How Tempus Is Using AI To Transform Diagnostics With Eric Lefkofsky

[Music] welcome to FYI the four-year Innovation podcast this show offers an intellectual discussion on technologically enabled

disruption because investing in Innovation starts with understanding it to learn more visit ark-invest tocom Ark invest is a registered investment adviser focused on investing

and disruptive innovation this podcast is for informational purposes only and should not be relied upon as a basis for investment decisions it does not constitute either explicitly or implicitly any provision of services or

products by Ark all statements made regarding companies or Securities are strictly beliefs and points of view held by Arc or podcast guests and are not endorsements or recommendations by Arc to buy sell or hold any security clients

of Arc Investment Management May maintain positions in the Securities discussed in this podcast hello welcome to f Yi the for your Innovation podcast uh today we're

talking to Eric lovsky uh CEO of Tempest Ai and um and a great operator kind of in the startup scene for for a long time Eric welcome

uh how are you I'm good I'm good thanks for having me okay so Eric can you start out and let us know um what is Tempest AI um you know why did you start it you

went from um you know I guess most notably from a public perspective being in the group Groupon Circle to suddenly running an AI Healthcare Tech startup

this seems like a strange um move and it's a very successful and growing franchise so like what led to this moment so in my case you know i' been in

tech for a long by time I got into I got into technology in um like 1999 and had built a series of

technology companies all kind of uh interestingly applying technology to space that historically didn't have a lot of Technology just the way the way the career kind of went we started

applying technology to the printing industry and then uh transportation and Logistics like literally you know moving moving trucks around and then media

buying and these spaces were all kind of very Antiquated and and had not adopted technology and we had developed some expertise at building these kind of hybrid systems that were like half human

half automated and made these marketplaces more efficient than those companies grew and they went public or got sold or whatever and then when when Groupon came around uh it was a a kind

of a similar problem in that we were trying to apply technology a different kind of Technology basically Tipping Point technology to to to local businesses and local problems and so I

found myself you know pretty early on leveraging a lot of those same skills that I had now developed over over long more than a decade and you know trying to figure figure out how do you bring

technology to the hands of people that own a pizza parlor or a dry cleaner right very local folks have not have not historically adopted technology and I

you know you know from that point on really never thought I would get into Healthcare but about 10 years ago my wife was diagnosed breast cancer and so I found myself spending a lot of time uh

in a clinic while I was at this point the CEO Groupon and I was just amazed at how little technology was a part of her care and how

Antiquated what in theory I would have thought was a very sophisticated process was uh from a from a technology and data

and AI perspective and so I think I just had this aha moment that my entire career and all these skills I had learned at applying technology to spaces

that historically had not had technology was kind of meant for healthcare and so uh I just had a big p it you know I figured since I spent time learning

other Industries I could learn uh this might take a long time but I could learn it and that's what I did I began kind of immersing myself in this case in

oncology uh and have been immersed you know ever since how long ago was that what what is the start point of Tempest and then where is Tempest now because it's more than oncology or at least it

has Ambitions to be more than oncology like what does that look like today the company was founded in uh August this August will be 10 years so

uh whatever that is August of of uh of 15 I think and um and and pretty early on you know we we started off when I say

trying to apply technology to a space that didn't have technology was actually a much narrower Focus than that it was really trying to figure out how to make a Next Generation sequencing or these

genomic tests intelligent by connecting them to the clinical data for the patient from from they were ordered like my one of my big kind of aha moments early on was my my wife was

being sequenced when when she in the course of her treatment and these tests would come back not knowing if she was male or female not knowing you know anything about her what drugs she had taken whether she was eligible for these

trials and so it was a like a remarkably non-personalized report for what at the time was like the state-ofthe-art of personalized medicine and so I'm

thinking like this is crazy why don't they know that she's a woman why don't they know what drug she took why don't they know that they can't she couldn't she's not eligible for that trial like they just didn't know those things and so we set out to kind of connect those

dots and at the time I really having been in technology and and starting Tempest as a tech company I didn't think we would be sequencing patience I assumed we would be building that

intelligence layer that sat on top of other people's sequencing and so I went to those folks who did the sequencing at the time and said hey I I want to solve this this problem I'm pretty convinced this is the main

problem especially as these tests get more complicated and in multiple disease areas so let me let me build that that technology but the challenge is no one would give us the the molecular data so

we could contextualize or make intelligent those reports they refused they refused to give it to us they refused to give it back to the hospitals or doctors that were ordering these tests they had it locked and so that

opened up a kind of a point for us to enter the market we began sequencing and we started in oncology and obviously still are have our biggest roots in oncology but probably five or six years

into Tempest we realized that this that the platform we had built that makes these diagnostic tests intelligent and contextualizes them and makes them personal was equally applicable not just

in cancer but in cardiology neurology Immunology like pick pick any disease where there's kind of a lot of different Therapeutic Pathways and we can help uh

Physicians make data driven decision really interested in that Journey from technology through to through to bio have you along the way sort of reflected

and thought what is it that the bioindustry generally could learn from more studying how Tech is executed and maybe vice versa but but but certainly

in that direction everything I mean it it even even today like it's just it's I I get sent you know articles and people kind of talk about how like

technologically sophisticated the the you know the the the this space is because these are very technically sophisticated people but

they are technically sophisticated in a completely different discipline than the one at hand so and the challenge is like it's very

hard to be a Craftsman who you know can like make cars with your hands and then accept a world where you can 3D print a

car like it's just a very hard thing and and so um the there there is an absolute kind of lack of a skill set for computational biologists and

bioinformaticians and translational researchers and folks with phds and MDS to apply the you know Cutting Edge um AI tools and techniques especially those

that are being used today for large language model building and and the like and you have I think the first step is is appreciating that and being like okay

I I'm an expert here I'm not an expert there and so how do I find people that are and build those bridges which is what Tempest does kind of for living but you do you I always say to people like

swim in your own lane like I I don't I don't want software Engineers pretending to be phds or MDS and vice versa and one thing that's magical about Tempest we're

now once we close this acquisition we'll be approaching 4,000 people we have huge you know 300 phds and you know thousand software engineers and AI scientists and

people that are supporting that and so just a very large infrastructure of Technology talent that all has to come together to solve these problems because you you need the expertise of people with a deep science especially biology

and chemistry background but you need the expertise of people that can actually that understand how Foundation models work and how they can be uh you know modified in an agentic environment so that that point about how people swim

in their own lane is interesting because obviously you know we hear a lot about people building and trying to constantly make people more and more bilingual and able to talk each other's language and there's obviously some Merit in that

there's also Merit in focus and do you think then it's it is about a journey to make people more bilingual between the tech and the bio side or is it more about building high bandwidth ways that they can communicate and help each other

out and and yeah how does that Journey look as as we go forward and and as you grow the company yeah I mean I think I have long felt like it is going to

take um the journey I think candidly and and there's so many broken parts of the process it's you know I've said this historically like um if you asked if you

had said to me 10 years ago like design a ecosystem that would impede progress at every turn I would hand you the US Healthcare System this would this is

almost masterminded like it's genius in its in its complexity to destroy progress like at every turn right it's just I mean we could go on and on with

like you it's like it's like a movie um so I think I don't think this process fixes itself from within I just think it's it's set up to not do that I

think it's going to take you know private commercial Enterprises that have a totally different incentive to solve this problem and then

that's not good or bad it's just a reality um you know the the incentives of the current system from an academic

standpoint are often to publish and get grant money around a unique idea that you don't share too early um the incentives in terms of

biofarma are typically to generate a drug that maximizes its value during its patent life and uh and the incentives of payers are complicated at best what to

pay for and when that goes on and on so if you look at the system today it just isn't set up to adopt technology in big ways and yet I think I'm quite sure the

only solution to our problem which at this point is um maybe among the biggest problems facing the US government you you can't support A5 trillion doll Healthcare spend that's growing at seven

or eight percent a year it's just it'll engulf everything the only solution is problem is technology and AI so I think it's going to take companies like Tempest that kind of have a clean slate that have an appreciation for all the

different constituents but have a different economic model and one that is sustainable at scale to make enormous investments in deploying technology to really build products that help

everybody be more efficient including people that make drugs and people that treat patients and people that are treated by people that treat patients I'll hand the mic to brat in a second but I I just wanted to say on that

there's a there's a fantastic South Park episode I watched on the uh the plane to the recent JP Mogan conference the end of obesity which is essentially the mission the mission that they have to

try and get Cartman his as mic is is is only to just navigate the American Healthcare System highly recommend to watch it's lots of fun and and the segue I was going to give to to Brett we just did we just completed a piece of

research which is going to go out in a few weeks is that because of companies like tempus and you know what you all are working on and and lots of other great innovators in the space we've done a fair amount of calculations saying

that actually the tide is very much turned and and and really the I on the R&D going in to to biopharma is is is very very positive probably the first

time in in know many perhaps decade so I don't know if Brett was going to go there but really interesting time for the industry at this point it's almost like these problems have existed for for

for decades within the healthcare system and we've also seen Innovative companies kind of like throw themselves up against it and kind of get ground up by the

friction uh it's interesting to see here how you basically had to vertically integrate cuz nobody would let you in if you didn't how do you think about like

tempest's kind of stance with regard to the all of the skewed incentives that are in the whole ecosystem uh and how youve I mean so far successfully

navigated that I'd say through the oncology Channel and then how you think about kind of like expanding the franchise or you have you know a

consumer facing or patient facing AI app now which uh you know seems like obviously a useful thing to do but not

necessarily obviously something that becomes like a monetization engine for the company so how do you think about kind of that that strategic stance you start by saying like okay what am I trying to do here right I'm trying to

basically make Diagnostics intelligent okay so how do I make Diagnostics intelligent well the diagnostic has to know who the patient is and then recontextualize itself around that patient and what so okay what what

Diagnostics every diagnostic right blood test CAT scan MRI mamography genomic tests I want every I want every diagnostic input to be intelligent

contextualized it's the only way to help Physicians and patients get navigated to the right care because Diagnostics sit at the center of every major decision like if you're ever sick you would know this like no one makes a decision

without ordering a merried of tests and then interpreting that data and then setting you down a path so we want that to be super intelligent the problem with that is

it's very expensive to kind of try to harmonize and normalize all Health Care data in the United States or in the world in real time and then deliver

insights in real time like that's no simple task so our approach has always been to try to figure out how to build products that are sustainable as one

individual puzzle piece in that bigger puzzle right and have them build on each other so that the puzzle becomes clear at some point so we started with sequencing saying okay can we build a

you know good business in sequencing one that grows and it delivers margin and cash flow and it's a good business and then we're like okay if we can do that and if that data is connected to

clinical data so we have like Rich molecular data connected to outcome and response who would find Value in that data oh biofarma okay so how do we build

products that biofarma wants that allows them to invest in these data that's so that that can help sustain this endeavor and then you're like okay well if I have those two going then what's feeding that

business what what's feeding that is all this real-time connectivity of genomic data and clinical data that's kind of flowing in real time and if I'm building these pipes to generate that can what

other businesses can I layer on top can I help find help patients get routed to the right clinical trial can I help close care gaps can I help say something algorithmically that I can get paid for

and each of these is like a puzzle piece that ultimately you hope if you do it well is generating lots of Revenue lots of gross profit lots of of of free cash flow that can be contributed back into

the hole and probably if I look at Tempest today the two biggest hurdles that we've gotten over that it seems like others have not gotten over is one our broad connectivity the fact that we

have these connections to like 3,000 hospitals the United States right it's a huge amount of the US that is connected in this bidirectional way with Tempest you know we're sending us data we're

producing insights we're putting them back in the second is that we've been able to generate a business and if you look at our guidance for next year which I think is something like 1.23 billion

and cash flow ea. positive you can do the math like you're generating I don't know 800 million or something of gross profit it's a tremendous amount of money if you follow those Trends and if you

look at the amount of money we need to run the business it's a minority of that gross profit right the the majority of the gross profit dollars are being reinvested in product and engineering

and Science and studies and whatever so we're at a point now we're we're able to invest whatever 500 million 600 million eventually a billion dollars like back

into growth in a sustainable way and if you look at the the inflection point of of like truly great businesses and I hope tempus is one of them that that math continues so if you take a look at

Google if you just run that math forward for another another 20 years you know if you if your business is growing at 25% a year for you know and you're at a

billion dollars of revenue for you know a decade or two you're you're investing not a billion dollars into new products you're investing tens of billions or

eventually a 100 billion and so if you want to know why search is so good it's because Google probably invests 2x every year what we invest in the entire NIH

Budget on curing humans just in making search good so it's really good and people die and we need to create a model that I think allows that same kind of

flywheel of investment you can imagine there's like various stages in this journey in which you could partner with others versus kind of like build it

yourself specifically on the AI side as in I you know I imagine your ambition is not to be like I'm going to build a competitor to chat GPT but you do have a language model attached to the back end

of your um system presumably powering the Olivia app um and there is um reason to believe I think that kind

of like injecting fundamental at least like some of the clinical trial understanding into the language model in a fine-tuning manner or maybe reinforcement learning actually could enhance the set of answers that are

provided on the back end of that where like to what degree does tempest become defined by and driven on an R&D Side by kind of like their ability to kind of develop

specific AI systems versus kind of just connecting kind of like all of the rich biological information with AI backends we've always had a very large product

engineering team and a team with deep AI expertise I mean we've been it's just if you look at our investments on an annual basis and the size of our investments since we've begun raising Capital this

is where the majority of our money has gone like it's just we just spent a lot of money on Tech it's hundreds of millions of years it's no small Endeavor and it's been that way for a long time

and and the vast majority of our of our of our technology stack today is proprietary code I don't even know the total number 70 million lines of code or

some it's some enormous amount of code that said we don't look to rebuild uh core underlying technologies that are kind of perfectly good and oper

operating in in what it feels like a commoditized manner meaning multiple people have them they're quite good so like we didn't build our own optical character recognition system we used one

that was available at the time um and you know and I think to some extent I think of some of these large language models in that same vein where there are

now multiple models whether it's gemini or llama or chat 4 or whatever you know pick a model that operate really well lowc cost um High Fidelity and so we

spend more of our energy building it's a product we call Ag agent Builder which now other people have similar products but you know we have I don't even know the number probably approaching a

thousand agents we built internally that are very good at taking these models and applying them toward multimodal Healthcare data which is radically different than the data these models

were trained on so you can't just like Drop in uh one of these models on a dcom file or a or an electrocardiogram raw

wave file or a a bam file be like hey what does this mean it would be like I don't know there's three trillion A's and T's and C's and G's in random order

it means like G I don't know they would it would I you know build the largest four-letter sentence in the world I mean you'd be like I don't know so it doesn't know that like that's actually

counting um you know counting base pairs to figure out if a gene is mutated so I think we have work to do in in in applying these models but it doesn't

seem to me to make a ton of sense to try to rebuild these models they're incredibly powerful there are multiples of them they're kind of easily accessible so it seems there's no

there's no need to invest money in building a entirely bespoke Foundation model that said there are certain data types where these models don't work at

all they just weren't designed for that and so I think we will likely have to build smaller models models that are tuned to a specific um data element for healthcare but in general our we apply

models from all the big players today and they are just getting better you know by by the by the month what do you think is the Eric the biggest Advanced multimodality will give us that we

haven't got yet from a multimodal analysis of using AI over biological data like what's the big unlock and is that does that constitute some sort of maybe it's overused like a chat GPT

moment in health or is it something different there's really kind of basically a few ma massive unlocks I mean the single biggest unlock is is

essentially if I look at every if I look at every every outcome or response okay there's no reason why at some point given the size of these models the

increasing scope of these context windows and the ability to run compute at low cost I won't be able to draw associations between like lots of molecular data and outcome response so

like that's just like literally lights on versus lights off like that that which I think companies like Tempest will will make real very soon in the

next whatever one two three years whatever it is that will be I think considered the first true moment of precision medicine like the fact that I found an egfr mutation that gave that's

kind of targeted medicine it's targeted it's kind of cool right targeted but so is chemotherapy targeted to you know this particular chemo is targeted to to that disease true Precision medicine is

like I know exactly who Charles is and I like he's not going to respond to this drug like that's not going to work and so that becomes like whoa wait a minute we now need molecular data because

that's how we're going to actually determine your individual course of therapy which will not be the same as Brett or I even though the three of us may have the same exact disease

phenotypically so that's like a huge deal that's going to be and that will work I think equally well in oncology as it will in diabetes and you know in in in cardiovascular disease and

neurological disorders like it'll work a lot of places and then I think the second part of the aha moment here will be you know in addition to like

molecularly pointing you in the right direction when we have all these different data modalities being fed in your your pathology slides your your scans this those same models will likely

find all kinds of things maybe even in predictive ways so that we get we we get to you before you develop disease and that'll be even another kind of aha moment here where I'm gonna you know

you'll show up and your A1c score is 5.8 but you know you had you were riding your bike and you had a you had a little piece of uh thing went in your eye and see had a retinal scan six years ago and I'm looking at that retinal scan and

your A1C and three other things and I run some algorithm in the background I'm like Charles you're going to develop severe type 2 diabetes in 3 years and you're like whoa okay okay so I'm stop eating Donuts every morning one of the

things that originally attracted me to The Tempest research was actually the really lovely work you did a few years back with sort of phenotypic embeddings and how do you position like the Deep phenotyping concept the data driven how

do you think though about obviously we're seeing in the natural language space hallucinations are becoming one of the big issues in in in the whole thing and and it's somewhere some places in

BIO you could imagine hallucinations being either low impact or tolerable or even encourag right like in drug Discovery in the early stage as long as you can close the loop and you can get to Grand truth and validate a hallucination might be kind of another

form of scientific creativity at least you know you can tolerate false positives and then the further and the closer you get to the clinic whether it's in development and you know it becomes more and more expensive and more

impactful if you know chasing false positives and then in in actual clinical deployment uh false positives could be very problematic because even if they can be ruled out you you know someone

might end up on like a diagnostic odyy or something like that how do you think about that is this is it more about just identifying those cases where that you know there where there's tolerance to something like hallucination and then

you know addressing those or is it do you think that the models just get better and better and hallucinations become less of a problem those kind of type one false positive Arrow I think the models get better and better I think

hallucinations are already like logarithmically down you know you could we I don't even know what percent you know from from chat one you know 1.0 to

4 40 or 01 or pick a model but even beyond that we don't I don't Advocate a system without humans okay so let's just let's just make it like let's get it to

here autopilot works really well okay like we got you know 777 has Auto takeoff autopilot and Auto Land like these things have very automated systems

I want two pilots up front like I'm not advocating the pilots can like not be on the plane and I wouldn't Advocate a physician is not in the middle no matter how sophisticated these systems get over

the next decade or two I very much want a doctor being like yeah makes sense I'm going to hit go um and I one day maybe I

they you know that's unneeded maybe I don't know but nowhere in the near Horizon like I If I Were if I had a kid going to medical school today which I don't but if I did I would not have any

hesitation to say you have great job security throughout your entire career but yeah 100 years from now will we need doctors I don't know but we certainly will need them for the next you know 30

40 50 yeah I mean also the way I think about it is it it's kind of like you know if you look at kind of like the Radiology is an example where even five years ago people were saying hey with

image classification what do we need Radiologists for um and it's actually you know Radiologists were spending something like five seconds per slide you know there was a lot of stuff that

actually they would prefer to have a system that can kind of like cleave off those easy cases so they can devote kind of their training to paying more attention to the things that matter more

you know and I think that that's the like a lot of times or my experience with medical professionals it's like of course they don't have time to read my entire chart or my mother-in-law's entire chart which is incredibly complex

and like they can't actually process all that information in the time and kind of like monetization scheme that they're afforded when they meet with her so if they had a better like mechanism to

embed all that information so it could say hey this is what we think is going on then kind of like their kind of training and learning can be applied in a much more effective way to kind of

impact her health outcome positively I think the other thing that's kind of interesting if you look at that there's a deep mind paper from a few years ago on the automated mamography reading which was very interesting because

essentially you can dial in the sensitivity specificity I sensivity as as folks appreciate being essentially the number of um false negatives or a

measure of that and and specificity being the lack of false positives um and you could because you can dial it in to that point about you can the pairing of the human and the and the Machine is really nice because you know very

specifically because you can run enough tests on the machine what setting you're at so you you if you had like a paired radiologist you can have the the AI radiologist if you like being set very

very high detection rate even if they send over some some false positives and then the human can deal with it which is you know overall you end up with a better system actually which and it's way cheaper and faster so that that I

think that is just one early little signal and I think Tempest is is really industrializing some of this that you know I think leads to dramatically better and cheaper Healthcare which is very exciting it's as you say it's the

augmented human 100% when you when you you know it's like when you start with A system that for the sake of argument is somewhere between 30 and 50% inefficient

some pick some number okay these improvements that you get like you won't even kind of feel them at the macro level I mean here's a great example like

uh uh so I you know I'm 55 so I grew up in a world where there was no you know no no smartphones and no GPS so you basically got lost every everywhere you went you got lost literally if you were

driving you were lost you were pulling over to a gas station asking someone like where to go so okay lost on the road is gone now like no one's been lost

in a decade like no one's ever gotten lost unless they don't have a signal and like do we use less gas or there are the are the roads like empty like in other

words that whole inefficiency is gone like all the miles we drove in the wrong direction are gone and yet it's really crowded and gas prices are really high

and blah blah blah and so I think it's the same thing here like you know there's going to be massive efficiencies but I wouldn't structurally when we take these leaps to like what

are all the doctors going to do and you know what do we do with a million nurses like they will be like there will be a shortage of nurses a decade from now I

like would guarantee it so and but but hopefully what does happen with technology is you get significant improvements every year so that five

trillion either stays at 5 trillion but can handle another 50 60 million people and can handle them living a lot longer so that it as a percentage of GDP it comes way down or you get real savings

you know you get that 5 trillion to four trillion and all of a sudden you've got some dislocation no question but you've got a system that's way healthier I wonder if

you could unpack for me a one there's like this throughout my whole career of looking at like technology and Diagnostics um kind of there people have asserted that there's no money in

Diagnostics basically because they become interchangeable uh and maybe that's that narrative has diminished some specifically in in oncology where

kind of there's the ability to kind of like demonstrate differentiated testing but it certainly is true in like you know standard blood testing and stuff um

do you think one like one how do you think about that from kind of like building sustainable businesses within Tempest and then also is is there something different about oncology

versus other kind of parts of healthcare that just lend it s to this sort of business model working and then it'll be just more challenging in other areas I think it'll be the either the same

challenging or probably less challenging in other areas because we'll we we'll the road map and oncology will likely kind of set the stage for other areas nonology had to be first um so there's

always change is always hard but look I mean and you know the guys that are you know know this pretty well but I mean before Tesla if you just said like you just said there's no money in making s

what a what a horrible you know like the biggest company in the world's worth 25 billion and they you know they've been around for 100 years and you know whatever so I think this notion that by

the way you'd have kind of said the same thing about movies you'd have been like who wants to like Blockbusters is not that great I don't want the world and here's Netflix worth I don't know I have

no idea 20 times what blockbuster was ever Worth or some probably some number like that so te when you when a technology Paradigm Shift can can kind of open up entirely new businesses and I

would suspect that if Tempest is ultimately a successful as a Tesla its journey is going to be very similar right which is like okay we made a

better car like we made better Diagnostics and so that produced some growth but long term all these other things you layer on

top create really the sustainable Advantage like you know in the case of Tesla you know things they're doing in terms of autonomous driving that could

one day open up you know taxi business or uh their their recharging stations or um you know robotics all these things that get layered on top of this core

like I'm going to build a better electronic you know mouse trap than the current gasoline powered mouse traps that exist and so I think it's opened up the world's idea to like oh wait a

minute you can have a trillion dollar car company and I suspect the same thing will happen in Diagnostics I would bet a lot of money that you you will see Di

diagnostic technology enabled diagnostic companies that could be very very big because again they sit at the heart of all major decisions so why why would they kind of give all their why would

they just seed all of that insight to somebody else why wouldn't they become the people that deliver that insight and the fact that some of the diagnostic players historically have been rooted in

very lowk in high volume tests is the same as the fact that we had car companies that you know didn't didn't become Tesla like you just have it you know or we had you know video store rental companies that didn't become

Netflix like sometimes you need a a new entrance to kind of change the game there's also a lot more complexity right diagnostic tests in a good way because they're just doing more are becoming more complex harder to do therefore also

harder to commoditize and we've seen that in some of the mrd and you know some of the certainly the early detection work as well we've got these quite layered tests looking at several different epigenetic signals Etc so when

we decided to acquire amre you know people would say to me like again they bring up this first of all by the way when I started Tempest for the record the first thing people said to me is don't open up a lap uh NGS is being

commoditized this is a this is a commodity space this is a race to zero and all that's happened in the past 10 years is the exact opposite of that um and then when we decided to buy Ambry

people were like hereditary testing is uh that's you know it's like isn't that a commodity like that feels like it's kind of best days are behind it and I'm like well how' you how do you how do you how' you conclude that and sometimes

public investors are like well I bought in V stock and I bought married didn't work out well that doesn't it's not a space like that doesn't Define a space right now in this country I don't know

some micro percentage of people get you know kind of whole genome or whole exom or large panel hereditary testing and I would suspect if you and I could fast forward 20 years from now it's like the

whole country maybe 50 60% of the whole world so I don't know I mean how big could somebody

like an Ambry get like you know giant you know literally I mean so and I and I don't know why that wouldn't happen we're going to ultimately understand from that molecular data because of

these large language models and large image models like everything that's going to happen to you oh like you know you're you're you're kind of at really high risk of of of developing apib at

some point down the road or you're really at high risk of uh of of early onset um dementia whatever it is and you're going to want to know that and you're going to want to take action and

so and when you get disease that same data can be used like over and over again that's the best part of molecular data I can just keep ring reusing it so for example what drug are you taking

like do you metabolize that drug well like maybe you need double the dose maybe you need half the dose like all these insights will flow from molecular data so I do one of the reasons we chose

to like start with molecular data is we thought there's no way to make Diagnostics intelligent if you can't generate molecular data that would be like Tesla trying to build an electronic

vehicle with no battery and no ability to like get a battery like hoping someone else would give them a battery and I just think that would be tricky place to start over the course of like

Tempest the underlying technology like the technology that's able to power what you're doing has improved like we think you know in in our work um AI systems

cost per performance are improving 75% per or falling by fourfold per year as in same dollar out you get fourfold the

power uh um year by year so much faster than Mor law um so you must palpably feel that and how do you how do you like

business plan around that kind of underlying um Power explos expion like how do you think about like even both strategically and then how you're

building products like how do you kind of like build on that Foundation that's moving so quickly yeah we probably fill it no question I mean a bunch of the agents we built that run models today

like would have been cost prohibitive when we started we when we started we started building agents pretty early like we we were very early in this game we've been at it for you know a few

years so we we we we were early and some of the models we be and running some um we would like we'd run a query and it would cost us like $2.7 million to run

like a very tiny query we're like okay this is bad like even even we who spent a lot of money on cloud and compute can't absorb this it's like inum and now those same models that cost us 2 million

bucks to run now cost you know I don't know 2,000 bucks to run so it's literally it's like that extreme or 10,000 some some some tiny fraction of what they were so the

there's no question we're benefiting from that but remember to do what or not remember but to do what Tempest wants to do at scale our main business eventually

is to contextualize all Diagnostics and generate real-time insights so everyone's on the right theraputic path so to do that I essentially have to kind of run compute across all healthc care

in real time every time there's a diagnostic Insight so that's a layer like we're not first of all we're not ready for that Tempest isn't ready ready for that the US Healthcare System is not ready but I think it will be at some

point but you're talking about a scale that's really unimaginable it's like a different it's a different level of scale and so you know we spend a lot of time we're try we're TR Cloud today but

we spend a lot of time thinking through with our Cloud Partners uh like how's that going to work you know because it's that is going to come and people like uh Google and

and Microsoft and others will have to figure out how to how to how to make that possible we touched on Olivia for a second but it'd be interesting as that was a recent launch just to talk a bit

more about that and I gather even the naming of that was yet another um personal story of how how cancer at least come come close to you uh and family so any any insights or just

thoughts about that and the road map around how you see that developing and helping patients would be super interesting it was a personal story a friend of my daughters who who passed away and um you know we just it was a

great opportunity to kind of recognize her but the challenge she that her parents had and the challenge that we all have if you've ever been in this

situation is like managing uh a complex care is miserable like if you have multiple doctors that

you are trying to get opinions from and weigh in and what should I do and da da da it is like no simple task especially when you have like scans and slides and

tests and all that so you know Tempest if you think about it like we we've spent a couple of billion dollars building these technologies that make sense of multimodal messy Healthcare

data and we thought look let's put that into an app and give patients access to those same tools so they can basically kind of click a button and not just get

like some of their structured data from you know my charts or apple or whatever but really get all their data like literally all of it so that if you want to move from one doctor to another or

get a second opinion or just have an archive of it like it's there and you have it stored in this little personal Locker that's in your pocket you can send that data anywhere you want because

it's your data and you can ask at questions and in theory as these models get better and better the questions you'll be able to ask will just keep getting smarter and smarter and we want patients armed we've been you know I've

been saying this for a long time I started telling uh Hospital systems and doctors that you know I've been doing this for a while and it felt I could it felt to me

like there was this tsunami coming only unlike other tsunamis this one was like 5,000 feet high traveling like at the speed of an asteroid like this was

this was like really not a fun tsunami and they needed to get ahead of it and I think that tsunami I couldn't really articulate what it was 9 or 10 years ago

but now I really feel like I can't that tsunami is that the average patient who today walks into their doctor in a completely uninformed way looking to

just put all their trust in somebody else with no understanding of what's going on in a really short period of time will walk in as if they just met

with a hundred of the smartest people on the planet that spent weeks understanding their condition and have given them the 10 questions they need to ask and make sure they get right

so like you're going to go from like uninformed patients to like super informed and uh and then you're going to say something and they're going to be

like uh did that did that like what do you think it's going be like oh no no no no no no like no don't do that bad so you know like that's G to happen so at the end of the day you're G to have

really sophisticated tools that patients can use and we want we we we view we view information flow to everyone as nothing but a good thing

like period end of story I want payers paying for the right drug not the wrong drug I want drug companies being as efficient as they can be never having a failed phase two or phase three want patients to know everything about their

care I want good doctors to be great doctors and great doctors to be super doctors like you just want information flowing and the people that the only people that should suffer in that world are the people that are like nah forget

this technology it's garbage garbage it's it hallucinates I'm not going you know what it hallucinates I'm done with it and if you what's crazy that that like there was a great stat I can't find

it I wish I could they someone read an article a while ago they asked Fortune like 50 or Fortune 100 CEOs back in the year 2000 if they thought the internet

was a fad and then they looked at the stock this this the stock market cap of those companies stratified by the CEOs that said it's basically a fad versus those that said this is like going to

change everything and it's like it's perfect it's like a perfect correlation like those people that saw it coming got ahead of it those people that didn't didn't and we want everyone in healthcare to get ahead of it Olivia is

available Nationwide now right like I can go and and sign up yeah it's going to take us we we were flooded with people that wanted to come in we made a decision to give everybody a free for it's like I think it costs like $12 a

month or something so we wanted to give people a 14-day free code because we didn't want everyone just you know not trying it so I we've been we've been metering out how fast we

get those codes to people and how fast that all works but everyone will come off the weit list in a a day or so and then as people sign up they'll get they'll get the code it is striking from

the you know as a patient and then with people in my family that are patients the the difference just from um asking

Google what's wrong with me which you know leads me to WebMD which leads to me to you might have cancer to like actually plugging in you know diagnostic information readouts to chat GPT and

being like what is going on here is such a profound likely more useful tool and so then having that but the the friction there is like I can only plug in

whatever I have can copy out of the EHR into the context window chat GPT and I didn't have to like kind of like randomly insert other information that like I think is relevant to it so it

seems like having that like the full patient experience packaged and then fed into an intelligent language model really could change the consumer experience of healthcare and do you

think that like in some ways like the payer system seems so broken kind of like the the healthcare system is broken in part by the fact that the consumer is in some ways divorced from all the

decisions that get made do do you think that what Tempest is doing um catalyzes and or requires their to be a more consumer-driven kind of like Health

decision framework that happens here but I couldn't agree more with you that um these exper by remember like for us Olivia is like chat 1.0 right this is like very early so it'll it'll keep

getting better but what I what I say that what I really kind of believe is if you if you look at Google for a moment Google has really two kinds of searches they have regular search and Incognito

and in a world of healthcare they actually need a third kind of search which is you know everything about me like literally you have my entire medical record and when I ask you a

question Google I don't want a generic answer I want you to literally I want you to be like I would you know I want you to know everything about me and then say things with that in mind so that

you're not recommending if you know I'm taking anti-depressant don't don't send me like to something else like just obviously my meds are wrong you know you know whatever it would be like I really

want you to render result results that are high not just kind of semi- personalized but entirely personalized like and I think the challenge is some of the big search engines and and I

think also you know could be this way with some of the big uh llm companies and general b companies is that they they may have a hard time with that right they're they're kind of like well privacy and I don't want all that data

it's a lot of Phi and what do I do so I don't know how this is going to shake down it may be that companies like Tempest end up kind of building more impactful products because we don't we want to have that relationship like

we're not afraid of it to the contrary we think it's invaluable but I do think eventually you know everyone will have to adapt and allow you to talk to your

AI agent in an incredibly um you know identified way great well thank you that was fantastic to hear some of the vision uh and

bringing it about to patients all the time and how technolog is helping and will help what what do you think is the biggest obviously as as they say making predictions espec is hard especially about the future but as you look like 10

years out you you you do a lot of envisioning the future what do you think is the biggest single thing that will have changed in an individual health Journey that is not the case now and

will be you know hopefully a surprise to the upside something we're that the health system is just not factoring in right now the markets are not pricing in people are not thinking about the

biggest single thing the biggest single thing will be that that the the error in the current system the the the US healthcare system has some amazing attributes including Freedom right this

freedom to innovate and blah but but the error the system produces is is unbelievable it's like I don't know one of every four decisions is entirely the

wrong decision and sets a patient and a system down a really bad Road and I think in roughly 10 years I could see

that being mostly gone where errors are very infrequent certainly by 20 years so some somewhere between now and 20 years but and in a world where there's no

errors and people never get the wrong drug in the wrong dose at the wrong time then you're going to have real improvements to longevity you're going to have probably a you know you could

easily see a three to five year extension of Life something something of that magnitude you're not going to see 10 or 20 but you're going to see you know you're going to see three to five

you you feel three to five you know one felt like covid was one year so you'll you'll you'll when when we're living five years longer you'll you'll feel it in in every part of the economy

it's like no small shift and so I think that's going to be the biggest change as people will live longer and healthier lives because when they have problems they're going to be on the right path and then I think as I mentioned the next

big one that comes after that is is is being predictive to avoid problems before they occur it's also coming and by the way you worked on that it's like that space I think had some again like everything

else remember we started siging patients 20 years ago it's only now that maybe we're getting somewhere and it's probably the same thing in terms of a lot of these early detection and predictive test like the early entrance

typically aren't the solutions change over time and you got to find the right solution but I think you know 10 or 20 years from now we will be very good at predicting lots of stuff and avoiding it I remember you remember we spoke about

probably early days of freenom early days of Tempest probably was around 2015 2016 we got connected so very excited to see all the progress you've made and you know couldn't be more excited about the future you're bringing for patients so

thank you Eric thanks for having me Eric liosi thank you and uh I guess people should try out Olivia uh if uh if you want like better answers about your

health and uh um really Tempest is I mean at least all the way going through you know reading the S1 and and talking to you uh leading into the IPO it really

isn't a company that is uh approaching things from an interesting in first principal perspective of how to I think remake human health ultimately and the definition of convergence in health

which we like thank you thank you thank you for listening to this episode of FYI the for your Innovation podcast if you enjoyed this episode please hit that like button and make sure to leave a

comment down below letting us know what you thought and if you haven't already please make sure to subscribe to the show and follow us at ARK invest on your favorite social platforms so you stay up

to date whenever we drop new episode Ark believes that the information presented is accurate and was obtained from sources that Arc believes to be reliable however Arc does not guarantee the accuracy or

completeness of any information and such information may be subject to change without notice from Arc historical results are not indications of future results certain of the statements contained in this podcast may be statements of future expectations and

other forward-looking statements that are based on arc's current views and assumptions and involve Known Unknown risks and uncertainties that could cause actual results performance or events to differ materially from those expressed

or implied in such statements [Music]