to speed on, and we can just ask chat GPT for like a reasonable eighty twenty answer. um like, hey, is there any problems with this NDA? Right? Like we can just do that. And then if it's really critical, we obviously like get a an informed opinion. So yeah, it's suffused into almost everything that we do. What excites you most? about the frontier that you're exploring. that you have to sort of hold yourself back from spending time on because you're focused on the things you are. Yeah.

53:46 So I I love everything that we do. And If it isn't clear, I really like smell. And I really like making smell and experiencing it and sharing it. And so Yeah.

53:57 generation launching generation is Kind of a dream come true. But it's not the last thing that we'll do. Um What's really wonderful is

54:06 by building all the systems that are gonna allow us to create beautiful scents for folks and generation. That platform is gonna help us. on our mission to understand human wellness with scent. And that is one of the next mountain peaks for us. Is

54:22 Yeah. Mm. What In the smell that we are exuding. contain information that help us make better health decisions.

54:31 And there's already a link there between what we do at generation and what we will be doing on our um journey for the next R D frontiers. Scent has powerful effects on our mood. And there's already a very deep tradition of aromatherapy, and you know, look, this science I think needs to be expanded there, and I think that will contribute to that.

54:51 But You know, as we're turning emotions into data, incent into data, and then data into products for people. um to build better businesses, to build launch better products. Like All of that goes into one platform that is gonna help us push back the frontier.

55:08 concepts behind technology platforms is that it's very hard to predict what people will use them to do. Yep. And it seems like every time you digitize anything in technology history. crazy stuff happens that we that we can't predict. Totally. And I'm sure that's gonna happen here too. That's the idea, is like we know what we need to do now. Right, we know the markets where we can be valuable now. And so we're not gonna

55:29 Waste any time. On anything else other than Building a business that Makes people happier and makes other people's businesses run better. Make sense more beautiful sense faster.

55:41 But it's really hard to predict a future, right? Like Computers haven't had a sense of smell. Like I can't see on the other side of that wall. Like what happens when our sensors actually can fit in your pocket. Like what happens not just to the products that we build or the partners that we have that are building on our platform. But like

55:58 What happens to society? Right. Society's different'cause That computer in your pocket can see in here. It's like I think it's largely better.

56:06 Right. There's more information flowing. You're remembering more things, you can hold on to moments, like there's some beauty that's there that wasn't there before. And I think that There will be Beauty.

56:19 In what comes out of what we're doing. As one's young. Uh not that old. Can you describe the story of what you what you would define like the defining moment so far in the company's history?

56:31 I think on the The journey. Of asking whether or not what we hope to do is possible. Like, hey, we've got this crazy dream of actually digitizing smell. Is that possible?

56:44 Like we had to get together this crazy team that had never been assembled before and then tackle this technology that had never been built before. And then when we actually Teleported The plum the first time, and we smelled it.

56:57 And it was actually A freaking Um And it was beautiful and almost like hyper real. I just

57:05 fell out of my chair. I mean that was like a really dream like moment. It's like this works, right? Like Yeah, we are we we did it. Like computers can smell now. Like it's in the lab. We'll make it. smaller and cheaper and better, but like It's no longer zero. We've gone from zero to one here.

57:22 And then There's All the stuff that came out of that, like all these capabilities and tools. That were turned into the ability to design scent even better and capture other scents.

57:33 train the AI models and kind of build the entire olfactory intelligence platform. Um But like that moment when I smelled the plum was Really special. I think what you're building is is

57:45 Singular, very unique. There's there's no company that I've really encountered quite like this one. And what I find so cool about it is that it's taking advantage of the technology that everyone is so excited about. So thanks for letting us in today. It's been an incredible. experience seeing around. I can't wait to see the next iteration of the smells and of the factory and and and of all the machinery. When I when I do these I ask everyone the same traditional closing question.

58:08 What is the kindest thing that anyone's ever done for you? Moments. When People could have closed the door. on me or said no thanks or

58:21 I don't believe you or I don't want to take a bet on you. I really divide the people that took a bet on me into the three chapters of my professional life. First In my academic life. Josh Burke.

58:35 And Then Bob Data. took a huge bet on a no nothing kid. Two mentor and to grow as a scientist і I'm

58:45 indebted to them for taking the time and uh kind of taking me, you know raw form. In helping mold me into uh a scientist. And Then I I know I moved from academic science.

58:59 Into industry, into entrepreneurship. And there I I have a a number of people to thank as well. Uh it's Brian Adams, first of all, for taking a bet on me to co found a company. with him, which we ultimately sold to to Twitter. And then when I move to Google Brain D scully

59:16 with really no reason believed in this idea of digitizing Wolf action before anybody else did. uh Google Brain. And then Jeff Dean. saw what we were doing and said, You know what? This weird little thing, let's just let this flower grow. Let's see how it goes. And so Jeff Was instrumental

59:34 In kind of giving us cover or a force field, just to grow this very delicate young idea into what it's And then when we converted from an industrial research project

59:46 And we decided that the right way to scale this was a company. A whole new cast of characters. Um Took a bet as well. Uh, where there was really not a lot of evidence that they should have.

59:56 Uh. I'm thinking of Andy Palmer. Who believed first of all that I could do it. And when I wasn't really sure that I could. uh Krishna Yeshwant who was there every single step of the way, both when I was inside of Google and then when I'd spun out the company.

1:00:12 uh Josh Wolf for playing the Instrumental role. Of I mean what is Tagline at his fund is is we

1:00:21 believe before others understand and he lived that very, very deeply with me to bring Osmo to life. And then uh more recently Colin Byrne at uh Two Sigma Ventures has been following along with the story and it's just been incredible. Cheerleader and continues to bet. on the company. Look, this is my board. These are the people that bet on me and they're involved in actually growing the company.

1:00:43 And I'm super grateful that those people are one and the same. So you know, look I'm leaving people out. There's so many other people to thank, but it's just these little moments where people say yes that can make all the difference. It's a wonderful way to put it. The little moments where people say yes. Great great place to close. Alex, thanks so much for your time. Patrick, thank you so much. It's so fun to bring you here to the lab to show you what we do to kind of share this passion and I hope we can do it again soon sometime. If you enjoyed this episode, visit joincolossis.com where you'll find every episode of this podcast complete with hand-edited transcripts.

1:01:17 You can also subscribe to Colossus Review, our quarterly print, digital, and private audio publication featuring in-depth profiles of the founders, investors, and companies that we admire most. Learn more at joincolossis dot com slash subscribe.

What you see above is a preview of the first minutes. One unlock costs 10 credits and covers this episode forever: full segment and word-level timestamps on this page, plus .txt, .srt, .vtt and word-level JSON downloads, as many times as you like.