EP#439: The Nuts and Bolts of Quantum Computing - Rick Altherr - Circuit Break - A MacroFab Podcast

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Rick Altherr (00:10):
Before we start, I do need to say that while we
will be discussing my employer,IonQ, any views or opinions
expressed belong solely to meand do not reflect the views of
my employer.

Parker Dillmann (00:20):
Welcome to circuit break from Macrofab, a
weekly show about all thingsengineering, DIY projects,
manufacturing, industry news,and quantum computing. We're
your hosts, electricalengineers, Parker Dillmann

Stephen Kraig (00:31):
And Steven Kraig. This is episode 439.

And this week, we welcome Rick Altherr on the
podcast.

Rick is a full stack engineer having worked on
everything from ASIC design touser experience in embedded to
hyperscale. Their career haskept them close to the hardware
software boundary, primarilyworking on computer systems at
Apple, Google, and OxideComputer. After a detour through
firmware security, Rick is nowdesigning instruction sets,

(01:03):
microarchitecture, and real timeembedded control systems for
trapped ion quantum computers atIonQ. Well, thank you so much,
Rick, for joining us on thispodcast.

Yeah. Thanks for having me.

So before we jump right into your your
current occupation of working onquantum computers, let's, I I I
how do you start working onquantum computers? Let's just
start with that.

In my case, it was definitely an accident. Right?
Like, most folks that I workwith, come to through quantum
computing through academicacademia. Right? Like, quantum
computers are still very much anactive, like, research and
development type space.
And so you have a lot ofphysicists and a lot of, you
know, specialist engineersworking, you know, in, like,

(01:56):
cryogenics and things like that.And they usually come with that
academic path. In my case, Iactually was applying to be
director of security for IonQ. Ihad been doing, firmware
security and, like, PC platformsecurity for quite a quite a few
years, and that's what I hadbeen doing at Oxide Computer.

(02:18):
And, it turned out that Iapplied, and they're, like,
well, we're actually not gonnabe hiring for that role right
now, but we think you'd be agreat fit for our embedded, you
know, controls team.
And I'm, like, sure. Why not?Like, let's just talk. And, so
that's how I got in. So I Iliterally came in with no
understanding or of how quantumcomputers worked.
And they're just, like, don'tworry about it. You know a

(02:39):
little bit about control theory.You're probably good to go.

Don't worry about it. We don't know either. Right?

You know, actually, like, that was
honestly part of it. And and mymanager, like, he he worked in
radios for Collins and, like,GPS receivers and things. And he
comes from a very strong mathbackground from the control
systems approach. And he's justlike, this is the kind of stuff
we're actually working on. Like,that's how the machine really
works.
And so you don't really have tounderstand the physics to do a

(03:07):
lot of the embedded systemdesign, because it's actually
classical control systemsproblems.

So I'm curious, about getting into the nuts and
bolts of of actual quantumcomputing. So not necessarily
the software or the algorithmside, but more of the nitty
gritty of what actually isinvolved in building and
operating a quantum computer?

There's a lot and it's also kind of a hard topic
to to get into because quantumcomputing these days is kinda
where classical computing, asthe quantum folks call it,
right, like, digital computing,was back in the 4 early forties.
You know, you're looking atfolks knew it was kinda possible

(03:51):
to build digital computers,like, store program digital
computers, but they didn'treally understand what the best
way to implement memory was orhow to encode instruction sets
or any of that stuff. It was allnew, and they just didn't know.
And so there were a lot ofdifferent designs that were
happening in parallel, differentpeople taking different

(04:12):
approaches, and just doing a lotof experimentation and
refinement, and they were incompetition with each other.
And, eventually, you know, westabilized on what we now think
of as a computer, but that tooka very long time.
Quantum computing is still inthat very early stage where a
lot of folks are trying a lot ofdifferent things. So,
ultimately, the nuts and boltsof of quantum computers is

(04:33):
you're using some physicalelement, to encode quantum
state. And quantum state is kindof a a complicated concept, but,
essentially, it's usually, itfrom a math perspective or,
like, a physics perspective,it's a a 2 tuple where each

(04:54):
element is a complex number.Right? You're encoding
essentially, a real andimaginary component of, like,
two parameters of something inin physics.
Could be the spin of theelectron. Could be a a charge,
electron charge state. It couldbe There's a lot of different
options. But you're working init at that level, and you're
trying to use, essentially,individual ions or neutral atoms

(05:17):
or or some sort of physicalproperty, and then you're
exciting it in some way tointroduce or to perturb the
state of it in a way that youcan then do some other thing to
detect an outcome. And sothere's a lot of different
approaches.
IonQ happens to work withtrapped ion. More commonly, you

(05:38):
hear you see, like, the bigthing that looks like a a
chandelier hanging down, andthat those are superconducting.
So, in that case, they'reactually using, silicon
lithography, just like making aa CPU die or or any other IC.
But they're building, microwaveresonators with a special area

(05:58):
that they can manipulate atcryogenic temperatures to
create, a synthetic, like,cubit. Right?
It's like a it's it's a nonnatural state that it gets kept
in, where in a trapped ioncomputer, you're actually taking
some sort of elemental source,like, Ytterbium, and you're,

(06:24):
ablating that with a laser andthen catching individual ions
from that plume of of an atomicsource, right, you're capturing
it in a electromagnetic field,where you're actually holding it
in position with a combinationof r f and d c to hold it inside

(06:47):
of the ion trap. There's a lotof other different ways of doing
it. I mean, I could justprobably I could probably name
off 6 or 7 other ways, butyou'll mostly hear about
superconducting. That's the onethat tends to get the most news,
and trapped ion just happens tobe what I actually work on.

So is is your control theory that you work on,
is that what's regulating the RFthat's trapping these ions?

Not so much. So the the trapping part is really
usually a pretty constant state.So effectively, you, you the way
it's often described is you'retaking the ion, and if you think
of it like a ball, and you putthe ball on, and you you take,
like, a salad bowl, and you putthe ball on the the bot you

(07:34):
know, the top of this invertedsalad bowl, it's on an unstable
surface. Right? And so it'sgonna roll off in any direction.
You don't know where it is. Whatyou're doing with the RF and the
the electromagnetic field iscreating a a modulating pattern
that's kind of like you took a ahorse saddle and we're spinning

(07:56):
it really fast. So it's an everchanging pattern, but it's
repeating in a in a, like,merging complex sinusoidal
pattern between the 2, and it'sholding the ion in that neutral
space in the middle. Right?Like, where the interference
pattern is.
But, most of the time, what youdo is you you're only
manipulating that when you'retrying to actually move the ion

(08:18):
within the trap. So it happensduring the initial loading part,
where you've captured the ionfrom the plume and then moving
it to the position where youwant it in the trap. So, there
is some aspects there and thereare times where we would do
wanna move it around for variousreasons, But most of the control
electronics is actually aboutthe other part, which is, well,

(08:39):
now I have this trapped ion. Howdo I actually influence its
state and then also domeasurement of it? And that is
entirely done with lasers, intrapped ions well, in in in ion
q's trapped ion system, it'sentirely done with lasers.
There are other systems that useother ways of doing that. But in

(08:59):
the ion q trapped ion systems,they're it's a jeez. I would
have to go and count somethinglike 14 or 15 different lasers
that are actually beingcontrolled, that are doing a
variety of things because you'reactually doing cooling, of the
ion. So you're, you know, lasercooling, is a is a thing that

(09:24):
happens, because as you'reinfluencing the ion, you're also
introducing energy into the ion.And, eventually, it will get too
hot and, cause it to, like atthat scale, heat is energy.
And so if you insert too muchenergy or the heat gets too hot,

(09:45):
it actually will thenpotentially kick the ion out,
you know, kick the electron outof the ion or or other problems.
But by using lasers, you canactually reduce the amount of
energy in the the ion, and andso you kind of are constantly
doing operations and thenrecooling it. The detection
mechanism used there is iswhat's called optical pumping.

(10:07):
So, essentially, you're you hitthe ion with a laser, which
pushes the energy state up acertain amount. And depending
upon where you started from, youwhen you then hit it with
another laser, it will eitherfluoresce or it will not.
So that becomes your detectionevent. Right? So, like, when

(10:28):
people talk about quantumcomputers at more of an an
abstract level, they're oftentalking about there's this
hidden state that you can'tmeasure. Right? Like, when you
measure it, it the state isgone.
And what we're really talkingabout, you know, from a physical
construct, like, nuts and boltspiece is, it's very hard to keep
track of the state inside thethe the electron. Right? And you

(10:53):
can't actually know it. Thatthat is actually, you know, a
fundamental part of quantumphysics. But you do have to time
your laser pulses that influencethat state according to the
rotation of the electron in theion.
And so there's a whole complexpiece of doing accurate

(11:14):
timekeeping to know when to timeyour laser pulse so that you
actually cause a rotation at thetime you want and by the amount
you want. And so a lot of thecontrol system is actually doing
modulated laser pulses at veryprecise timing, potentially of
many lasers simultaneously,alignment and can only be

(11:41):
tracked for so long. Like, wedon't know what the actual
natural phase precession, or,you know, like, the natural
rotation freak speed is, but weestimate it closely. And so
there's always some error. And,eventually, that error accrues
too much to the point where youlose coherence, and then the
state is also gone.
Right? So there's a lot there's,like, complicated things

(12:02):
happening at the physics layerthat translate to, I need to
actually track the timing of mylaser pulses at 4 nanosecond
resolution, but I also have tohave the precision and
correlation of multiple lasersmodulating within picoseconds of

(12:24):
each other. And so that's wherethe real time control systems
are are in there and and how itreally works. Now, of course,
that's all in the how it worksat the, like, physics and
electrical level. The what doesthat actually mean in terms of a
higher level is a, you know,it's easier to think of it in a
very different different model.
Right? Like

So you're talking, you know, picoseconds,
being able to like, real timesystem that is in that frequency
range. Are you running, like,FPGAs or custom ASICs, or how
does that work?

Yeah. So it's, it's many, many FPGAs. The,
they're they're large FPGAscoupled with ADC or with DACs
and and some ADCs. There's,servo loops involved and and
other things. But, largely, it'syou can think of it as a very an

(13:20):
arbitrary waveform generatorthat has 60 plus channels.
Right? And you get to programevery single channel
individually, and each one ofthose channels can do, like, 2
tone modulation. And those areused to emit like, the output of
those arbitrary waveformgenerators are coupled into

(13:42):
other elements like electroacoustic modulators, where or,
or sorry. Electro opticalmodulators or acousto optic
modulators, which are, you know,devices that either based upon
the sound pressure or basedupon, electrical impulses act

(14:05):
like shutters or modulators fora laser pulse. So you have,
like, a a fixed laser goingthrough it, through these AOMs
or EOMs, and the the outputsfrom the AWGs are actually
causing the modulation on thatlaser to that external device.
So you can kinda think of thosedevices as, like, upconverters.
Right? Like

And and every one of these hypothetical channels
from this, function generatorcan all be synced or desynced or
however you choose. Correct?

Yeah. They all operate off of a common clock,
and so they're all running at250 megahertz, you know, 4
nanosecond time base. But thattime base is synchronized within
picoseconds across the entiresystem. Okay. So you

Yeah. That that answered my question. Now my
other question is, I don't knowif you you can answer this one.
Oh, man. I just lost it becauseit was more about, I guess we
just forget it.

Well well well, I Okay.

So come back to me.

Yeah. I have I have one question. You were
mentioning, basically, gettingout of phase or out of sync with
the ion itself over some periodof time. Is that something that
you have to regularly reset soyou can be back in sync with it,
or is that not even possible?

Yeah. It's, so we talk about it in terms of the
coherence time. Right? So,essentially, the the ions are
always moving at some phaseprecession, and that is
influenced by its physicallocation inside the trap, the
ion itself, and then what laserpulses have actually been

(15:48):
imparted on it. So there's,like, a a multipart thing where
you can kinda estimate what thatrotation is, but it's always
rotating.
Right? Like, every particle isalways or every atom is always
rotating that way. So,initially, you don't know where
it is. So when you start, youbasically do a a laser pulse

(16:09):
across all of the ionssimultaneously that establishes
a baseline time. And then you'recalculating your phase offsets
from that point forward.
So you you kinda start with, Idon't know absolute phase, but I
establish a a common referencepoint for relative phase. And
then it's that you can trackthat relative phase for some

(16:30):
amount of time before your errorgets too large, and then you've
lost track of it. And that's thecoherence window. And so when
you run a quantum circuit,you're actually breaking it down
into these chunks, which youcall shots, which is, you know,
doing that sequence. So it goesthrough what we call CPD, Cool

(16:51):
Pump Detect.
So there's actually, you know,you cool the ions, you, run your
actual circuit, then youoptically pump them, then you do
your detection. And thatsequence runs. At the end of
that, because you did thedetection, you've already lost

(17:12):
all of your quantum state.Right? Like, this this is that
whole thing of, if you measureit, the quantum information is
destroyed in the process.
So you're collapsing the quantumstate down to a 0 or a 1. And so
then at that point, it doesn'tmatter what the phase is, and so
you start over at the beginningand you establish a new relative
starting point, and that becomesa new shock. But you can reuse

(17:34):
the ion many, many, many times.Right? Like, the ion remains in
the trap for a very long time,but your coherence interval
might be, like, one second.

Can can you just leave an ion in the trap? You
know, as long as you keep thesystem on and running, can you
just do it basicallyindefinitely?

You can do for a very long time with the trap
diagon approach. Right? And thisis where, like, the different
approaches come in play. Becausein a superconducting approach,
like I said, they actually sortof artificially construct their
qubit. Right?
It's not a natural state. Andtheir coherence times are, like,
microseconds. So they have to doeverything. All of their

(18:14):
operations have to happen ordersof magnitude faster, which is
why they use microwaves insteadof using laser assemblies.
Right?
And it's, like, there'sdifferences in the the physics
itself around how fast youroperations need to be and how
long your coherence times are.And so there's different trade
offs. Now, in the trapped ionapproach, you can hold on to the

(18:36):
ion in the trap for a very, verylong time, but your coherence
interval is gonna be, you know,1 to 2 seconds, which is nice,
but using lasers is slow. Soyour gate time, you know, your
operation time, the way quantumcircuits are represented is is
thinking they're they talk aboutas quantum gates in a quantum
circuit. It's not really like anelectrical circuit diagram.

(18:56):
It's more like a a mix of, like,a circuit that progresses over
time. Right? Like, if you had aschematic that changed or that
showed a progression over time.I know that's probably not
terribly helpful.

4 d schematics.

Yeah. I mean, they're a lot simpler, though.
It's, like, you essentially haveall your cubits running left to
right, like, you would, youknow, on a on a musical, scale.
Right? But then you're showingoperations as it's between you
know, it's this type of gatehappens at this time across
these cubits.
So you're showing connectionsacross it, and they just call

(19:34):
that a a quantum circuit. Butthose individual operations, how
long they take also varies withwhich type of system you're
using. So, like, you you know,your gate time on a trapped ion
computer might be microseconds,whereas your gate time on a
superconducting one has to bedown in the, like, you know,
nanosecond time range.

Okay. I I do remember what I was gonna ask
about, is because you're using ait's all lasers on the system
that you're on. How does a lasercool an ion or anything? Because
it's it's it's energy. How'sadding energy to a system enable
it to cool down or lose energy?

So I'm gonna piss off I'm gonna piss off all the
physicists because I know thatI'm gonna completely botch this
from accuracy. But think of itthis way. The idea is that
you're removing energy from thethe electron or not from the
electron, electron. But you'reremoving energy from the the ion
itself. Right?
Now, there's a lot of ways thatenergy can be re related, but

(20:39):
part of that is, like, the speedof rotation, part of that's,
like, how high up in the energyshell it is. And, essentially,
the idea is, like, using thelaser, you can time where and
when it's going to hit, and youcan do your modulations based,
like, around the expectedrotational speed of the the the

(21:00):
electrons around the in in theatom. So if you time it right,
it's equivalent to breaking.Right? Just like you think of
acceleration in a car.
Right? Like, acceleration can beeither actually increasing speed
or slowing speed. Right? You'reyou're doing so by applying
forces in different ways. It'sthe same thing happening at the
quantum level with lasers.

(21:21):
It's just that you're actually,by hitting it with the photons,
you're you are imparting energy,but you can do so against the
natural procession of it, whichremoves energy from the system.

Okay. Okay. That makes sense.

You're giving it a nudge in the opposite
direction.

Yeah. And there's other things you can do, like,
you can take an ion next to itand actually cool the ion next
to it and have it'll actually,you know, draw down heat because
it's it's adjacent. So you canhave what they call sympathetic
cooling where, you know,neighboring ions are used to
like, one ion was actually usedto be the cubit and the next ion
over is just a coolant. And youhave

to and you can do that after you've synced all
the cubits together?

Yeah. You can do the cooling, you can do pretty
much whenever you want, and it'snot a particularly, dangerous
operation or or anything likethat. The the main thing is,
like, when you've establishedthat relative phase time, right,
where where you you begin yourshot, your shot runs until
either your error has increasedtoo much for the result the the

(22:33):
quantum information, and itdoesn't you know, it's basically
gone back to a neutral state.Right? And this is probably
where getting into a little bitmore the abstract model makes
sense, because the the physicsside gets really tricky at that
point.
But, like, from the the abstractmodel of quantum computing is,
imagine you have a sphere.Right? And and what folks talk

(22:55):
about in the the quantumcomputing world is is called the
Bloch sphere, and think of it asthe North Pole of the sphere is
a binary one and the South Poleis a binary 0. So, and it
doesn't really matter. It'sconvention.
You could flip it, you know,whatever. But the point is is

(23:16):
that a qubit can represent canbe in a state of any point on
the surface of the sphere. Thesphere is a unit sphere. Right?
And, essentially, what you'redoing with all the laser or the
gate operations is you're movingthe point around on the surface.
So you're picking some point andthen you're rotating it, you
know, around one of the axes ormultiple axes, and you're

(23:38):
picking a point on the surfaceof that sphere. When you do the
detection, what you're doing isan operation where the location
on the sphere determines theprobability of whether it will
be a 1 or a 0. Essentially, thecloser you are to one of the
poles, the higher theprobability that your

(23:59):
measurement will will be at thatpole or at the representation of
that pole. So all quantumcomputations are really working
on this basis of you can movearound in an essentially
infinite space around thesphere. Right?
It's kinda like analog computingor like an analog voltage. In in
theory, it's infinite. Right?The problem becomes noise and an

(24:22):
error that gets introduced. But,ultimately, you can move
anywhere on the sphere, and thenat the end when you measure it,
it collapses to a 1 or a 0.

With some probability. Right?

Right. Which also means that you don't just run
the computation once. Usually,you run the same circuit for,
like, a 100 shots and look at ahistogram of the outputs,
because you actually have toanalyze the the probability
over, at least some relevantnumber of shots to know where

(24:55):
the the majority fall to knowwhat where your competition
ended up.

Well, that's fascinating. I didn't even think
about that. So

so if if the 2 hemispheres represent 1 and 0
and then all all other points onthat sphere represent some kind
of probability towards 1 or theother. Does does the equator of
this sphere equal 50% no matterwhere you're at on this equator?

Pretty much. And there there's actually a special
gate that that kind of like, youstart with the the 2 poll
locations, the the one and thezero, they call them the basis
states. So there's, like, thezero basis state and the one
basis state. And you'll see inthe notation, there's there's a
gate, and I probably would getthe pronunciation wrong, but,

(25:44):
there's a specific gate thatactually moves you halfway
between. Right?
It puts you at the equator. Andthat's because it is incredibly
useful to be at this point whereyou have a 5050 probability
distribution to start from. Sothat's often what happens at,
like, state initialization. Soyou establish that, reference
point where you start yourrelative phase measurement. And
the first couple of things youdo is actually run that that

(26:05):
gate to set all of your qubitsat the 50%.
Right? Put them all at theequator. Then you're running the
operations that are maybe thatwould ultimately be data
dependent. So you have inputdata where you're, like, you
know, I'm gonna run the sequenceof gates and and some of the
gates actually can be, you know,single qubit operations, like,

(26:27):
what and they call them 1 qgates, where it's just like,
this rotates you, halfwayaround, the sphere on the z
axis. Right?
That's an r z gate. But then youcan also have 2 queue gates

(26:47):
where it's essentially you willdo the gate, but it's
conditional on the state ofanother gate. But it's not it's
it's kinda like well, the gatethat we talk about is the
conditional NOT or the CNOTgate, And it is kind of like a
not, in that it essentiallyflips the probability. It

(27:09):
inverts where you are on the onthe on the sphere. The
conditional aspect though, it'sit's not like that immediately
takes effect.
Right? It's not inherently apoint in time operation that
persists where they becomeentangled, and future changes on

(27:31):
the control qubit will actuallyadjust what happens to, whether
or not it gets inverted.

Why? So

And this is about the point where I start walking
away from it because it getsinto algorithm development and I
don't understand how you turnthis into useful operation.

It's funny because I was gonna ask
something very similar to that.It's like, you know, to to also
potentially sound somewhatignorant, what what's going
through my head is, so what?Like, we can we can make
something possibly be 1 of 2states. How is that useful?

There I mean, this is a thing where there there is
a lot of research that's beengoing into this for decades on
how to turn these primitivesinto useful operations. And
there are useful algorithms.They often fall into the world
of, like, eigenvector valcalculations or, you know, and

(28:25):
and they often end up beingabout looking at energy level
states in physics simulations orchemical simulations. So there
are certain like, one of the toplevel things of quantum
computing is it is not a generalpurpose computation device like
we think of a computer. There'sa lot of folks who who talk
about quantum computers gonna bea revolution.

(28:46):
Everything's gonna be a quantumcomputer. A a more conservative
outlook is a quantum computer isreally an accelerator like a
GPU. It's really good atspecific types of operations
that classical computers are notparticularly good at.

But my question is, how so you're talking about
all these gates. How are thegates implemented in this kind
of computer? Because you've gota bunch of lasers. How's that
play?

Well, so that's that's exactly it. Right?
There's, there's a comp acompilation stage where we take
that list of gates, and eachgate turns into a series of
modulated laser pulses. So eachgate has a sequence where it you
know, on a on a 2q gate, it'sactually gonna be a set of

(29:32):
lasers because, at least in oursystem, like, one of the one of
the key differentiators in a lotof systems is the connectivity
between cubits. So in a lot ofsuperconducting systems, because
of the way they're constructed,you have, like, a 4 corner
system.
Right? You have a a grid, and soyou can talk like, each qubit

(29:52):
can interact with its adjacentneighbors in any of the four
directions. But that's it.Right? If you can't skip over 3
and do that.
So you have to optimize your youhave your compiler and your
optimizers have to rearrangeyour algorithm to work through,
I can only talk to my nearestneighbor. In the trapped ion

(30:13):
laser or, like, Raman laserapproach has all to all
connectivity within within thetrap. So, essentially, you can
steer the lasers to aim at anyparticular 2 ions. And so part
of the laser controls isactually the steering
mechanisms. Right?
Like, part of these AWG channelsare actually controlling
steering of the lasers. Then youalso have to have background

(30:33):
lasers that are just, sort of,constantly on across all of the
ions. And then you have theactual laser pulses that are
going to those 2 target ions forthe 2q gate, and each one and
those are modulating the actualeffect that you want. The
details of exactly what thatpulse looks like gets into a

(30:54):
level of the physics that I justdon't really understand.

Because I'm I mean, that this is what's going
through my brain is, like,because you were talking about,
like, the knot operation. Mhmm.I'm assuming you don't you your
when your laser does what itneeds to do, I'm not going to
pretend to know what that is,Hits it with some kind of
modulated waveform, or energypulse, but it doesn't it's not

(31:21):
like a inverter gate that wehave in silicon where a one
becomes a 0, a 0 becomes a 1,it's more of a I'm just going to
make sure I'm going to hit it ina way that I know with
probability it will flip itsstate, and you don't really know
what that is until the end.Right? Well, so in the case

where you think of a a a digital not gate, you're
having or a digital inverter,you're you're flipping the state
from 0 to 1 or 1 to 0. But ifyou think of that as, like, a
vector operation, right, you'redoing an in you're you're taking
if you have a vector pointingfrom 0 to 1 and you invert it,
you go to negative one. Right?Well, that's closer to what's

(32:02):
happening because you're dealingin a complex two dimensional
space. So the the point on thesphere that you have when you'd
run the the not gate or the cnot gate, when it actually
indicates that it should flip,you're actually going to the,
like, polar opposite location.
Right? So you you would, like,figure out the point on the
surface that is directlyopposite on and that's where

(32:24):
you're gonna try rotate to. And,really, the laser pulses are not
literally rotating electronsaround. What they're doing like,
the the point on the sphererepresents some notion of the
rate of spin or the particularcharge or, you know, some other
physical property, and that'swhat the lasers are actually

(32:46):
modifying. They're eitherintroducing energy or they're
adjusting, like, the rate ofspin or, you know, a variety of
things.
And and it depends on the systemthat you're using and the exact
configuration as to what exactlyyou're controlling, and that's
where every system is differentand every, you know, every
company that's making a quantumcomputer is taking a different
approach.

Because you don't have to use the same parameter.
Right? There's multiple Right.

And there's a totally different trade offs for
each one of

them. So so I I wanna rewind, for a second.
There there something came up inmy mind. The so so you mentioned
earlier that the the actualapparatus or mechanism that that
you use to trap these ions. Youhave a material.
You ablate it. It basically youcreate, like, a cloud, right, of
of ions. And then you captureone of those ions and move it

(33:36):
over to the trap. Right? OrMhmm.
Correct me if I'm saying any ofthis wrong. But how do you know
you actually got one? It's notlike you can lift the lid and be
like, oh, yep. It's right there.I see a cubit.
Like, what is the indicator orwhat's the actual mechanism for
you saying, yep. We got one?

So remember that we the detection that we do is
we cause the ion to fluoresce.So what you actually end up
doing is putting a camera withan optical path into the trap,
and you actually run a lasersequence that causes all the
ions to fluoresce. And you'rethen, essentially, using the

(34:16):
camera as a photon counter, andyou're looking to see whether or
not you got counts.

So are you is there just one ion on there? Or
is it you you grabbed a chunk ofthem, some cloud of them? So you

load them individually. Right? You do
capture 1 at a time, but you canthink of the the trap as, at
least the traps that we workwith are are linear trap. Right?
So it's essentially you'reconstraining the ion in 2 out of
3 dimensions or 2 out of 3 axes,and the the 3rd axis is is the
one that you can move it.

(34:50):
You can move it along that axis.And then along that axis,
there's a multiple zones, andone of those zones is the
loading zone, and that's whereit actually has the hole for the
the plume to come from theablation area up into the trap
and where you do the initialcapture. But then you have a
separate area that's a littlebit cleaner, where you actually
move it along into that area andyou line them all up in a row.

(35:14):
And so you can hold, like, howmany you can actually load in
there becomes a function of, thesize of your trap, the number of
electrodes that you actuallyhave in that space to for how
fine a resolution you canactually hold the positions, and
then also how well you candetect the fluorescence. So in
some systems, you might use afiber optic array where you

(35:37):
actually have a fixed number offiber endpoints and they're
pointed at very specificlocations along the trap.
That's gonna set the maximumnumber that you can actually
effectively use. You couldprobably load more, You just
can't see them. But if you use acamera, you can see everything
more, but there's trade again,it comes back to trade

So in in the trap area, you you're not just
capturing one ion here. Right?Is it is it is it an array of
ions? Basically, are you doing abunch of different cups that
hold them? Or is it literallyone trap?

Well, so the loading area you catch one ion
at a time from the ablation Andthen you move it along and you
you merge it with the line ofthem. And so you build up a a
line along the length of thetrap. Okay. And so you like, the
current systems we have, we talkabout being, AQ and some number.

(36:37):
Qbits.
It's more of a a a top levelmeasurement of performance of
the system, but it's also, youknow, you can think of it as we
get 36 ions lined up. Right?There's probably more than 36.
I'm not gonna give you an exactnumber, but 36 of them are
usable as qubits in that space.

Is is this environment, in a in a vacuum?
Is it, is it is the entireenvironment cold or are you just
trying to reduce the temperatureof the ion itself?

So, again, this is where, like, different systems
do different approaches. So thetrapped ion is is nice and that
you really only have to keep thetrap at cryo temperatures. So
the actual cryo chamber is quitesmall. I would say, think
basketball sized. Mhmm.

(37:30):
Right? Roughly. In asuperconducting system, there's
you have to actually keep a lotmore of the system at cryo and
at vacuum, or sorry. They don'tnecessarily have vacuum, but
they have more cryo. So theyhave to keep it at a colder
temperature, and they have moreof it.
So they have to build a muchlarger they literally call it

(37:50):
the fridge. Mhmm. Right, andthat's what the chandelier hangs
inside of. But in a trapped ionsystem, it's a much smaller
package that does have to bekept at vacuum, relatively high
vacuum, and then also at at cryotemperatures. But not, like,
like, there's different gradesof cryo.
Right? Like, you it's, like, howcold are you really gonna get?

(38:10):
Are you talking 30 Kelvin or areyou talking 5 Kelvin? Like,
like, that's

That's fascinating. So how, how was
parallel computing done? Yousaid you had, what, 14, 15
different lasers, but they allhave different functions. Right?
Could and and and you have theability to steer some of these
lasers.
Are you interacting withmultiple qubits at the same time
or is it just one at a timesequential?

Well, it it depends on if you're running a 1
q gate or a 2 q gate. You know,in a 2 q gate, like, you
potentially could do more, like,you could do a 3 or 4 q gate,
but from there's been a lot ofresearch that shown you can just
like in digital computing, youcan build everything out of NAND
gates. Right? There's a subsetof gates that are necessary that

(38:57):
allow you to build all possiblequantum gates. So you don't have
to build a 3q or 4q gate.
It might be advantageous forperformance or something, but
you really only need cert acouple of 1q and 2q gates. And
so the systems are oftendesigned their native gate set,
like, the actual operations theycan do, is a set that allows

(39:20):
them to emulate what's called,like, the common gate set.
There's a specific name for itthat escapes me at the moment,
but there's like a and that andthat's what most people program
in is is this, like, common gateset, and then it gets translated
to the actual native system. Sojust like in a modern computer,
you have the instruction setarchitecture, which is what
everybody writes their programsin. But internally, in the

(39:40):
microarchitecture, it'ssomething completely different,
and there's a translation thathappens.

Oh, so

I could write an algorithm and give it to 2
companies with quantum computersand they could just translate it
to theirs and and give me aresult Yes. Back.

Okay. Yep. Exactly. Interesting. But it
does mean that you arecontrolling, you know, multiple
qubits simultaneously.
And the whole thing is that youwant to you utilize that quantum
entanglement, like, that's a keyaspect of writing quantum
algorithms. If you onlyinteracted with 1 qubit at a
time, you would only get resultsfrom 1 qubit behavior, and you

(40:15):
actually want those 2 queuegates for that entanglement
behavior. Yeah. And therethere's a really good website
that goes through a lot more ofhow the algorithms work. It's,
quantum dot country.
It's a little long, and itmight, you know but it does go
through, for example, how you dowhat they call, like, the

(40:38):
quantum search algorithm and andhow it works. It's not it it
takes a lot to of building up,you know, the the knowledge base
to get to the point where it itkinda makes sense, But it goes
through how this would befaster. Now in practice, real
quantum computers can't actuallyrun the quantum search algorithm
currently. And and that comesaround to, you know, some of the

(41:02):
constraints of, like, why are westill in the r and d phase? Why
aren't we just using these?
Like, the you know, and they'recomplicated, but we've been
doing all this research. And thething is it comes back to those
coherence times. Right? There's2 major parameters that quantum
computers get evaluated on. 1 isthe number of cubits that you
actually have.
Right? Because each cubitultimately turns into a 1 or 0

(41:24):
representation at the end. So itdefines how many how big your
data is that you're working on.Right? Like, how many bits wide
it is at the end.
The other parameter is how manygates can I operate? And that's
purely a function of how long Ihave before my error adds up and
I've lost coherence. And so,different algorithms require a

(41:46):
different number of gates. So ifI need to run an algorithm that
uses 8 qubits for a 102q gates,like, sure, that fits on most
systems. But something like asearch algorithm might need 4000
cubits and a 100,002qgates, andthat doesn't fit on any quantum
computer that exists today.

(42:07):
So there's a lot of theinteresting applications that
would be advantageous just aretoo big, either in terms of the
number of cubits required or thenumber the gate depth required
to actually operate on thephysical machines. Now the other
side of that coin, is, you know,when does that trade off happen?
Like, when when does this becomeuseful to run them on physical

(42:30):
quantum computers? And there'sthis whole aspect of you'll hear
folks talk about quantumsupremacy. I don't really like
the term, or quantum advantageis a slightly better term for
it.
But it's like, when will quantumcomputers be better at running
an algorithm than somethingelse? Right? Where a classical
computer can't actually achieveit in a reasonable amount of

(42:54):
time. That's what the benchmarkis. And so the problem is is
that as quantum computers getbigger and better, so do
classical computers.
And so you can simulate aquantum computer on a whole
crapload of GPUs. So GPUs keepgetting faster, and the quantum
computers keep getting faster,and so the point at which that

(43:16):
crossover point would happenkeeps moving. But we haven't
gotten to a point where itactually crosses over. Now, in
the nearer term, there's also athing where you might talk about
might hear some folks talkabout, commercial advantage,
which is it's not that you can'tsimulate it. It's just that it
becomes more cost effective toactually run it on a on a

(43:38):
physical quantum computer forspecific applications.
Right? It's not a universal,like, you can't possibly
simulate the quantum computeranymore with classical and
reasonable time frame. But forspecific applications, the
quantum computer will be faster,or more power efficient, or less
expensive.

I I got a I got 2 questions. The first one is
alright. So in in programming,this is the hello world. What's
the hello world for quantumcomputing quantum computer?
Because, you know, in inprogramming, it's print to the
console or print to serial, andthen you got blinking LEDs and,

(44:20):
you know, you know, embeddedembedded hardware.
What what's the hello world orblinky for Quantum? Like, how do
you know that you're, like, yes.It's working finally.

Well, okay. That that's that's interesting that
you put it that way becausethere's really 2 different
things. One is, is the machineworking? Which is which is very
different from is my algorithmcorrect?

That is true. Like, your your your your
investor stakeholder shows upand is like, I wanna see this
thing working. Like, how do youhow do you do

that? Right. And and a lot of that is actually
running things that are notnecessarily meaningful quantum
algorithms. Like, one of theways that quantum computers get
evaluated is by literallyrunning random circuits and
recording the performance of itin terms of what what they call
the fidelity. Right?
Like, how much error did youaccrue or or, actually, the the

(45:19):
inverse of the error rate? Howhow accurate were you, as well
as the number of cubits that youran and how many gates you were
able to run? There's advantagesand disadvantages to that
approach, but there a lot of thecalibration and, like, proving
that the machine is actuallyworking is running very specific
test patterns, you know, justlike you would do on any other

(45:40):
system, and validating that, forexample, the laser is pulsing at
the correct time and, you know,doing a and there's a lot of,
like, looking at scopes and, youknow, running plots of different
patterns and looking at the whatthe the actual fluorescence
rates are and all that kind ofstuff. When you get to the
algorithm stage though, you youessentially just think the
machine is working. Right?

(46:01):
For, like, people writingalgorithms, they don't they're
not particularly concerned aboutwhether or not the machine is
working. That is the machine opowner operator's problem. Now,
the Hello World, like, what doyou actually write? I'll be
honest, I've not written aquantum circuit. I look at them
because of the work I'm doing,but I'm more concerned with how

(46:22):
do I run them rather than how doI write them, which is kind of
an odd thing, but, you know, ithappens in the computing world
too, in the classical computingworld.
Now there is the the one thingthat's probably a bit surprising
to folks is that most quantumprograms are actually written in
Python. So a lot of theframeworks, like Qiskit is one

(46:45):
of the more popular ones.There's also, CUDA Quantum
Quantum. Yeah. CUDA Quantum.
There's a couple of others. Theythey're Python frameworks that
let you write out, like, here'sthe circuit that I wanna run. I
I need this many cubits, andthen I wanna do this. And then
they have the plug ins for thedifferent back ends for

(47:05):
different, computers. So itit's, in a lot of ways, like
writing a GPU or program.
Right? You're writing it in sortof a neutral language, and then
you have these back ends thatactually compile it down to the
specific machines that it'sgonna run on. But a lot of the
examples are, like, just draw upsomething that doesn't

(47:28):
necessarily have to bemeaningful, but that you can
actually figure out what theoutput should be. You know, so,
maybe, it's allocate 2 qubits,run the thing that puts them at
the 50%, and then run a rotationthat should force them to be,
you know, 1 to be a 1 and theother one to be a 0. And you run
that and you just see, does itactually come out to be a 1 or a
0?
Right? And then you get to lookat the probability distribution

(47:50):
of how often did that actuallyoccur, and that tells you a
little bit about the accuracy ofthe machine that running on. A
simulation might show that asperfect, but a real machine is
gonna have error in it.

You know, errors brings up one of the questions
that I I had. You you mentionedthat, the the coherence time is
that correct? Coherence time?

Mhmm.

It is or one of the ways you identify that
you're nearing the end of yourcoherence time is is an increase
in error. But how do you knowwhat your error is? If how do
you know that the system isbecoming or or there are more
errors in the system? Is it justthat the histogram starts to
smear out more and andeverything looks more random?

There's definitely that is one of approach, and and
it is one of the, like, one ofthe key ways of doing it is that
you actually look at thehistogram output of an algorithm
that you know should behave in aspecific way for these inputs,
and you see whether or not thesharpness of that histogram.
Right? Is it actually startingto to distribute the probability
further out? There are othertechniques that are more based

(48:57):
around trying to isolate wherethe error is coming from,
because the the you know,looking at just the histogram
output is telling you about thefidelity. Mhmm.
It's telling you how accuratelyam I getting to the the result.
But the source of the error cancome from many different places.
And so that's where you starthaving different tools where you

(49:19):
might run specific algorithmsand look at the output, and it's
not meaningful in terms of whatthe result would be, but where
the distribution of that, thatit clusters around might tell
you a source of error. Or youmight look at actual output
timing data from the controlsystem and see, for example, oh,

(49:42):
I was off by one clock cycle,and that actually causes all of
my phase estimations to bewrong. So, therefore, I, you
know, incurred extra phasethere.
Or, you know, my phaseestimation uses a fixed point
number with which is a finiteprecision. So that's inherently
gonna have some noisecharacteristics that come in
just by virtue of, havingreduced precision.

So so not only are you looking at the system
itself, the, the output of thealgorithm or the output of the
cubits, you're also monitoringall of your inputs to it. And if
any of that gets off, thenthere's error.

Yeah. You could think of it like in a modern
computer system, there's a wholebunch background processes that
are running, like, ECC isrunning. Right? Like, you you
have things that are telling youif it's misbehaving or, like,
your your modern PHYs areconstantly looking at signal
integrity and adjustingparameters. In a quantum
computer, it's so large scalethat there aren't a whole lot of

(50:41):
those automated systemshappening in the background all
the time.
Instead, what you do is you stoprunning customer workloads and
you go and run a bunch ofprograms and and calibration
sequences and, basically, haveautomated systems that look at
the results of that and thenmake a decision about what to
do. And maybe that means, like,oh, it's time for me to to dump
this ion chain and reload. Ormaybe I need to actually, you

(51:05):
know, run this analysis of it,and I come back with corrections
for some of the other printlike, calibration parameters.
And so you run those calibrationsequences periodically, and then
start running jobs again. Right?
So there's, like, periodicchecks of is the machine still
in good state and returning thefidelity that we guarantee.

Is is a lot of that found empirically? Like,
how often you have torecalibrate?

It's, I I would definitely characterize it as
there's a lot of research andsimulation that happens when
designing and in the earlystages of of getting the machine
commissioned. But once it goesinto actual operation, it's much
more empirical. Mhmm. Right?Like, going through the
commissioning phase into the theactual operation, it's, like,

(51:50):
there's a point where thesimulations can only take you so
far because each machine has itsown its own quirks.
Right? Its own littletolerances. And, like, if you
take a machine down to swap theablation target out. Right?
Because that's I mean, you'reliterally ablating a material.
Eventually, you wear it out. Soif I replace that, well, I had

(52:12):
to potentially warm the machineup, come back up to to
atmospheric pressure to be ableto change that, and then I have
to bring it back down to vacuumand back down to temperature,
things shift. So now you have torun all the calibration
sequences again. So it's, like,it's very time dependent. It's,
you know, so you you can only doso much of where you expect it

(52:34):
to be, but you have, likesimulation gets you into a
here's the bounds of, like, whatthe operating area should be.
And then then you have to spenda lot of time running
calibration scripts to find out,where does this particular
machine at this moment actuallybehave?

So my my second question is, so electrical
engineering is already like,when when you're in school and
you're talking to a bunch ofother engineers and everyone
keeps saying, like, electricalengineering is, like, the
hardest. And the main reason whypeople say that it's because
you're dealing with electrons alot of times and you don't
really see this is the hardwareelectronics, I guess, But you

(53:13):
don't really see what's reallygoing on. Right? This is even
further down that hole of nowyou can't even really see what
you're working on, especiallynow, Rick, that you worked,
before the podcast we weretalking, he works fully remote.
I can't even imagine trying towork on something like this,
fully remote and also not evenkinda hard for me to say this, I

(53:38):
guess.

Well, I I mean, you're right. It it is very
abstract. Right? Like, even if Iwanted to know what was actually
happening inside the machine, Ican't. Like, that's that's
fundamentally impossible.
Right?

There's no it's not like a register where you
can go, okay, if you put theseinputs into this ALU, it's gonna
do this thing with the bits.

Yeah. Debugging gets really hard.

Yep. Once once you, Lisa embedded, like, once
you figure out a debugger, like,those those exist, then that
this that process of writinglike assembly code and stuff
starts making more sense.Whereas this there's not that
yet or maybe not impossible withQuantum.

I mean, you can do a lot of work at the machine
level. Like, for for the peopledesigning the machines, there's
a lot of things we can look at,because a lot of it is actually
classical systems that are thatare operating the machine. And
we often separate the machineinto, the the sort of physics
part of it, the the actual trap,cryo chambers, all that kind of

(54:45):
stuff, where the actual quantumoperations are happening. And
you can't really know what'shappening inside that space
without like, you have to runliteral physics experiments to
figure out what's happening,where all of the control systems
in front of it is, like, all thetest equipment that they're
actually using for it to runthose experiments. And those, of

(55:07):
course, we can apply all thestandard techniques to.
Right? I can hook up a scope anda logic analyzer, and I can use,
you know, integrated logicanalyzers and FPGAs, and I can
run regression tests on softwareand all sorts of things. But
you're right. Like, when I'mdeveloping an algorithm, a
quantum algorithm, or I'mworking through, tracking down
error in a quantum computer,there is a certain point where

(55:29):
you're just getting down to, howdo I run these experiments where
I can't actually see what'shappening at the moment? I can't
stop midway through and inspectit.
I can change the algorithm toend at that point, but that
won't necessarily tell me what'sgoing on. So you end up having
to construct these differentscenarios to kind of induce
behavior that you wanna see andthen think about it and come

(55:53):
back with this analysis of,well, when I did this and I did
this and I did this, then thesewere the behaviors, and that
means that it probably is this.And then we go run another
experiment to test thathypothesis and figure out
whether we actually found thereal real cause or not.

I can totally imagine a boss, you know,
something's going in incorrector or or wrong and a boss comes
up and says, I need to knowexactly what's going wrong and
it and you're like, it'sphysically impossible for me to
tell you what's going wronghere.

I I actually have some stickers on my work laptop
that I got from a a friend ofmine. And I have one that says,
uncertainty, do not attempt tomeasure. Right. Like Love it. So

this this is a this is a statement from you
when we were we were talkingabout this podcast, and Steven
put this in here and I love thislike just touch on it a little
bit and it's a quantum computersaren't useful for anything
except R and D today and thatlikely won't change for at least
a few years. What needs tochange?

Well, that comes back to that that commercial
advantage thing. It's, you know,on one hand, you can look at it
as as, like, the quantumadvantage, where you actually
get the machines good enough interms of number of qubits that
are there, with sufficientconnectivity between those
cubits and good enough fidelity,which also means good enough,

(57:22):
you know, fidelity after somenumber of gates. Right? Usually,
it's, like, your your fidelityis a target number and you're,
like, I'm 99.6% fidelity afterso many gates. Right?
And, like, quantum advantage iswhen you get those 2 parameters
to a point where the time to runthe same algorithm on GPUs would

(57:47):
take much longer than running iton a quantum computer. We're a
long ways off from that. Like,scaling up these systems, you'll
see again, this comes into thetrade offs of the different
systems. Like, superconductingsystems, it's actually fairly
easy to scale up the number ofcubits. And so if you look at
IBM's systems, though, you'llsee the cubit numbers keep

(58:07):
climbing.
And that's because, like,they're literally just making
wafers. I I I mean, I say justmaking wafers, like, that's an
easy thing to do. You know, it'sit's complicated, but it's also
it's a known simpler process.Right? It's it's using a a lot
of existing practices aroundsilicon production to create
those qubits.

(58:28):
But they have the problem of thenearest neighbor connectivity
and their their, fidelity timesare pretty limited. Right? They
they kinda have so you get endup with a a lot of not so good
qubits. So you end up having touse a lot of those qubits as
error mitigation. Right?
You might have to have Whereas,in a trapped ion system, you

(58:51):
have a lot of Whereas in atrapped ion system, you have a
lot fewer qubits, because it's alot harder to hold on to all
those ions, but the fidelity andthe coherence signs were a lot
longer, but it's also slower.Right? So there's just, like,
different trade offs in thespace around these different

(59:11):
approaches, but the commercialadvantage piece is really
looking at when do we get analgorithm that is advantageous.
It makes sense to buy thequantum computer to solve that
problem versus doing it all withGPUs. And that's a trade off of
not only how long does thealgorithm take to run, but also
how much power it consumes to dothat and, the actual cost of the

(59:37):
machine.
Because one way of scaling upsome of these problems, if you
have a problem that scales to,like, n log n cubits, for the
size of the problem input, Well,like, that that's growing, you
know, more than linear. Your youcan scale that out with GPUs by

(59:59):
just keep adding more moreservers and more GPUs. But your
power increase per system thatyou add is pretty large. And so,
eventually, you get to a pointwhere you're, like, the amount
number of servers that I wouldhave to buy, I'm buying, like,
whole data centers worth ofmachines in order to actually
run my computation. Whereas, ifI can fit that in a quantum

(01:00:20):
computer, the quantum computermay cost orders of magnitude
less and be about the sameamount of time to do the
computation.
So it's this is something thewhole industry is trying to
figure out is, what are thosealgorithms, How many cubits do
you need? What fidelity do youneed? How do you get your system
to do that? And the thing isthat the trade off space between

(01:00:41):
all these different approachesmeans the algorithm that's gonna
be commercially advantageous foreach approach might be
different. Because a system witha lot of a lot of noisy cubits
might be better for somealgorithms than one that has
super high fidelity but fewcubits.
I think those are, you knowMhmm. Being a two dimensional

(01:01:03):
space, it's there's a lot ofplace to play. But even in those
cases, again, it's a it's a racewith GPUs keep getting faster.
So it's just gonna be, you know,no one has a good answer for
this right now. There's somecandidate, algorithms that seem
like they would be viable whenyou get out far enough, but you

(01:01:25):
also then look at, well, that's4 or 5 times the number of
qubits that we currently have.
How easy is it gonna be for meto get there?

Stephen Kraig (01:01:36):
So in your opinion, you think we're pretty
far off from quantum computingbeing in our everyday life?

Rick Altherr (01:01:43):
Oh, I don't know how much we will ever get to
everyday life. Like I said, it'skinda like g if you think of a
quantum computer as more like aquantum accelerator. Right? It's
similar to a GPU and then it'sonly useful for certain types of
problems. How often are thoseproblems gonna show up in your
day to day life?
For everybody, I kinda doubtwe're gonna be doing, you know,

(01:02:06):
physics simulations constantly.And and I don't mean, like, the
the video game physicssimulations. Right? Like

Parker Dillmann (01:02:15):
I want my bridge simulator to be perfect.
Right. It's so incredible. Spaceprogram running with physics
simulations on a quantumcomputer?

We gotta make doom run on the quantum computer.
Right? That's the whole

That's that's the first step. Right?

Right. But the like, the question about when
does it get to everybody needs aquantum computer, part of the
problem is, is there a problemthat everyone faces that would
even be in where are quantumcomputers even relevant? And we
don't know. We just don't know.

Pretty early on in the r and d phase, it seems
like. And and and it seems likeseems like this technology is
moving quickly but in the grandscheme of things on like a
consumer level, quite slow.

Well, with any sort of technology development
like this, I mean, if you goback and we we look at the
development of digital computersthrough the the lens of history.
Right? And if you go back tothat time, did anyone know that
a stored program computer wasgonna be possible? And, like,
how fast was it gonna take youknow, were they gonna be able to
develop such a thing? It's thatsame level of uncertainty.

(01:03:32):
There's gonna be some series ofbreakthroughs and developments
that make it, you know, not onlypossible to build the machine,
but that it actually becomescommercially viable? How long
did it take to go from thosevery early systems to buying one
to put in your business to run,

you know Point of sales.

Yeah. Well, I mean, even before that, right,
where they were, like,specialist applications that
needed that computation power.Right? Like, it went through
phases, and there's no reason toto expect that it'll be any
different for quantum computing.But it also is we can't expect
that the time of development isgonna be the same either,

(01:04:13):
because it relies uponfundamental science
breakthroughs.

Yeah. It seems like there's a lot of, a lot of
the environmental issues wouldhave to be resolved before it
becomes something that is ineveryday use. In other words,
people are not gonna have avacuum tank that gets to cryo
temperatures just to havequantum computing in their
kitchen.

Right? Right. It's not gonna be on a wristwatch.

Right.

You know, it's like but, we are moving to the
point where there are multiplequantum computer manufacturers
who are selling the machinesthat are able to be deployed in
slightly modified data centerlocations. Right? Like, it's
it's getting to a point wherethey're more self contained

(01:05:01):
systems. Certainly, we'retalking, like, 6, 7, 8, 19 inch
racks. Right?
Like, these are not smallsystems. But if you again, go
back to classical computing whenhe was, like, how big was
UNIVAC? Right? So there'sdefinitely gonna be a question
of how fast does this progressin terms of building up scaling
up the capability of the systemand also the the miniaturization

(01:05:22):
of the system. And we just don'tknow.
I mean, has there ever been areal demand for microscopic
cryogenic systems? I I can'tthink of 1.

Yeah. There there is a there is a manufacturer
here in town in Denver that,that that makes them and they're
they're doing the fridge style,and I think their smallest one
is is, you know, generally thesize of a of a small car. Yeah.
So we're at that stage.

Well, in in a couple of years ago for a
trapped ion system, it wouldhave been well, first, you start
with 2, 4 by 10 foot opticalbenches. Right? And then your
cryo system, and then and soyeah. And these already have
been making a lot of strides interms of miniaturization, but

(01:06:13):
there's a long way to go.

Do we wanna wrap up? Yeah.

I think so. This has been really, really
fascinating, and I'm sure wecould go for quite a while
longer but we really appreciateyou coming on, Rick.

Yeah, it's been my pleasure. So so

Rick, where can our listeners get in touch you
if you, if they want to talkabout quantum computers?

So I am on mastodon@mxshift@social.treehouse
dot system, I think. Yeah. It'sdot systems. That's, where I'm
usually at on the social media.I also actually offer, free

(01:07:02):
mentoring and resume review andand other things.
And so I have a a Calendly,calendly.com/mxshift, where you
can sign up for a slot for mockinterviews or have me look
through your resume and givefeedback and those kinds of
things. Very cool.

Awesome. We'll put that in the show notes. So
thank you so much Rick for,talking about more of the
quantum computing actually likephysically works versus, I guess
we did talk about some theory,but not too much.

Well, I mean, you wanna know about the electronic
side. Right? That's the wholepoint.

Yeah. Yeah. I I'm just so fascinated that it's
like another level down the Ican't see stuff, hole. So, so
thank you so much everyone forlistening to circuit break from
MacroFab. We are your hostsParker Dolman.

And Stephen Craig. Thank you so much Rick.

You're quite welcome.

Alright. I'm gonna do the outro. Breaker for
downloading our podcast. Tellyour friends and coworkers about
circuit break podcast fromMacroFab. If you have a cool
idea, project, or topic, or youwant to talk about quantum
computers, because apparentlythat's what we do now, let
Steven and I and the communityof Breakers know and then talk
to Rick as well.

(01:08:25):
Our community where you can findpersonal projects, discussions
about the podcast, andengineering topics and news is
located atform.macfab.com.

All Episodes

EP#439: The Nuts and Bolts of Quantum Computing - Rick Altherr

Episode Transcript

Popular Podcasts

On Purpose with Jay Shetty

Stuff You Should Know

Dateline NBC

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}EP#439: The Nuts and Bolts of Quantum Computing - Rick Altherr