Welcome back to Impact Quantum, the only podcast where we explore

the frontier of quantum computing and ask the real

questions, like how many SAT words can we fit into a single

episode? I'm your host, Frank Lavine,

joined as always by the indomitable quantum curious,

Candice Gilhooly. Today's guest is Clark Alexander,

a mathematician, quantum thinker, co founder of

Energuice. No, it's not. A startup selling

kombucha and self professed flania. If you've ever

wondered how quantum computing, AI and energy

markets intersect or how to irritate IBM with a single

slide, this episode is for you. We'll dive into quantum

advantage, energy efficiency, and why you can't just

build a Lego tower to the moon. Expect some strong

opinions, academic wanderlust, and at least

three existential crises about your electric bill. Let's

get into it.

Hello and welcome back to Impact Quantum, the podcast where we explore the emerging

industry and field of quantum computing and where you don't need

to be a physicist, but it does help if you're curious.

And with me, as always, is the most quantum curious person I

know, Candace Gooley. How's it going, Candace? It's great. I'm

really excited to be here today. We are going,

we're going, it's all good. We're going today to speak with

Clark Alexander, who is a mathematician and

he is co founder of Energuice. And it actually sounds

really exciting, his company. So we're definitely going to be asking him some

questions about that. Yeah. So welcome to show Clark and tell

us, tell us all the good things you're up to with Energuice,

which is a portmanteau of energy and juice.

And in the virtual green room, we were, we were busting out with the

SAT vocabulary words. So.

Right. I like, I like Portmanteau. I once got

an improv comedy show and they're like, give us some words that were SAT words.

I was like. Well, we've had two so

far. There was Flenore, which I was like, the

only person I've ever heard use that word in public was Nicholas

Nassim Taleb. And turns out you're familiar with his works.

And then we had Portmanteau immediately followed. So this is going to be the

SAT vocabulary word show. So not only we learn about energy

and quantum computing, but also maybe pick up a new vocabulary word or

two. But not like in the way when I'm stuck in traffic and my kids

learn new vocabulary words. Those are different types of vocabulary.

Well, thank you very much for having me. This

is exciting. I like to talk about what I'm working on and I like

talking about quantum computing and how it's affecting industry. And so I think we've landed

the right place for today. Awesome. So that's a good, that's a

good segue. Like where are we with industry? Right,

because we had a guest recently kind of talk about

how it's going to be an industry by industry type of

takeover. Not takeover, but it was like it's going to grow industry by industry.

And he's like, you know, will the airline CEOs care about quantum computing?

Well, probably not for another 10, 15 years, but if you're in the defense or

mathematics or even chemistry,

you're going to care about that in a much shorter time frame.

Sounds reasonable to me. But what's your take on that?

Yeah. So I want to pitch back to just one

week ago I was in Egypt for the first ever national

hackathon of Egypt. And it was co sponsored by

Open Quantum Institute, IBM Quantum Quantum,

the Bibliotheca Alexandrina was there, ICAFE out of

Netherlands. So Saleem, who you may have talked to, and then

Yusuf Eldakar were some of the organizers. They had

invited me to one be a juror on that at that

hackathon, which was amazing to see the, the progress being made by the

university students in the, the wider MENA region. And then also

they had me give a talk. And you know,

my thing is I follow energy. I was an energy trader a few years ago

and you know, I work in AI and I work in quantum computing. And right

now I'm looking at what are the energy limitations of

quantum computing. So this was, this was my talk. It ruffled a few feathers, but

it got people actually really thinking about it. So

sort of to put our listeners in the right mindset, those

viewing, I love to start with this question. This gets us sort of in the

right mindset. And the question is this. How tall

a tower can you build out of Legos? You know,

like just, just the bricks. Just take a bunch of two by fours. How tall

can you build that tower? Okay. And if you think about

this for a few minutes, well, there's, there's kind of two

obvious answers. There's the math answer which is just keep sticking the bricks together

for infinity. And then there's the physics answer and you

start asking, well, can I build this to the

moon? What happens to gravity? Can I build this

past geosynchronous orbit? How tall can you actually Build this thing,

right? Plus wind and like birds flying into it and stuff like that. Like,

so the, the analogy that we're trying to get here is that there's a math

answer and there's a physics answer, and in the world, live in this

sort of mesoscopic world. Here's a good SAT word for you. So

in the mitoscopic world, this middle thing, the math and physics really agree

really, really closely. Extremely closely. But

when we're talking about like galactic style stuff,

right? How do you measure how far away a star is,

right? You're not measuring the centimeter. You're

not measuring, you're measuring this to the nearest like astronomical unit. But you also

have to consider like how gravity is bending light, right? I mean

this, this is a very different realm of physics. The mathematics

is the same, but the physics has actually changed. Now the same

exact phenomenon happens at the quantum level, right?

Quantum mechanics has its own set of rules. There's physical rules that are not in

this world that we live in, right? They're mostly counterintuitive.

So we have things like the uncertainty principle, right?

In the, in the, the mat. The big world we live in, we don't

have to worry about this. And there's, you know, I'll give you a joke, right?

A friend of mine once said, I got pulled over for speeding.

And the cop said, do

you know how fast you were going? And my friend said, no, but I know

exactly where I was.

I mean, he was a physicist. And like, that was a really nerdy joke. But

people who have studied quantum mechanics are like,

actually that's, that's a good point. But you know, in this world we can know

how fast we're going and where we are kind of simultaneously, right? There's

some, some error there. But we're not concerned at 10 to the -35

electron volt seconds. That's not a, that's not in our

consciousness, Right, Right. So I mean, the,

this, this ends up being the point, right? At quantum computing,

there's this energy scale that we have to consider. There's actually a

large energy scale and there's a small energy scale.

And so to, to start with the large energy scale, let's start with the one.

We kind of understand this, right? How much build, how much energy does it take

to build a house? How much energy does it take to build a skyscraper? We

can actually measure that pretty closely, right? So

I'm looking at, say, these superconducting qubit technologies.

IBM is maybe the most forward and out there

according to their Blog. They use a 25 kilowatt

refrigerator, which they have to run for 96 hours to get

their qubits cold enough. Now, I gave this talk last

week and one of the guys from IBM who I

actually really quite like, he said, I think it's a 50 kilowatt refrigerator.

Like, okay, that's a lot of energy, right? So let's, let's say 25

give IBM the benefit of the doubt. Their scientists have figured out some

extremely awesome refrigeration technology.

But it's good to be an H. Vac tech, isn't it?

Sorry, I didn't mean to cut you off. Yeah, yeah, but

you do the math. It's 2.4 megawatt hours

of electricity to get to that computation.

And in this world, we can't ignore that overhead, we can't

ignore that time overhead, and we can't ignore that energy overhead. And so

you ask this second question. How much can you get done in four days

using 25 kilowatt hours of electricity? That's like 400

laptops running at full tilt, right? For four days.

Like, can you get a pretty good approximation of

literally anything running that fast? It's like

not everything, but an extremely large set of problems you can

get a good approximation for, right? And so

I was in a business meeting a few months ago with the

former head of Renaissance Technologies, and I pitched this question to him,

right? I can find you an approximate portfolio of stocks

that you want to trade which will give you, let's say,

28.1% return. Or I could run for

four days and I could get you

28.100000007

return. And he's like, well, I'll take the first one all day, right? By the

time the, the stock market's already changed in that four days, so

that, that 0.00007% return is, it's actually

negative, right? Because you're paying for that in time and volatility,

right? So what this, this does, this puts us in

what quantum computers can and cannot do and where they actually are going to be

advantageous, right? So for me, I like to, I like

to sort of say exactly what is advantage and what is

supremacy in the world of quantum computing? I think these words get used a

lot without like really defining them. So

I'm going to dig deep into my mathematical self and I'm going to give you

the definitions and your listeners and viewers can disagree with me all

they want, and that's totally fine. But from my

perspective, there's three things that we measure in modern

computing. There's speed, there's memory.

And the kids who have studied the beginning computer science algorithms will realize

you can trade off speed and memory. You can sort a list really, really,

really fast if you can memorize all of it. Right? So there's a trade off

there. Okay. But the third one now is really energy,

right? You look at the large language models

opening, reopening nuclear facilities, data centers, how much water

they're like Tulsa, Oklahoma had to go on water restriction a couple of days last

year to like cool these data centers down. So this is no longer

this sort of thing we think about at an industrial scale. This is

the main metric. There's energy, then there's speed,

then there's memory, right. Or energy and then time and then

storage. If you want to think of it this way, for me, energy

is like the prime metric now in quantum

computing space. I think advantage means that some

quantum computer chip system

has outperformed a supercomputer in at least one of these three things,

even on a specialized task. Okay. And

Supremacy would mean that a quantum computer is outperforming a

large, large, large set of problems in all three of these tasks.

Okay. So advance. We've probably seen

Willow, probably this Marco Pistoia when he was at

JP Morgan before, before he joined Ion Ionq.

They did this certified randomness. I

think that's advantage. I think that is advantage. They have built a very

specific chip to outperform in speed.

Building randomness on a classical computer.

I'll give them this, right. I think, I think that actually happened

for Supremacy.

I think because we have, at the moment, we have

this huge time and energy overhead, I don't think

we're actually going to be able to get ahead on time based problems.

Right. So I've worked in supply chain optimization and I don't have four days

to cool down a computer. So I can make a decision. I have to make

the decision 12 hours from now, right? If we have this

overhead that can't be discounted. And so there's no way a quantum computer can

actually beat that in time because they have

this overhead that you can't get around, right. There are physical rules to it. It's

not like, oh yeah, I have a quantum computer that's just always on, right?

With that amount of energy, if would. You throw energy into the mix, then yeah,

that becomes an issue, right? And I was thinking like, well, what if you rotated

it, right? Like you have one on one cooling? And I was like, well, you're

still spending. You still have. Absorbing.

Not absorbing. Yeah. You're still running a lot of energy. Yeah, that's

right. You know, and a few years ago I was talking, I

interviewed at Oak Ridge National Lab for their quantum machine learning group and

they were, they were installing Frontier at that time, which was

at that time the world's fastest and largest supercomputer. It's now moved to

second, but when it came online was the most energy efficient per

computation that had ever been built. And the guy directing

the building of this computer said, you know why we didn't build it twice as

big? It's because we couldn't afford the electricity bill. I'm thinking you guys work for

the doe, right? Right. Seriously, if anyone could, you

know, sign off on new nuclear reactors and whatnot, like, it'd be them.

I mean, this is them telling me they couldn't afford the

electricity bill. So there's some, like this metric has

like catapulted into like, this is the thing we actually really need to care about.

Right. At an industrial scale. And, you know, he worked out the math for me.

Roughly as you square the number of operations, you cube the amount of electricity

necessary. This is a serious,

this is a serious problem. It's funny because now you, you pointed something out

that, so I live between Data Center Alley in Northern

Virginia, which is Loudoun County, Virginia, which is

near Dulles Airport. So if you ever fly in a Dulles airport, all those

buildings are probably data centers and Three

Mile Island. Right. So one of the big

controversies here is they want to plow through a lot of farmland and

like remove, put in a new power line.

It goes basically straight from the Pennsylvania grid to Virginia.

And there's going to be, there's a lot of political drama, NIMBY

type stuff going on. NIMBY meeting, not in my backyard. It's not another

SAT word really. But.

But I mean, like, it's like it's serious and it's just like basically

the way the, there's a lot of shady deals going on where Maryland

customers are going to have to pay a surcharge for this reliability product project,

which is the electricity is basically going to go straight over our heads into the

next state. So I mean, this is a very real problem. Right. And you

can look, you can look online about, you know, kind

of stories about, you know, communities

that have had data centers put in and it wasn't exactly the wonderful

thing that they were told it was going to be. Right. So like, it's, it's,

it's interesting to see that. Now this is an issue. Right.

I long sometimes for the days when nobody cared about computers but other

computer nerds. Yeah, yeah. I mean I'm

in, in some ways I make computing great again. Right, right, right,

right, right. Mpga. That's what we want to do. Make it obscure

again. Again. I like that.

Yeah, we got the acronyms going today too. So

anyway, this, this is where I, where I am about how quantum computing

is going. I don't think supremacy is in the cards because there's a large

set of problems that we

can't either outperform on memory or time. Right. One,

energy or time memory is not even in the discussion yet. Right.

Story. Quantum storage is not even in discussion. I know there's a patent on

qram and I took to Mohammed Zadin who has that patent. I talked to him

last week and even he's not really a believer in

quantum memory over performing classical memory ever.

And he has the patent. Right. So it's not like

it's not some rando on YouTube. Right, right. This, this is the

folder I saw the patent itself actually, which was pretty cool. So

any case, he's, he's not necessarily a believer in this,

this third one, the memory piece. So

I think going way back to the earlier point,

what we're going to have to have is quantum hardware built for

specialized problem sets in which they can perform an advantage and

maybe two or three, two of the, two of the metrics that probably be able

to over forum. I, I see this happening. Right. And

to give yet another analogy, I was speaking with

the IEEE subgroup yesterday. We were working on our, our final paper for

quantum cyber security. And I told them this, that we're, we're discussing

Google's Willow chip. I'm a big Formula One fan. I've

been a big Formula One fan for a long time, since 92 actually, Nigel Mansel,

but you can look that one. Nigel Mansel, my man.

Weird dude, but good driver in, in modern

Formula one, they take the cars apart after every race and they

rebuild them and they sort of rebuild

them to be advantage, advantageous to the track they're about to race on.

Right? So this, this is some like really, really, really specialized race car. At each

track, it looks roughly the same, but they can tilt the front wheel a little

bit and they can, they can balance the tires a little bit. So if they're

going to be turning right a lot more than turning left, if there's banked turns

right. If there's a very, very long straightaway, they'll they'll let

the, the back wing come down, you know, a tenth of a degree more.

It's built specifically for the track. Right. They're not

allowed to memorize the track. That calls the disqualification. A couple years ago with Renault,

they had memorized the tracking in the brakes

that caused a disqualification. But they, they build

the car to, to the specifics of the track for the week. That's legal

right, to within, to within rules.

That's the kind of thing I think we're going to see in quantum computing. People

are going to be building specialized systems to solve specialized problems

and kick ass at doing this. Right, right now the,

this where, where we're actually going to see some advantage. You know, again, I

was sort of jostling back and forth with IBM about this.

This is. Well, I can solve something that will take 3 million years in 5

minutes. Okay. If that thing is worth 3 million years of

advantage, then I give you 4 days, I'll give you 8 days to cool your

computer down. It doesn't matter. Right, right. But stock trading

doesn't fall on this thing. But if you're talking about

improving, we end up speaking about the

Habermach process for making ammonium. Right. If you're

improving that by a fraction of a percent, the

payback is so, so, so enormous over, over just

a year that that energy usage is going to be wiped out.

Right. If you're doing something that's like a long term massive energy

reduction problem and you can solve this faster, that's

advantage. That's really a thing that has happened. But stock

trading, supply chain optimization, it just can't. Right? You

can't, there's, there are like physical barriers which you can't do that.

Right. So it really has to be for now

forget nisq. This is like specialized quantum hardware to solve specialized

problems. And I think, and, and I'm, I'm okay with that. I think that's a,

that's a really interesting scientific and

engineering problem to go into like solving. I want to solve this thing

better than it has ever been solved in history. Right.

That's, that's a, it's a worthwhile, at least scientific endeavor.

Well, so I'm the curious one, so I get to ask questions that sometimes seem

silly. But when you're saying the

quantum hardware that's able to do this kind of precision,

why would that not be different kinds of software?

Like I'm trying to understand the difference as to, as to

what would allow you to do this

kind of computation. And I thought that was more of a software thing

than a hardware thing.

Well, at this level, at present, they're not really separated.

Right. Because I think, I think where we are in the world of quantum

computing is we haven't even decided what a qubit really is.

Okay. There, there are nine known types

of. Geez. And,

and what I'm hearing this is, this is from the IEEE discussions are saying, well,

each one has its own advantages and disadvantages.

I mean, so have we decided what a Qubit is? Well, IBM

has decided what they think a Qubit is, but IonQ has decided something else.

Right. Because there's, there can be used for different sectors to solve

different types of problems. Like you have the ion capture and then you have

the super. You know, when we started learning about qubits and learned there

were nine different kinds and you know, every time we feel like we've

got our handle on the information, there's just a little bit more that's

released that we're like, no, we don't know anything. Two,

like mathematically there's two types, right? Annealing has these, like these

wires. Right. So if we're going to talk about topology a little bit,

the annealing is just like, it's a one dimensional qubit. It has, it's just

spin, positive or negative spin. And the,

the neutral atom or

the trapped ion or the superconducting cubits, they're like

full electron spin. So the, the annealing 1D wave

is like an S1. Oh, the circle. And then on the, the

gate side you have like S3. So a sphere sitting in four

dimensions. Right. This an S3. Right. So

even even that technology is like mathematically they're super

far apart. Even how you program them is different. Right. So it's just like

the analogy of a punch card computer to

modern digital computer. Just even that technology is different. So they're going to do

different things. Although punch cards are not so useful at this

time. No, I know what you mean. You mean like what's the type of architecture

we have now? Not von Neumann, but I know what you mean.

Like the typical. It'll come to

me later. But speaking of sat. Yeah.

Computer science, AP terms. But yeah, I know what you mean. Like a traditional,

what you would call a conventional or classical computer, that type of thing. Punch

card computer is a little harsh. But, but I know where you're going with

that. Two types of quantum

computers are not, it's not that far apart. But you know, I just want to

make the analogy so that the listener understands that

annealing and gate computing are really separate technologies and they

require a separate set of mathematics and a separate set of programming. Right. A

digital computer is like, or to be reductive, it's a little bit of

light switches. It's just a whole bunch of light switches, zeros and ones on and

off. A quantum computer has a fundamentally different set of physics

and so it needs a fundamentally different set of rules to program. Well,

annealing and gate computers are also fundamentally different. Like

topologically, they're a distinguishable type of things.

So it's not just a new species, it's like a new

category of species. Right. Like just program

a gate computer to do an annealing task. They're not the same.

Okay, so is that why companies like D Wave, they're, they're

heavy on the annealing side of things and they're, they're more

commercially around longer and maybe that's an

easier problem to solve? Well, annealing is,

you know, kneeling's been around for several thousand years. Right. And so I think

it right itself In I guess 99 when D wave

started to say like, yeah, actually we could probably do this quantum

annealing thing and make, make a specialized thing, right?

D Wave, D Wave has a specialized solver. It's, it's kind of a one trick

pony. And I don't say that in a dismissive way like it's an

amazing trick that it does, but it does a thing it's not going to

be doing. It's not, you're not going to have a D Wave GPT,

Right. That's the kind of thing it's going to solve. You're going to have these

really hard optimization problems which you can pitch as

binary optimization problems. So special purpose

computing. Yeah, yeah. And it can solve a

lot of really, really hard problems or solve or

approximate very closely a lot of hard problems.

But it's in a specialized realm, not just a general computer.

Right. So where do you think

the first breakthrough is going to happen? True

breakthrough? Like, is it going to be in precision? Is it going to be in

pharma with precision medicine? Is it going to

be in energy with EV batteries? You

know, is it going to be in finance

for, you know, what was that? The random number generator?

Like, what do you think will be the first true

breakthrough?

I don't, I don't want to maybe guess because

what prediction is hard, especially about the future.

Go more quotes. But I'll say this,

what I see, you know, I, part of my talk is I, I talked about

how cyber security is safe from quantum Computers forever and ever and

ever and ever. It just is. RSA 1024 is safe from

quantum computers because of this. There's an energy limit on the bottom. We can talk

about that later if we want. But they're only

safe from this particular style of attack.

Right. This quantum Fourier transform source algorithm is going to top out because

you have to do this. You have to rotate these

electrons so little, right. So the

readout becomes random. It's just noise. Right. You can't, you can't say,

I'm going to rotate this 10 to the

-7050. Right. That's zero. That's,

that's zero rotation that the rent. The readout will just be random. Okay.

There's no, there's no way you can produce so little energy to actually make that

rotation physically meaningful.

Right. Mathematically, it's fine. Rotate as little as you want. It'll

work. We've proved Shor's algorithm works mathematically in the 90s.

Physically, it can't. Right. There's, there's an uncertainty limit there. But

what, what, what all that does is that tells us

that the, the path that we're going is going to have some sort of limitation

when you're trying to get some so specificity, right.

You're. You're going to run into a limit the way we're doing it

now. This does not, however, preclude some totally

other algorithm and totally other way of doing

things from coming up. Right. Going back to Nassim Taleb, we'll go to

the Black Swan. Right. Earlier this year,

Ken Ono, who's an amazing mathematician, he's a number theorist, actually, I

think was working with Katie Ledecky, the swimmer. They were, they were doing some.

So he helped her with like, cracking the statistics. But anyway,

he's, he's a number theorist and he and two of his students,

Greg and I think Vaughn Iterson, they put a paper this

year redefining what prime numbers are.

Hmm. And they said, actually

we found out that if you take this polynomial and this partition function, partition

is just the number of ways you can add up a number to get there.

So five can be added up as four and one. It could add up as

three and one and one or three and two or two and two and one.

Right. These are partitions of five. How many partitions of the

integers there are this polynomial times this partition

function plus another polynomial times another partition function.

It works only on primes. It's just this sort of like, magical thing.

We've been thinking about primes. The Same way since Aristhenes, right.

2600 years ago. Right. We've been thinking about primes this

way for a long time. And now, just this year, 2025, we say,

actually there are infinitely many more definitions of primes.

This is the black swan, right? And so, you know, let's. Let's go, let's. I

love this. This quote is from The Zero Effect, 1998.

Bill Pullman's character says this thing, and even though it's a

comedy movie, it's so, like, philosophically deep. I kept. It says, if

you're looking for something, something specific, your chances of finding it are

very bad because of all the things in the world. But if you're looking for

anything, anything at all, your chances of finding it are very good

because of all the things in the world. And I think,

like, this is. This is where we are in quantum computing right now.

For. For specifically for Internet security.

Right. Somehow there's. There's a magical way in

which most technologies have two different sets of competing

technologies, but Internet security has never been that. It's just been key exchanges.

Okay? And so factoring large numbers has basically been

what Internet security is. Well, now you just need one

algorithm to break any one of infinitely many definitions.

I'm looking for anyone at all in any way, shape or form.

I'm not precluding this possibility at all. In fact, the

chances of this not happening are one in infinity, right? It's going to

happen, right? This thing is going to happen.

By whom? I don't know Where. I don't know by what type

of algorithm, I don't know. But the fact that there are so many

possibilities now, it opens it up in a way that we haven't

been thinking about, right? And this, this is brand new. This is four months

ago that this paper came out, right? So we're. We're

not there yet. So Internet security,

maybe at least the. The integer factorization part,

what I see actually happening, and

maybe I'll ruffle some feathers here. A good friend of mine I

went to undergrad with is now at Flatiron Institute. And if you follow Flatiron

Institute, these are four guys who, they take all

these claims about, oh, quantum breakthrough happens. Chinese researchers

have done XY thing that supercomputer could

never do. And about six months later, they say, actually, we did it on a

laptop. They've done this like four or five times.

Flatiron Institute, they do awesome stuff. And

for me, talking about quantum impact, being on this, like,

particular podcast is important that the real

measurable economic impact of quantum computing is

it is causing these guys like Flatiron Institute and guys like me who work in

evolutionary programming to rethink what our classical algorithms are

doing. We are getting better and faster and smarter

classical algorithms which are costing less energy and less memory

to do better things, to sort of push quantum advantage back.

This is a quantum inspired algorithms at the.

Generally the. Okay, some of them are, and some of them are just like,

oh, you know what, there's this randomization scheme we just weren't looking at,

right? Some of them are just pure randomized algorithms with

like a really clever way to do stuff,

right? Now I give this example, like someone showed me this is the

slickest line of code I've ever seen. And it was,

it was for a video game where when you're looking around in a video game,

what they want to do is make the, you know, all the

vectors are normal. So like when you're looking at the spot, you

turn around and you look. And what this looks like mathematically is you have to

take this ray of vision and you normalize it to

length one. Alright? So going way back to vector analysis,

you take the vector, you divide it by its length,

right? So square root of it. And so this guy found this way to

just take an inverse square root really, really, really fast.

And the way he did this is basically he got a really good first guess

and one linear approximation. And that's

absolutely brilliant. That's what he did. He took a really, really, really good first

guess. And so this thing can sort of run and it causes much less lag.

And you can, you can, you can see this happen in like, you know,

area game. So you, he's reduced the lag across the entire network of

all video gamers worldwide by just this clever,

right? That's pure, pure classical algorithm. But it was like a really awesome

randomized first guess. He figured out how to do that,

right? No quantum nothing. It was just like, oh, if you start

near the solution, you only have to do a little bit of computation to get

to the real solution. So some of it is quantum inspired

algorithms. Absolutely 100%. I work in that sort of area.

Genetic algorithms, Monte Carlo simulations. I think there's this like

biased field diagonal cross.

Optimize something. It's a, it's a terrible acronym that has

no word to it. But this to me

is like the, just the quantumized version of this 1992

algorithm called MCMCMC. There's three

MCs which for the listeners will be

Metropolis coupled Markov Chain Monte Carlo algorithms,

in case you're wondering. So it's not a rap group from the early 90s, though

it's unfortunate. Not. No, it's not in the native tongue school. Right.

I know Latifah and De La Soul would have put out the album of the

three MCs, but that'd be awesome. And

it got into like, philology in. In biological

classifications. But I use this actually for supply chain optimization

because the point is that instead of just guessing this one spot like

Monte Carlo algorithms do, it allows you to guess

many different Monte Carlo algorithms. And so it allows you to

find multimodal probability distributions a lot

faster. It converges so much faster. Right. It's just

pure probability. And I think, I think that actually inspired the quantum

algorithm for the biased field diagonalization,

at least to my reading. That's how it looks. Right. So the

quantum algorithm is classically inspired, not the other way around this time.

Gotcha. It goes both ways. Right.

This is a. So you wouldn't have thought of that kind

of. Naturally, you would not have thought that the classical

inspired algorithms would. I don't know if the. The authors

of that algorithm were thinking of it that way, but, you know, having. Having

used the other. The classical algorithm myself multiple times and,

and having read their paper, at least to me, it was just like,

you know, my neurons were lighting up, my neural network was saying, oh, these are

the same algorithm. These are the same algorithm. That's how it

rang to me. They might not have been thinking about that. And that's cool that

they have like a totally unique algorithm. But, you know, I,

I've, I've seen this algorithm before as a classical thing,

but even, even if they didn't know about it, you know, this same

sort of technique landed. Right. It's like,

it's like the name Soren. Soren is a Persian name, but it's also a Swedish

name. They just sort of landed on the same letters. Right.

Interesting. That's my take. Could be wrong,

but that's just how I read it.

I'd love to just take a little step back if I could. I mean, what

you do sounds legitimately

fascinating. And you know, what you're uncovering and

you're. You're at the, the frontier of

innovation, you know, I'm going to ask you, like,

walk me through a little bit of your career journey.

That got you. That got you to where you are

right now. Okay.

I hope you guys like random walks because. Oh,

that's all the type of walking I do. Fantastic. Okay.

You'll appreciate this. I've, I've done a couple

of, I've been to support

this program in Mexico a few times called Clueless. And one of my

former students from Northwestern is one of the founders of this. So he invited me,

said come talk. And, and one of my favorite events at this,

this week, it's like a one week intensive where instructors from Mexico and United

States come and teach like one week intense course on some sort of

science to high school seniors, college freshmen, college sophomores in

Mexico. Right. Because there's a lot of talent coming and they just don't have the

resource that was the point. But Wednesday night of this week,

whenever they do it, they have like the, the Science Cafe and they have the

instructors, me and some professors from University of

Chicago, from Harvard, they come and they ask us questions

and someone asked me about how do I

think about work, life balance, something like this. And I said

whatever you're expecting in the future is wrong.

That's it. But I don't mean start there. Yeah, but

what that means is that some things are going to far exceed

your expectations and some of your expectations will never

even get close. Right. Okay. Right.

So that's, that's kind of, and that's, that's kind of how my life has worked.

So I'll give you like just the really top down overview.

I finished my PhD in 2008 in non commuter

geometry and mathematical physics. What I have

learned, reading a lot the last two years is that historiography is a chaotic

system. If you start the story one year earlier, it changes the whole story.

So 2007 there was a whole hiring spree for non

commutative geometers and mathematical physics. And this was spurred on by Alan

Cohen's idea that he may have solved Riemann hypothesis using these mathematical

physics techniques. There's like this glut of non commutative

geometers getting in postdoc positions. The 2008 comes

around and I, I didn't get one of those postdoc positions. I got a teaching

job at Temple University. Go Owls. It was

awesome. But I was going to say at. Least you didn't work for a mortgage

company. So. Right. Well, almost, almost happened.

You know, think, think all it didn't. Right. You'll go randomness all the way.

2008, you may have remembered there was like a massive financial crisis.

So there were a lot of postdocs of three postdocs for three years got

shortened to two postdocs of two years. So I got double

caught up in that. And Then I basically came what

they would call in the sports world the journeyman. I went to southeast India to

Institute of Mathematical Sciences for a year. And then I came back.

A professor had died days before semester was supposed to start, and I

just ended up getting a job at the University of Wisconsin Parkside to fill that.

That was completely random only because I had known someone here in Evanston who was

doing this. I taught there for two years. Then I

landed at DePaul lecturing one year. One year. One year. I was at DePaul

for six years. And then I moved to UIC for a half a semester. And

I never was going to make tenure. Right. That those days had

kind of passed for me in some sense.

And so a friend of mine who I'd gone to Northwestern with had started a

company, and he. He called me and said, clark, I'm doing this

thing in data science, but it's not. It's not traditional data

science. I need some real mathematical firepower, and I don't have it. You want to

come work with me? And he went

to my wife and said, you need to convince Clark to come work with me.

And so my wife said, clark, you need to go work with him.

My friend Rami pulled me out of academia and started me into industrial

mathematics. And I didn't know how to program a computer, and so I learned

there. Rami, unfortunately got sick and he. He died

a few years ago. And so I kind of have made my way

from there. He got sick and then. Then Covid

happened and I left that company and I joined an

electricity trading company through another roundabout connection that I knew from India.

Completely random. A mathematician was like, I want a mathematician to help me trade

electricity. Okay. So I did that. There was a

electrical storm in Texas, you may remember, like, there was an ice storm.

Everyone lost all the money. So my company went under. I lost a job again.

Fantastic. Started just applying

everywhere. That was when I was applying at Oak Ridge. And then I ultimately took

a job at a credit card company that didn't work out

for whatever reasons. And then I joined a logistics

company where aforementioned, my friend Rami was

supposed to be the head of AI, and when he was. When he was

really sick, he had called the CEO and said, hey, you need to take Clark.

And so that's how I landed there. Interesting.

Mentioned a couple of times. Energy trading.

What's the dollar store description of what

energy trading is? I'm not quite sure because I know it comes up a lot.

Usually when there's a crisis, people are suddenly experts on energy

trading, but like the Texas crisis, plus there was some

drama in the early 2000s in California and I think

the most infamous energy trading company in the world is still. Enron,

many orders of magnitude. So. Yeah,

well, if you're going to blow up something, blow it up big.

But what, what is energy

trading? I don't quite get it, right. Because like, and this has come up, you

know, I'll tie it back to the issue with Maryland and

Virginia and Pennsylvania, right. Like they're talking about they buy

energy from here and they do that I don't quite understand.

I can understand how the math would work in terms of optimization and

probably what you do, but I don't understand the industry. And I realize this is

the Quantum podcast, not energy trading, but what,

what's like a good two dollar description of. Okay,

fantastic. I'll give you two really easy problems and then I'll tell you why quantum

computing is important. Okay, so good, you're tying it back in.

That actually just happened recently. It'll tie all back in. Great.

So there's, there's two ways the energy trading sort of works, right. The easiest

one is you go to a city, let's say Madison, Wisconsin.

Right. Wisconsin goes to the regional transmission

operator or independent system operator, depends on how they're named. So you've heard maybe

of Caiso, that's California ISO. And then

you've maybe heard of miso, which is where I am, Mid

Continent Independent system operator. So the ISO or the

RTO controls like all the energy flow and it controls the

pricing. So the city of Madison, Wisconsin will say, okay, I want

to buy, is basically a futures contract. I want to

buy this many gigawatt hours of electricity

that you give me from January 1st to December 31st

of this year. And I want to pay this much per

kilowatt hour for it ahead of time. And in this way Madison,

Wisconsin can now sell to their residents at

whatever marginally marked up price. Right. So we want to buy it

for 12 cents a kilowatt hour for the entirety of the year. And we're going

to make a deal for, let's say 500 gigawatt hours,

whatever they make, I don't know how much Madison uses. And then so they sell

it to all the, the, the independent

households and the schools and the businesses for 15 cents a kilowatt hour. And that's

just the price of electricity for the whole year. Right. That's one way to do

it. That's a, that's a four year contract. Okay. They make the deal

the one in trading. So you could do that if you're, if you're a

municipality, you trade this way. If you're an individual little brokerage

house, you will say, okay. The ISOs and the

RTOs actually set the price of electricity. And what they do is they say,

okay, 9:00am today, so this is just a few hours ago.

They set the price for tomorrow's electricity

pricing. They set it at 5 or 15 minute increments, depending on where you are.

So say every 15 minutes we're going to charge this much for electricity.

Okay. This is called the day ahead price. Okay.

And so what, what happens is these little traders can come in and say,

okay, actually I think it's going to be less than that.

Okay. It, the real time price is going to be less

than that. So what I'm going to do is buy the real time price now

and sell it at the, the actual. I'm gonna

buy it, buy the day ahead price and sell it at the real time price.

Right. So they make some money. Or you can sell it as like short

selling. Basically you can sell it, you think it's going to be too expensive, you

sell it and then you buy it back at the, the real time price.

Literally. I think this is called day ahead real time. So in, in trading they

call that the DART model D A, R, T. Right? That's,

that's the simplified version. And then there are options, all

sorts of exotic options and, and hedging and

all kinds of stuff. You know, they run it like a hedge fund, except that

the commodity they're trading is time based. Very, very, very strictly time

based. That's how it works. Okay. So you know, Con

Ed is kind of like the supermarket. And then whatever the

supermarket buys their food and their groceries and distributors is

kind of like that, the back office to all of that. Right. And so

the RTOs and ISOs have this question like how do you set the price?

And so what you want to do in. So this is a massive, massive

optimization problem. This is probably the most important, most

worthwhile optimization problem you've never heard of, called the AC opf.

This alternating current, optimal power flow. So

what you want to do if you're making the electricity, if you're a generating plant,

you don't want to just distribute more than you've made and you

don't want to have shortages. So you want to balance best you can

in real time the supply and demand of electricity.

Okay. And this takes into account

congestion. Where there's construction, there are voltage angles, there's

like you know, there's all, all sorts of things, pricing. So if,

if you're a mathematician, this is the most exciting problem because it's non

convex, non linear, time dependent, directed graph,

acyclic graph, cyclic, whatever, whatever non thing you

can think of. This is the problem for you.

Like a 0.1 percentage in improvement. I think I did the, the math on

this. If you improve the efficiency of this solution by 1% and

are actually able to successfully trade on it, it's like a billion dollar a day

benefit. Oh wow. So no wonder why it's

run like a hedge fund. Yes, but like bigger than that. Way, way,

way, way, way. Right. Because the electricity market is

so much bigger than the stock market because everyone uses electricity

all day, every day. Right, right. And it's, it trades on

companies and trades on everything. And there are, there are options and there are municipalities,

there are big players, there are little players. This is a big market. We're talking

like size of 4x. I mean massive, massive market.

So wow. The AC OPF extremely

difficult. The way that people make money is that the, the

optimization is called the D.C. oPF and D.C. oPF is direct

current, optimal power flow and that has a convex solution. So you

can simulate this and solve it very quickly on a digital computer.

You need a supercomputer, but you can solve it quickly. Right. Minutes.

Right. It's a minute solution, not a, not a

millions of years solution. So one of

the main problems, the ACOPF has sort of sub branches. One is about

pricing and one is about actual energy delivery. It looks like

IonQ has recently worked on the unit

delivery problem. So given a particular

power plant, where does it deliver its units of energy? I guess they're

doing them, they're probably scaling them in kilowatt hours. Where does it

deliver kilowatt hours at 15 minute intervals? That is an

extremely difficult problem. And it looks like IONQ has tried to

tackle this at least at a small scale. Right.

So this might be one of the major breakthroughs. Just the problem is the amount

of memory needed. I think it will

overwhelm any quantum computer that currently exists.

But this might be one of the major things. But

ACOPF is like worth not a little bit of money,

is worth a lot of money, extreme amounts of money.

So wow, this has been interesting and I like the fact that

this is literally every time you flip the switch, like this is a

mathematical problem. So kids, if kids are listening,

math is super important and

that cannot be said enough. Seriously, Seriously.

But look at the exciting things. He's doing because

he started with math. I mean, this is just

outstanding. Interesting. Like, this

would captivate any, you know, any Gen Z

kid out there. Like, you know, you can tell she's Canadian, she lives in Canada.

She's not. Right? Because I say I born New York,

born New Yorker, born and bred. But now I say Zed because I'm in Canada.

Still on Mid continent ISO.

Honestly, Clark, you've been absolutely fascinating. I've

loved every second of this and I absolutely want to have you back on

because I have so many more questions to ask that, that we didn't get to.

So I, I just, I'm blown away right now. I've learned. I've learned a lot.

I've learned a lot. And I have to like, digest, you. Know,

to just the explanation of the electricity

markets and how they function is worth it because I just. All I remember

is, oh my God, Enron did all this

fraud and then you didn't, you didn't hear about it for years

until everything went sideways in,

in Texas. It's like, oh, well, the energy companies blame the energy

traders and blah, blah, blah, blah. These people do this. And I'm like,

oh, these people again.

Yeah, yeah. So that was, that was a totally different issue.

Maybe we can get into that if we, if we go again. Yeah, another time.

Yeah, yeah, yeah. But where can folks find out more about you and

what you're up to? Basically, I'm

mostly on LinkedIn these days, starting another

venture called Argentum AI, which we're trying to do energy efficiency in.

In AI training. Right. And we distributed

training. So Argentum AI is one of my things introduce.

We're trying to do some projects with the DOE,

but mostly LinkedIn. I'm. I'm kind of just

mostly there most of the time. Yeah. And.

And if you're into soccer, I'm the local soccer commissioner in Evanston, so come out

and see me on Sunday. Cool. Awesome. That's

awesome. And we'll let our AI finish the show. And that's a

wrap on another episode of Impact Quantum, where the topics are dense,

the qubits are entangled, and the guests are occasionally

flaneurs. Huge thanks to Clark Alexander for joining

us today and proving that mathematics isn't just useful,

it's a passport to energy markets, quantum

hardware and mildly unsettling jokes about the uncertainty

principle. If you enjoyed this episode, be sure to, like,

subscribe or entangle yourself with our past

interviews. You can find Clark on LinkedIn,

energuce on the cutting edge of renewable innovation and

candice trying to remember which qubit type is currently

trendy. Until next time. Remember, classical

computing may be fast, but quantum computing has

better party tricks.