Numeracy, Probability, and Business – Embracing Randomness for Better Decisions

I went from Midtown New York to Dumbo, which is this cool

warehousey, lofty space in Brooklyn. And I was like, what's

this guy in the corner do? And she's like, oh, he's a data scientist. And

I was like, what the hell is that? That sounds made up,

you know? And she's like, oh, he tells us about the correlations

of X and Y and the charts. I was like, huh?

You're like, I know that. I know what the correlation, I mean. Right.

You're speaking my language. And I met him, I chatted with him, and he's like,

this is what I do. I was like, this is so cool.

Hello and welcome back to Data Driven, the podcast where we explore the emerging field

of data science, artificial intelligence, and all of it is possible

through the hard work of data engineers, Like, my favorite is

data engineer Andy Leonard. How's it going, Andy? Hey, Frank, it's

going well. How are you, sir? I'm doing all right. I have been a

busy man. I no longer work at Red Hat. I work now

at a company called ClearML, but we will be rebranding to a new name.

And as soon as I have authorization to share that new name wider, I will.

Today, we've had a really fun conversation in the virtual green room, which

sadly we will not share. But we have

Ilan Mann, who has an interesting career.

Currently, he's the founder and CEO of Paradox Machines,

and it's a full-stack data and AI company. And they deploy a

managed data platform into your cloud environment and embed a

senior team to run it, aggregating the data from the systems you already use

into a single structured layer that your leadership can

actually query. I think that's cool. Product and service kind of bundled. We'll talk about

that. Mm-hmm. But the most exciting thing, 2 exciting things. One,

he's in Brooklyn, and 2, he started his professional

career as an actuary, which is probably what people

called data scientists before the term data scientist was

coined. Welcome to the show, Ilan. Thanks, Frank. Thanks, Andy. Thanks for

having me. And I'm very excited for this conversation,

mostly because you 2 might be the only 2 that think being an

actuary is the most exciting. Adding a little bit of commentary on

you, or a commentary on me and how uninteresting

my life is. But let's go with it. Let's go with it. Yes, being an

actuary is definitely a little bit tedious. I mean, it's interesting

because it really, I think the rise of big data, and you started your

career just as big data started taking off as a commercial opportunity,

but the rise of big data, I also noticed that you immediately switched to data

Squarespace. according to your LinkedIn profile. So, so tell me

about that, because when I made the switch from Windows development into

data science, my wife, who has a background in mathematics from Carnegie

Mellon and things like that, I told her all the courses I was taking and

what I wanted to do. And she's like, she turned to me and said, so

you want to be an actuary? Yeah, she, she

probably wouldn't think it was cool like Andy and I do. So yeah, but I

love her anyway. Yeah, I'm happy to share. So I started my my

schooling. So I went to the University of Toronto in Canada. I

grew up in Toronto and I started as a computer engineer

because that's just what you did. Wait, if you're good at math and physics and

kind of science and you're a nerd like me. And I was tinkering with

So it was super fun. Turns out that I didn't make a very good

computer engineer. In particular, I wasn't good at at

programming, at understanding how to, you know, there would be the exercises

of, okay, program a chessboard in Java. And I just didn't really

understand classes and static, public, all

this stuff. But I was always good at math, right? I kind of aced my

calculus, my linear algebra. That just kind of made sense to me. So I went

to the counselor at school a year or two into my degree

thinking, okay, I don't think this computer engineering thing is for me. Everybody's blowing

by me. But, you know, I'm pretty good at math. And

they said, well, have you ever considered being an actuary? And I said,

I haven't because I've never heard of such a thing. And so they

walked me through it. And all I took from this conversation

was, with them was, you do a bunch of tests

and the more tests you do, they're math tests. This is how it was

pitched to me. You do a bunch of math tests after you graduate, mind you.

Yeah. And the more you do, the more senior you get and the more money

you make. And I said, well, this is perfect, right? I'm in testing mode.

I'm good at math. I like the meritocracy of you

just like, you know, the harder you study, the more money you make. I was

like, it just made all the sense in the world. Of course, that's not how

the world works. But when you're 18, 19, 20 years old and all

you know is an entire lifetime effectively of

schooling and this notion of you do tests and you move up,

I thought, great, what a continuation of what I know already.

And so I did it. You know, I toyed with the idea of being a

math professor, but then I learned more about what that was,

and you do even more schooling and you make even less money. And I was

like, okay, forget that. Let me just be an actuary.

And so I did that. I did that for a number of years, bounced around

a couple different companies, mostly like large organizations, reinsurance companies.

And it was like reasonably interesting. I did the tests, I

passed the tests. I kind of got to this quote unquote

peak as a young person, very young person, very

inexperienced young person blowing through all the tests.

And I moved to New York. I started working for Ernst

Young, also as an actuary. And let me tell

you, like the math was just not as, not nearly as interesting

as one expects. And it

wasn't as, I think, technically interesting for, again, a math

nerd like me. And I think that was just the expectations of

what I had versus the reality were just misaligned. And by the way,

that's like what everybody goes through when they go through this journey. They

realize like, oh, the real, like, business math is not

academic math. And this is, you know, I was kind of sold like a bill

of goods. And at the time, so this is like end of the

with their article about data science is the sexiest job of the

21st century. And I was like, well, I want to, I want a sexy job.

And I met my now wife, who at the time was a product manager at

a startup. What the hell is a startup? Well, they had

exposed brick, they played ping pong,

they had a beer keg. Everybody there was in their 20s and I'm

coming from like a large consultancy. Yeah. Ernst Young

is very buttoned down and you're going to like presumably some Silicon

Alley firm. Yeah, yeah, yeah, exactly. I went from like

Midtown New York to Dumbo, which is like this cool warehousey,

lofty space in Brooklyn. And I was like, what's this guy in the corner

do? And she's like, oh, he's a data scientist. And I was like,

the hell is that? That sounds made up, you know?

And she's like, oh, he tells us about the correlations of X and

Y and the charts. I was like, huh? You're like, I

didn't know that. I know what the correlation— I mean,

you're speaking my language. And I met him, I chatted with him, and he's like,

this is what I do. I was like, this is so cool. Like, you're doing

statistics in code and you're applying it and

you're solving these problems that are coming at you every day. And you're

iterating, you know, this is before I even know the word iterating, right? But you're

iterating on these problems together and people are making decisions

based on your analysis. Like, this is crazy interesting.

And anyway, that led me down the journey of data science and

tech and startups. And it's a whole different

language and vocabulary I had to learn, but

fundamentally they're using data. make decisions

and doing an analysis and in between. And you are

the person who understands, who can map data and patterns

in the data to business decisions and outcomes

that the business wants. And I was like, that is such an interesting job.

And you're using code, not Excel. Excel is still obviously

used, but you're using Python or R or whatever these like

SQL, these different coding languages at the time that were arcane to

me to do this. And by the way, there's like

this huge path, a lot of excitement, so much

energy. Everybody's in their 20s, you know, at least for

in my circle at the time that I was like, this just makes

so much sense for me to do. And, you know, I did the kind of

fake it till you make it. I didn't know what the hell I was doing,

but I can like, you know, try to

network and try to learn what I needed to learn. And anyways,

to your point, Frank, like I then ended up at

Squarespace, which was like an up-and-coming website

design company, pretty beloved, I would say, in the New York area.

And yeah, the rest is boring history, I guess.

So I'm gonna jump in. I love your passion.

Right away I recognized that and I love it. I have that

same passion. I love using data to help

people and, you know, helping people make decisions

qualifies. And I absolutely love it that

you were sold a bill of goods, not unintentionally. It was— Yeah,

exactly. But, and the people who did it had, you know, had

your good in mind. They were actually looking out for you.

And they rescued you from, at that time, which sounds

like you're now more in, maybe in more in code, or at least you were

when you jumped over to data science, but you needed, I think,

that motivator to have an application

for your math that was code included

in that application. Is that right? How do you feel about that? Yeah, yeah. I

mean, code, you know, enables you, you know, to move fast,

to actually make real, to manifest these

symbols, right? These like integrals and,

you know, linear, like matrices, when you multiply them, you get a thing. And

I've always been so fascinated with the fact that I can,

you can kind of think through in formal

logic and like deductive logic or some sort of a mathematical framework.

You could write it down. It can make all the sense in the world, but

it's very abstract. Yeah. And it's very just like academic, an

ivory tower. And then you open up a notebook, like a Python

notebook or SQL or something, and you literally

translate into code the things that

you've written down on a, you know, with pen and paper that in theory makes

sense, right? Like in theory, when you integrate a distribution,

the sum of the probabilities gets you to 1, and that's 100%.

And that's what they told us in Statistics 101 or whatever you take. And

you just are like, yeah, that makes sense. That's what an integral does. What the

hell is an integral, right? Like, it's this, like, no, it's

just like so far out there compared to, or a derivative

compared to like arithmetic. Now, of course, it is arithmetic at

bottom, but you're so far removed. But if I write a for

loop or some sort of a loop in code, which is

effectively what an integral is, and I do a summation of

all the indices of the terms, and I define a

probability distribution in code. Well, it turns out when I loop

over from like 0 to 1 or negative infinity to

infinity at the limit, I get, and then I

take the sum of my probabilities, I get 1, I get 100%. And

I just, that was, maybe I'm very naive or something, but

it was so cool to me that I can see with my

eyes and I can like break it down, the thing that I wrote

on pen and paper that these mathematicians told me would be true.

And of course, sometimes it doesn't correspond, sometimes it doesn't work and you get rounding

errors or you get, you know, something that breaks down, but then you could use

the code to debug your thinking. And it

always was, you know, I don't know, I had this like boyhood kind of

like, I don't know, enthusiasm for this

because I could see it. You know, I could, I could see the code doing

the thing that the math promised would be true. And I don't

know, part of me thinks that people take for granted that it just works,

that they're like, yeah, yeah, you just, here it is. It just works. And I

say, no, it doesn't just work. You have to like do it. You know what

I mean? Right. There's a certain magic in statistics.

There's a certain magic of that. And for me, it's about finding the signal amongst

the noise. Mm-hmm. Right? The data is telling you things.

You just have to know where to find it and where to look at it,

right? And it's fascinating. It's fascinating. And I also think too, like something you

said was very funny was one, you know, the

whole fake it till you make it thing. In the early days of data science,

there was not really a well-worn path for this, right? Everybody was just kind of

making it up as they go along. I think there's a lot of that too

in the AI space too. You hear about AI and agentic patterns and things like

that. I'm like, no one's really got this all figured out, right? And anyone that

tells you they have it all figured out is selling, selling you their solution. They're

selling you a bundle of goods, right? The other thing too is I didn't realize

that I had my Tim Hortons cup today because there's a Tim— the first Tim

Hortons opened in Maryland near my house. So, oh wow. I

stopped by there and so I figured— I didn't realize you were Canadian,

but I'm sitting here sipping this. I'm like, wait a minute.

So we do a lot of, we do a lot of like distractions like that.

Yeah. Yeah. Tim Hortons is an institution in Canada.

I've been to a couple of Tim Hortons here in the US. They're not the

same. Sorry to say. Yeah. I mean, it's the same

branding and it's ostensibly the same if you didn't know any different,

but somehow the coffee hits different. Uh, the Timbits,

which you would call Munchkins, I think, in like Dunkin'

Donuts. Yeah. Yeah. Yeah. Timbits, uh, hit different. Uh, the sandwiches a bit.

Still good, but I think, I don't know, it's missing that je ne sais

quoi. But Frank, I think you kind of stole the

words out of my mouth. Nobody really knows

what's going on. And in particular, those folks

who claim that they do are the ones that are kind of selling you

something. And that's been my MO the whole time is,

you know, it's okay to have an idea, to guess, To

take a stab at something. I mean, you got, you can't just like sit on

your hands. You gotta make a decision. You gotta put your money somewhere. You gotta

spend, you know, allocate resources, but don't delude

yourself into thinking that, or that the person on the other side of the call

knows. They might be confident, right? They might know

something. But I always am quite

skeptical. And maybe this is my statistics brain, which always thinks in

distributions and thinks in, you know, spectrums. I'm always skeptical of

those folks who have just high degree of certainty, right? I'm always

saying, what are your— it's fine to make a prediction, but like, talk to me

about your error bars. Are they wide? Are they narrow? How do you

think about that? And if people don't, and sometimes

they don't know how to talk about error bars, which is fine. Like,

I don't know, my error bars are 20. Like, what the hell does that mean?

Doesn't mean anything. But so then I say, okay, what would you do if

you were wrong? Yeah. What are other decisions you make? How do you

iterate? And if people don't have good answers to that, that

tells me that they're only, they're thinking kind of in a single direction

and they don't have, they couldn't fathom it going wrong.

You know what I mean? Like, what if AI blows up

tomorrow? And if they're like, well, I don't know, then you say, okay, then you

must be pretty confident that it won't blow up because I guarantee

you. If you had thought about it and you thought, well, what if it goes

wrong? You would've come up with some contingencies or some

remediation measures. And if you didn't, I don't think it's

'cause you're not a thoughtful person. That could be true. But I think

because you're just like super confident and I would question

the amount of confidence you have in whatever it is that you're doing.

That makes sense? Well, that's hard to say. That's 100%. Statistics, when you study

statistics, It changes the way you think, and you can

never hear things quite the same way again. Right? You know, one of the

talks I give is about facial recognition and the ethics behind that. And

you know, one of the—I'm not going to name them because I'm tired of dealing

with lawyers. Long story there. But but

they said, well, you know, they've been involved in a lot of false

arrests and things like that, and lawsuits right now. And they're like, you know, their

material says that. Or at least what the police said is that we were told

this was 100% accurate. And again, if I hear

100% accurate, immediately the hairs on the back of my neck stand up and I'm

like, oh, yeah, 100%, you say?

Yeah. Is anything 100%? I mean, what is like— The

only thing that's 100% is it's probably not 100%. That's the

only thing I can say. Exactly. Exactly. No. And, you

know, so I tell my team often as well, like, Just because you're,

you run the simulations and there's a distribution of results and maybe

you're 80% confident in, you know, the going down,

going on the left and you're 20% confident going on the right. Like,

it's good to think through that. Ultimately, you gotta make a decision. So it's

fine to go with your gut. It's fine to go in the

direction that you're not 100%, you're gonna go in the direction that you're not 100%

confident in. That's not the issue as much as it's, uh, you know,

how do you mitigate your risk, right? How do you make sure that you're

going, that you've taken all the precautions that are reasonable,

right? Sure. And this is where constraints come in, whether it's a time

constraint, a budget constraint, an information

asymmetry constraint. Like we all operate in a

world of imperfect information. That's fine. You just have to embrace that.

And just, you know, despite that, make

the best decision that you can, and, you know, kind of live to fight

another day. I mean, it's science, right? The

science in data science is about iteration, right? Sure. Hypothesis,

test, correct, repeat. I

think people forget that, right? And I think it's also interesting, you know, I'd

made a quantum joke earlier. It's interesting how people got used

to computing being very deterministic, right? 2 2 is

always 4, right? But AI has gotten us used to, or at

least you would hope, to probabilistic computing,

which is realistically all quantum computing is essentially about, right? You're

never guaranteed the same answer twice. Yeah. You know, I don't

know. Yeah, yeah. I'm really interested in the

fact that, um, so, you know, there's like literacy and there's, uh,

what is it? Numeracy? Yes, numeracy. Yeah. Numeracy. And

I'm really interested in society. And this is maybe, Frank, the point you're kind of

getting at, society getting more numerate. Right? Like, we talk about being literate,

but I don't think that we often talk about being numerate. And

statistics plays a big role in that, right? Because we don't really think

about calculus and linear algebra, but we think about numbers. How

much money you have, how many people live in a country, how many people

vote, what is the likelihood of a president getting elected, and the

inflation rates and stuff like— these are numbers that float around

our society all the time. And I think all around you, and once it's all

around you, one of the things that warps your brain when you study statistics is

like you see it everywhere. Weather forecast, weather forecast is probably the

most obvious, right? Like, what does it mean there's a 30% chance of rain today?

Like, so let me, let me tell you something that I find— I'm

like, I, I feel like I'm talking a little bit too much here, but no,

let me— You're the guest, man. Yeah, Andy and I

bang on all the time, so So something that I'm, that I'm

kind of curious to get your take on. So, and it's okay,

COVID happened, right? We all remember that. And during COVID

there were a lot of conversations around vaccines or

different treatments. Putting aside your thoughts on them, let's just

talk about the numbers. And I don't, I don't have the exact

numbers off the top of my head, but I remember a lot of

conversations Where it was of the frame,

hey, you should get vaccinated because it has— and I'm going to make up some

numbers— it has a 90% effective kind of effective rate.

And they're great. Okay, fine. It's 90% effective. That's what public policy

is, what the experts are saying. Fine. And, you know, people have been

vaccinated for different things for, you know, decades now. So it

wasn't like a new concept that vaccines have some sort of a rate of

effectiveness. whatever that rate happens to be. So let's say

it's 90%. Great. It's a high number, high enough that you think, okay, it's worth

getting vaccinated. So then you do. And then you'd hear a lot of

folks say, I was vaccinated and I still got the

COVID And there would be a disconnect in their heads. They'd be like,

but how could I possibly get it if it's 90% effective?

And there was this disconnect because And this is

to the point of society getting more numerate. The

disconnect in my mind, I mean, there might be a lot of disconnects, but like

from a pure numbers perspective, the disconnect is that

when somebody says from a public policy perspective, it's

90% effective, whatever the percent happens to be, what they mean is

something like if 100 people got

vaccinated, 90% would not get the

COVID strain, right? And that's because they

ran experiments and all the rest of it. That means 10% will get it.

Something like this, you know, I think— That's what I would

expect. That's a safe way

of interpreting it, right? Even though I think it's a little bit different when

you get into the nuts and bolts, but roughly. But you as an individual,

you either get it or you don't get it, right? You don't get 90%.

of the thing. You get 100% or 0%. So

the people, but there's a disconnect in how we think about

probabilities. And this is what I was saying before, where you have to make a

decision, even if you're 80% confident that it's left

and 20% confident that you go right, you still have to choose one or the

other. Right. So that was a very interesting, that was the first time it

dawned on me that, oh, because I was like, well, of course you might

get it. You're 90% safe. But like, you either

will or you won't. Like, you must understand that just because

you get it doesn't mean that the claim that it is 90%

effective is invalid. It doesn't have anything to do with that

claim because we're talking about broad

probabilities in a population level, like from a sample.

And those are frequentist probabilities. Those are, I flip a

coin 100 times to see if it's tails or heads

to determine if it's a fair coin or not. But any given flip of a

coin will either, will with 100% certainty

either be heads or it will either be tails

regardless of if it's a fair coin or not. Like regardless of if it's

weighted or all these like experiments. And this was the exact same

thing. And it was, and I'll get off my soapbox in a second.

And it was, it was like an interesting— social

experiment in my mind, how many people, and

obviously like people have a lot of feelings about vaccines and COVID and the

present. Like there's all kinds of emotions. There's a lot of other things around that,

right? There's a lot of other things around that. Exactly. That maybe couched

people's perceptions, but, or

maybe actually like because of that, and this is,

this is where data and math, I think, plays an out—

like should be playing an outsized role. Is

despite your emotions and your feelings, the

math doesn't care about it. And I think that is one of the lessons

is the numbers are what they are and

interpreting them, you got to do it with like an unbiased hat

on or whatever. And, and I think society was just not

ready for that and probably still isn't. But to your point, Frank, about like

AI is making things more probabilistic, I hope so. Yeah. Because the world is

probabilistic, believe it or not. And— Right. And it's good

for the consumers, you know, because AI is not just this

B2B, like enterprises use AI. Consumers are using AI

and they're using it in their chats and in their, you know, different apps.

And I hope that they understand that they're engaging

with a technology that is not guaranteed to give them the thing that they

expect whenever they hit refresh. or they rerun it. And that

says something really important. There's a really important point there.

And even if they don't understand transformer architecture and all the rest of it and

neural nets, it's still, I think my hope

is that the next, our generation's lost, right? But I hope like my

kids' generation that they grow up just appreciating

that things are random and we have to deal with that fact.

And so, Yeah, I always come back to the COVID example

of like, at least for me, this was like the first example of like

people just could not grapple with the fact that what the expert, what the

government was saying in terms of effectiveness was different than their

personal experiences. But that's okay. I mean, that's true,

that's gonna happen. But what do you do with that? And— Well, that's

why I think communication and data visualization is very

important because it tells a story, right? The math will tell you the facts, the

raw facts, But the raw facts are not interesting. Let's be real,

right? Like, if you look— Totally. Okay. Um, you always see

it packaged into neat infographic. And one thought I had while you were talking

about numeracy at the population at large, you could probably

tell how numerate a society or strata of society

is— stratum, whatever the singular strata is— by how

often they buy lottery tickets. Because I think if people became more

numerate, the lottery system would collapse overnight.

Yeah. Interesting. You think that would be the case? Because I know a lot of

numerate people. So, so I actually debate this with my

wife, who actually does have a degree in math. And I'm like, look, if it's

a billion-dollar jackpot, $1 or $2 for one

ticket, I'll get one ticket. But I'm talking about the people who buy like 50

tickets, you know what I mean? Like, yeah, yeah. Okay. It's like, because like my—

the effort to go from zero To just slightly

above zero. Yeah. Versus

the burn versus the opportunity. In my mind, once it

hits $700 million and up, it's like, hmm, I see the sign as I

drive around and I'm like, hmm, you know, but if it goes over a

billion, then it's like, well, you know, there's zero and there's slightly

more than zero, right? But I don't have fantasies. And I know people

that grew up with that would just buy lottery tickets every week. play in the

same numbers. And I, even then I'm like, random doesn't work that way,

but they didn't want to hear any more smartass comments from me.

Yeah. Well, I think so. So one of the things that's really

important in this kind of conversation, to your point, is

when I think about probability and statistics, I think about, you know, the term

expected value, right? Which is the probability of the thing

and its kind of impact. So I think those 2

things are really important. I mean, you can't think about

one without the other because you're right, who cares? Probabilities are very

abstract and there's 80%, 20%, but what is the

impact of the thing? So if the impact of the thing is you win $1

billion, well, that changes

how I think about the probability. And if it's, you win, you know,

$100,000 and, you know, I don't know, maybe it's not worth it as much. Like

it's probably not worth Burning the gasoline to go to the corner store. Exactly. Well,

you live in Brooklyn. It's not worth the calories of— It's not worth it. Yeah.

Although it's good to walk. Yeah, that's true. Yeah,

yeah. But, um, but no, that's interesting. So I,

I have one comment on the last bit, and then I want to return to

what you said earlier about the numeracy

conversation. I've always heard that lotteries are a

tax on people who are bad at statistics. I've heard

that. That's kind of cute. Yeah. And, um,

but one of the things that I said in your, your conversation about

numeracy reminded me of it. As a,

a person who's led, I led a team of 40 ETL

developers at Unisys Corporation for 2 and a half years.

And one of the things I shared with several of them

were math majors. As I said, you know, my take on

statistics, especially applied to management

principles, which I see a lot of disconnects

there, speaking of innumerable, I would tell

my directors as I was a manager that I had this saying,

we can use statistics about everything to do with

people except people.

Interesting. What do you mean? And I

think the way you expressed someone catching

90% of some disease is probably

the best way to articulate that. They

don't experience 90% of a disease or even 51%.

It's as you said, it's a Boolean,

it's a switch, and it's either you've got it or you don't have it.

And I think that I, as an electronics

engineer, I refer to a lot of things as an impedance mismatch.

And I think that's, you know, that's my word, my engineering word for disconnect.

Or, you know, there's other words, cognitive dissonance,

for that. I think that's what fuels that, is our experience

doesn't match what the numbers appear to be telling us,

or perhaps the way we're interpreting or maybe even

misinterpreting the concept the numbers are expressing.

And I love that example. I'm going to use that.

Yeah. Yeah. Yeah. Feel free. Feel free. I can talk about probability

and statistics for a while. It's a, I think it's a very interesting topic.

It applies. It's the most practical, I think, mathematical

topic out there just because it exists whether You

want it to or not, it's everywhere. And, you know,

if you don't, like, I feel, I sometimes feel bad for folks who get

intimidated by charts and graphs and scary numbers.

And because math and statistics in particular,

I don't think it's taught very well in school. I think

it's very dry. It's super boring. Like, what the hell is a p-value?

Like, please get me out of here. Right. All those weird Greek

letters and— Yeah. And then you lose people. And then it turns out,

you know, a decade later after that class, they realize like, man, I really should

have spent more time understanding interest rates and credit

scores and so on and all this other

stuff. And like, frankly, this is going

to sound weird. I'm not into like betting and sports betting and stuff like

this. Like, I don't think that's good for society. writ

large. But I think some of the most

numerate people are like the sports bettors

who understand the odds and the over-unders.

And they just, a lot of these, and people who play poker and understand. And

yeah, I kind of wish that it was something that's a little bit, that was

a little bit less toxic. But they're like, it would be

great if there are other spheres of society in which

folks who participated, who otherwise are

not, you know, into numbers and, you know, aren't mathematicians

or, you know, kind of anything like this, but they were just interested for the

sake of a game or for the sake of a community or for the sake

of spending their time. And it forced them to have to understand this

language because those folks get it. They're super, they're whip

smart. They understand how to make decisions. They understand risk

management. And, and yeah, so like every time I talk to somebody

who does like sports betting or anything like this, like they just get it really

quickly. And not 'cause they studied p-values and

probability distributions, it's like they wanted to partake in this

activity. In order to do that, you do it over and over and over and

over again, and you build up an intuition and you kind of just see through

the numbers. I think, I think that's— That's a great way to put it.

'Cause like I was listening to the audiobook version of Casino.

which is obviously that famous movie with De Niro and Sharon Stone

and Joe Pesci, right?

Fantastic. But in the audiobook,

there's a whole segment where the guy, I think his name was Frank, which

is kind of funny, was talking about how he gathered all of

his— there's a whole segment. He goes on a soliloquy about how like he knows

more about the team than the coach does, right? He knows if the head football,

the head quarterback has had a fight with his girlfriend or like his mom's

got cancer. Like he knows all of those extra variables. And that's

what made him— he credited like that level of depth and knowledge and

almost spycraft into enhancing his odds, which is why

it's in the movie, why they kept him alive. It's because he was

really good at that. Yeah, yeah, yeah, exactly. You know, and

then there's like less maybe challenging kind of,

um, uh, or, uh, pernicious examples like Moneyball

is maybe like the classic, right? That's another one. Yeah, yeah. Yeah. Right. And it

kind of broke through to the mainstream a little bit. Now there is a bit

of a backlash, I would say, for analytics in sports. I don't know if y'all

are Knicks fans. Frank, you might be. Kind of. Yeah. Yeah. But they had a

great year even though they weren't supposed to. So that's exactly right. They weren't

supposed to. And I forgot what, what, what, which coach said it

could have been the Sixers or the Cavs, one of those series or the

Hawks. But he basically said like, analytically, we beat them.

This is after they had lost to the Knicks. He's like,

analytically, we beat them. Or it could have been analytically, we should have won, but

it was something like this. Right, right, right. I kind of take offense to it

because I'm like, no, no, you're giving analytics a bad name because now nobody's going

to give— nobody's going to care about

analytics when you say something like that because they're like,

well, I don't care about your numbers because the Knicks won against the odds. Right?

And then it just so happened that they kept winning against all the odds,

like all the time, which was wild and very exciting.

But yeah, sometimes analytics can go too far, right? You stare at

the numbers and you say like, this must be the truth

because I ran the model, I ran the numbers, I ran the simulations, and

this is the most likely outcome. And it turns out like, well, the world

doesn't care about your model or your simulations. Right. It's gonna do what it does.

And there is randomness going back to probability and statistics,

right? And you don't control that and stuff

happens and you just have to keep moving. And so

there's like a dogmatism with analytics. Yes. That I think we need

to move away from and we have been, you know,

despite as data-driven as you want to be, as much as you wanted the

numbers to drive all the rest of it, Spoken

as a data person, the numbers are actually a

data point. There's like a meta-analysis here where the data

is the data point. It doesn't matter how much data you have, whatever it is,

is a single data point. You know, another data point, qualitative

interviews of your customers. Another data point is the

intuition that you and your board or your executive team or your product man,

like, you know, the experts in your company, their intuition and their gut.

is another data point. Another data point is the story you want to tell

your investors or the story you want to tell your team, right? So

you get all, and then there's exogenous data points. So you get all these data

points together into your decisions. That

to me is data-driven, not whatever the model says

we're going to blindly follow. Right. And I think that is an important check

on On analytics, on data. Spoken as, you

know, again, spoken as a data person. I think that's a great way to put

it, right? The numbers are always a model of reality. It's not reality. Reality is

something we are unable to simulate as of

today and probably for the next decade or two,

even then. It's that whole map versus the

terrain. It's the map, not the territory. So

tell me about paradox machines because I'm sure there's Knowing your mathematical

background and goes this deep, you must have a good reason for that name.

And tell me about that. Yeah. So,

so the name, so I'm, so we're building Perplexity inside of

a holding company called Infinity Constellation. And they, they

do AI services companies. So that's, they're kind of incubating us as we grow.

The chairman of that company is very opinionated with when it comes to

names within that, Within the Infinity kind of family.

And I was opinionated as well. So we had a really good, and he's a

kind of a philosopher type. So we had a few good conversations around it

and we both mutually landed on paradox. And

for me, data, AI, everything that we're in right now is a

paradox, right? It's exactly the conversation that we're having.

There is data everywhere, but you know, there's noise everywhere

and we're searching for the signal. Yeah. Yes. Right. AI at once

will create new products

and new job opportunities and revolutionize the world.

And it's this kind of like quote unquote savior complex we have. And at the

same time, it's gonna take our jobs. People

hate it. Young people in particular hate AI, but it's getting shoved down our throat.

It's also stupid and dumb and like always gives me the wrong

answer in my chat. So we have to hold these competing truths at the same

time because both happen to be true, right? Yeah. And I

think that there's data and AI, I think there's this really interesting paradox.

So we both landed on Paradox, both as a brand that

we both believe in and I really like it as a play

on, you know, it's kind of fun, right? It makes you think a little bit.

It's a bit whimsical. It hearkens back to, you know, the Greek tradition

and I'm kind of a I don't know, also a philosopher type.

And I like to think about the foundations of our knowledge or foundations

of decision-making, right? When coming back to data and statistics.

So that's kind of how Paradox was born, if that makes sense.

Very cool. And your model is you embed it, you embed a

full team and like do the full stack for an organization? Yeah.

That's interesting. So would you call yourself a consulting firm or are you trying to

avoid the C word? No, you know, I'm not necessarily trying

to avoid it, but we do more than consulting. So just like

very quick history. So I did my actuary stuff. I was a data scientist.

I started leading data teams in various startups across

different verticals. I joined a friend of mine, started a consultancy called

Brooklyn Data, and that was a very successful consultancy

where we did a lot of Snowflake implementations, kind of modern

was, when the modern data stack was ascendant. It's kind of peaked

in '23 and has been on a decline in

various regards since then. That company was sold to a

private equity-backed company. He left, I kind of took over the data

team. I then left and consulted for myself and then got

the idea for Paradox and kind of got in touch with Infinity. So

that was, that's kind of the arc of my career in a nutshell. And so

I have the experience of leading teams and doing kind of on-the-ground

work as a practitioner. I have the experience as a consultant at Brooklyn

Data and implementing other people's technology, right? So the

Snowflakes and the Databricks and the Tableaus and Sigmas and all the rest

of it, and the Azures and the Fabrics of the world as well.

But we never had our own platform. Right? We are a pure consultancy.

So with AI, I wanted to take a stab at

building a— and there's plenty of platform companies. There's a lot of SaaS tools out

there of various flavors. Before AI, it was

always very expensive and fraught and a long endeavor to build your own

platform, right? To build your ETL and your storage and compute and your

orchestration, your security, and all the rest of it, all the good stuff that we

want in our data platforms. With the rise of AI

and coupled with experts. So my team are all like

staff level, like really senior engineers and myself who, you know, I'd

like to think that I know what I'm doing as well. I wanted to take

a stab at building a platform, but instead of having to

raise tens of millions of dollars over the course of several years

to try to stand up a platform, we would build it with AI.

And it turns out if you know what you're doing, And you have some funding

and you have some use cases, some customers who can't

afford a Snowflake and a Databricks and a Fabric, can't afford to

hire a team of engineers, can't afford that VP of data and AI

who's gonna, you know, do the strategy and the change management.

They can't afford all that. But that's the cost of entry.

It turns out that with AI, you can build them a

cost-effective platform. With AI, you can deliver

services with margins that both support your

business and they can afford, and it's still a sustainable business.

Now that is a really interesting proposition and that really hasn't happened

up until, you know, AI was capable. And that's, you know, as of maybe

a year or two ago. So we started Paradox and this model of

services and platform is rare. There's a lot of services

companies, There's a lot of platform companies.

There's a few that do both, Palantir being maybe the most

popular among them, and they've had a 15-year-plus head start

and really paved the way for what these companies can do. Popular is an

interesting word to use with them, but widely known

might be a safer word. Yeah, widely known. I'm not going to touch that. That

is the third rail of this industry. But yeah. Yeah, maybe

notorious. That might work. That's a little harsh,

but yeah. Widely known is a nice, safe, neutral. Exactly.

They are widely known. Again, put your thoughts and biases and

emotions aside. It's like facts on the ground. They're widely known.

And anyway, so I wanted to, and I have a lot of

feelings about the services industry and consulting as a business model. And I think that

the billable hour model is going away. I think we're seeing that with Accenture stock.

with stock price decline. I think we're seeing that with BCG and

McKinsey and how they're reimagining how they're

doing delivery. I don't think consulting goes away as a

function. I think that industry is here to stay and I think

it'll still be thriving. But I think how services are being delivered needs

to change because of AI. And I'm— and I was very

interested in building a company from the ground up

which was a services company. So I'm not afraid of the services word. We are

a services company, but we don't just do consulting and we don't just

configure other people's technology. We can build our own

technology and we can build it in such a way, as I said, that is

affordable for you and doesn't

pretend to be the be-all and end-all. Like, I'm not pretending

to rebuild Databricks. Like, that is a massive company

with hundreds of billions of dollars and All the rest of it. But the

mid-market, which are smaller companies, not

sub-enterprise scale, they don't need a

Fabric or a Databricks or a Snowflake, you know, not

yet anyway. Maybe they will eventually, but they don't need all that machinery,

but they're left with a, you know, what, there's no middle ground. What

do I do? Right. So that's kind of the idea is to

solve their problem both with the technology, but also with

services, because they need a partner. They need a partner that's not going to rake

them over the coals, that's not going to keep charging them billable hours and

scope creep and all the rest of it. So it's been a very

successful endeavor so far. A lot of traction, a lot of organizations

like what we're, you know, what we're producing. So yeah, it's been a fun journey.

I absolutely love your focus on, you know, small to medium

businesses, SMBs. That's where my software company is

targeted as well. And I see what you

see in kind of an emerging

application. And I'm intrigued.

I'd love to chat with you maybe differently than

publicly on the podcast here about your experiences

and what you see, especially given your statistics-driven mind

about that. I have notions. But I wouldn't call what I

have innovation. And I'm building based on those

notions. You talked about considering the risks, counting the

costs earlier and stuff like that. My MO,

because of just me,

is, you know, damn the torpedoes.

That's kind of a, got this idea, I'm going to go with it. I'm

that person you were warning us about earlier. Not

considering the the error bars at all.

And I'm trapped in here with me, Elon, so I don't really have a

choice to do that. I had to pick a path and run down it full

speed ahead. And every now and then, you know this, even if the error

bars say don't do this, sometimes the nicks win.

100%. And, you know, something that I tell folks as well, and this

goes back to the conversation we're talking about, You gotta make a decision. Sure.

And, and, and you can't always trust the numbers or like, I shouldn't say trust

the numbers, make a decision based on them. I believe this to be true, and

this kind of undercuts what I do, but I'm a pretty

honest broker, I'd like to think. Sure. Is there have never been

massive step function types

of decisions based on

data. They, there is no Fortune 50 CEO

that is out there deciding something meaningful

and thinking, well, what do the numbers say? Again, the numbers are, are

a data point, but they talk to, again, the board. They have their

intuition. They just believe something to be true

and they push forward. And, you know, you can torture the data to tell you

whatever you want it to say. And I don't think you should do that, to

be clear. But I'm real with myself, right? I'm not

pretending that, hey, if you're a small to medium-sized

company and you have your dashboard and your data, there you go. Now you

can compete with the big guys. It's like, no, like now you can make better

decisions and now you can have more, you can have more tools

in your toolkit. You can have more confidence and you can have more

remediation measures and all the rest. Like this is a net good, don't get me

wrong. But it's not that now that you have the

data cleaned, you are going to— it's going to solve your problems, right?

Like big decisions are still going to be made. I

believe this for better or for worse, based on

what important people at your company.

It's the HIPPO thing, you know, the highest paid person's opinion. Now, I don't like

it and I don't think it's necessarily the highest paid person, but I'm not

deluding myself into thinking that like, oh, data is what drives

Really fundamental. Again, like on the margins, it drives a lot of things, or it

should. But when you have to like pull the trigger or you have to

push the button to do the thing,

it's gonna be based so much on your intuition, your instinct. And because you as

a CEO or you as the like important person in the room, that's what you're

paid to do. This is what I tell managers. This is what I tell a

lot of people that I coach often when they're like, oh, should I hire this

person or that person? Or should I structure my team this way? And that

was like reasonable questions. You can make a case for either side.

You know, you can look at the data, but ultimately you've been hired and been

elevated to and put in this position to make the hard

choice. And you gotta make that choice.

And that's what I feel. Sorry, I kind of forget how we got on this

topic, but— That means it's a good show. Like, at least—

yeah, but that's kind of how I feel about the most important decisions are

You just like, you do it and if it's wrong, maybe

you get fired. You know, like, sorry, that's just the cost of making

important decisions sometimes. Yeah. I mean, that's the

risk-reward function at play. Exactly. Exactly. For good or for

bad. So we are, we could talk to you for another hour or two, but

we want to be respectful of your time. Where can folks find out more about

you and Paradox? paradoxmachines.com.

is a great place to read up about us. My LinkedIn,

which you can just, you know, LinkedIn my name, I should pop up there. I

try to be active there. Those are probably the 2 best. If

you have any cyclists or runners, you can check out my Strava.

I'm pretty active on the Strava app. And

yeah, that's where you could find me. Excellent. With that in mind, we'll

let the AI finish the show. Or the theme

music. I haven't decided how I'm going to edit this. Thanks for your time, guys.

Hey, thanks. Thank you. Hey, thanks, man.