November 4, 2024 • 23 mins
Ray Kurzweil recently released his new book, *The Singularity is Nearer: When We Merge with AI*, a follow-up to his 2006 blockbuster, *The Singularity is Near: When Humans Transcend Biology*. I had the chance to see him speak at our local bookstore as part of the launch event.
A central theme of Ray’s talks is the exponential growth of computing power—how we’re constantly achieving more computations per second at a constant cost. While exponential curves are easy to discuss and graph, understanding these transformations on a visceral level is challenging. How can we fully grasp the scale of change that exponential growth reflects?
This episode tackles that question, showing the exponential advances in computing power at a fraction of the cost, using examples and everyday objects that everyone can relate to, regardless of technical background.
We’ve all heard that the smartphones in our pockets are far more powerful than the supercomputers of the past. But what does that really mean? How can we quantify, visualize, and bring this leap in technology to life in terms that feel real and relatable? We delve into comparisons—RAM, storage, and power usage—to illustrate the mind-blowing scale of progress and explore what these advances mean for the future of computing.
More importantly, we examine the impact of this rapid growth as we move forward along an ever-steepening exponential curve. Join me as I use everyday items to help us better grasp this extraordinary rate of change, connect with our technological past, and imagine the future as exponential progress continues to shape our lives in profound ways.
Exponential Curves: Supercomputer to Smartphone to Singularity | Turn the Lens podcast with Jeff Frick
#AI #GenAI #ExponentialGrowth #ComputingPower #Future #RayKurzweil #TheSingularity #MooresLaw #AIandFuture #SupercomputerToSmartphone #Change #RateOfChange #FutureOfTech #Rice #Chessboard #Curves #Understand #Visualize #TechEvolution #Exponential #5G #T1 #FLOPS #Bandwidth #Storage #RAM #Visualization #TechForEveryone #SuperComputer #SmartPhone #Singularity #ExponentialCurves #SmartphoneVsSupercomputer #Podcast #TurnTheLens
https://www.youtube.com/watch?v=meqrxgy435g&list=PLZURvMqWbYjk4hbmcR46tNDdXQlrVZgEn
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Hey, Jeff Frick herecoming to you from the future.
You know, I'mfinalizing the edits on this
latest episodeon exponential curves
and the fact that we havea supercomputer in our pocket
and trying to visualizeor create some comparisons
of things that are easier to grok,
so that we canreally understand
what's happening here.
And as I'm going through it,there's a lot of numbers,
(00:20):
there's a lot of math.
But ultimately I just wantedto get to some things
that we can kind of graspor visualize
to try to quantify better.
How much technologyhas changed.
And what's importantis really the concept
of how much, it'sgoing to change in the future.
So you'll heara lot of numbers.
Try not to get toowrapped up in the numbers.
(00:42):
I'll put all the mathin the notes
so you can goand run your own numbers.
But the conceptsto really focus on,
you know, we've gone from
a Cray supercomputer in 1973to our mobile phones
and our mobile phones are
thousands,if not tens of thousands
of timesmore powerful
for fractions of the cost,fractions of the weight,
(01:04):
fractions of the energy consumption.
So if you think aboutwhere we were at 1973
and where we are today, and
now you start to think aboutwhere are we going to be
two years from now, three years from now,
five years from now, ten years from now
on, these exponential curves,
it really begs the question,what will we be carrying
around in our pocketsor in our brains
(01:25):
or in our heads, or howeverwe're going to be connected
to this technology down the road,
you know, ChatGPT,
came onto the scene two Novembers ago.
They're alreadytalking about ChatGPT.
I think it's [ChatGPT] five
that’s coming out soonis a hundred times more powerful.
So it's two years, 100 times.
So 10x twice, to get here.
(01:47):
So I really
the most importantpart of this episode
is trying to visualize
how fast the stuff is moving
and try to think aboutwhere it's going to be going.
We didn't even get into
some of the interesting thingsthat Ray Kurzweil talks about
like mortality escape velocity,
digital twins when applied tobiology and people.
(02:08):
I'll put a bunch of his linksin the notes below as well,
so you can watch some ofhis interviews and videos,
because it’s a really interesting concept,and I think it's important
that we always kind ofthink of where we are today
as kind of where it's going to end,
and it's notat all.
You look backin any technology
and just a few years back,you know, what we were doing
(02:29):
seems almost juvenile sometimes.
And then you think aboutthe implications,
implicationsin something like,
medical technologywhere, you know,
artificial knees noware an outpatient procedure.
Artificial hips
are practicallyan outpatient procedure.
Stents and pacemakersand all these things
(02:50):
that we kind of take for granted
were amazing technological developmentsin the not too distant past.
So what does the future hold?
So keep that in mindas we go through
some of the numbersand some of the illustrations.
Don't get hung upon the math.
Run the math yourself.
Try to think of someinteresting ways to compare
what the power of computing
was 50 years ago,
(03:11):
compared to what the powerof computing is today,
and try to think aboutwhat is it going to be
five years from now,
10 years from now, 20 years from now,
50 years from now,
a very different future.
So with that,
I'll sendyou back to the past,
where we recordedthis episode,
which I've been tryingto get out for so, so long.
So I'm just going to hit publish,
tribute to my buddyKeith Townsend
(03:33):
and get it out there.
Hopefully you enjoy it.
Don't get too hung upin the numbers and the math.
Really try to focus on the illustrations
and if you got better ones,I'd love to hear it.
Thanks for listening.
Thanks for watching.
[clap, clap, clap]
OK
in three
two
one
Hey welcome back everybody.
(03:54):
Jeff Frick here
coming to youfrom the home studio
for another episode of‘Turn the Lens’
Going solo today
I got Ray Kurzweil herein kind of virtual spirit, virtual sense.
But he's not here in person,unfortunately.
But he recently came outwith his latest book
called ‘The Singularity Is Nearer’[2024 Viking]
And it's pretty wild.
And he's out doing the book tour.
I had the good fortune
(04:15):
of seeing him hereat one of our local bookstores.
If you get the chance,I would strongly suggest you do.
There's another great interview
where he talksabout a lot of the concepts.
If you don't have the,the chance to see him,
I'll put a link to thatin the description below.
But this whole conceptof exponential curves
is really hardfor us as humans to grok.
And I've know I'vementioned it before,
(04:36):
but I thought, you know,
let's go through some specificsand really see if we can,
give you some
tools to see
to make it more,
tangible I guess
digestible,recognizable,
of what exactly is going on here.
So Ray always shows this,
(04:59):
graph showing the scale
and the power of computation,
over time, looks like from 1935
to the present day.
And it's up into the right
and you say, ‘Wow’
looks like a pretty steady curve.
And the reason it doesn'treally express
the exponential-’ness’ clearly
(05:19):
Clearly is because it's price.
It's performance over price.
So this continues
to grow at a rapid rate.
And if you look to the leftand you see the scale,
you go from 0.0000001.
It's like five zeros and a oneas a fraction
to the top, which is not thousands,
(05:42):
not millions,not billions, but trillions.
So you go from,you know, a decimal place
with six placesto a trillion at the top.
So how do you even startto understand that?
What I wanted to dois just make
a real simple compareagainst a couple of devices
that most of ushave probably heard of,
(06:03):
if we're not very familiar with them.
And that is the Cray-1 supercomputer,
which came outin the late 70s,
and a modern smartphone,
like my Galaxy S24.
So we'll go throughsome of the basic
computing measurement parameters,
try to convert it into somethingthat's easier to grasp,
(06:24):
and really use thatto try to illustrate
how far things have gone.
Or, excuse me, how far things have come since the,
since the late 70s,which seems like a long time ago
50 some odd years ago,thereabouts.
But it's reallynot that long.
And what's importantis that the time going forward
were only on a, a steeper part of the curve.
(06:46):
So let's go throughsome of these,
some of these things.
So first off let's talk about
computing horsepower.
And that is measured
in something called ‘flops’ which is
floating point operations per second,which is doing the math.
The computer's doing the math.
And a Cray in the late 70s, a Cray-1
(07:07):
And I'll put all these numbersin the notes below.
Could do 160 Megaflops,
a modern smartphone
does 5 Teraflops
and a Teraflop is a million Megaflops.
So you go from 160 Megato 5 Teraflops.
(07:28):
So how much moreis that?
And it's more.
It's 31,000 times more.
31,000 times more between a Cray-1 supercomputer, and your phone.
So again, it's just like hard to even understand what that means.
So what I decided to do is
(07:49):
let's look at that in terms ofhow long would it take to compute
all the credit card transactions
that happen in the world today?[in a single day]
So right now,according to ChatGPT,
there's about 1.98, close to 2 billion
credit card transactionsthat happen around the globe
every day.
So let’s just call it 2 billionin round numbers.
(08:11):
So at 160 Megaflops,it would take approximately 3.5 hours
to process all those transactions.
So basically thinkold school batch processing.
All the credit cardsrun all day long.
You want to, run it throughthe machines that you can,
put all the money where it's supposed to go,
(08:32):
that would take about3 hours and 26 minutes.
With the state of the art machine
in the late 70s.
Using your modern smartphoneat 5 Teraflops,
that same calculationwould take.
How long do you think?
Again, the comp is3 hours and 26 minutes.
(08:54):
Modern smartphone.
It would take 0.39 seconds,
less than half a second
to do the equivalent work.
3.5 hoursto less than half a second.
That's how much smarteryour smartphone is
than the stateof the art computing.
Not that long ago.
So that's flops as a proxyfor computing power.
(09:17):
Let's look at working areaor RAM
And RAMI always like to think of
as kind of the tabletop of your computer.
How much stuff can you haveopen at one time?
You know, how much things can it be working on?
And the old, Cray-1 had4 MB of RAM, 4 megs.
(09:38):
Your modern smartphone again these are top end smartphones
can have 12 GB to 16 GB gigs,
so let's just go with 16 gigs.
So how much bigger is 16 GB gigs than 4 MB megs?
Well, it's 4,000x times.
4,000x times.
So if we think about thatin terms of workspace,
(10:00):
let's just say your modern desk.
And I just went and looked at a catalogfor a modern standing desk.
It's 72 in by 30 in
is about 15ft² square feetof working area,
which is a pretty good sized desk.
So if that represented 4 MB megs,
how much desk space would you need
(10:21):
to represent 16 GB gigs?[Gigabytes]
So compared to one deskon a Cray-1
how much does spaceis that equivalent to
in your modern smartphone at 16 GB gigs?
And it's basically,
a football field.
It's like 6,000 ft² square feetcompared to 15 ft² square feet
(10:45):
or, excuse me, 60,000 ft² Square feet
compared to 15 ft² square feet
So it's basicallycomparing a desk
working areato a football field,
including the end zones,which come in at about
60,000 ft²square feet.
Huge increase 4,000x times.
(11:05):
Okay.
Let's talk about storage.
It turns out the Cray-1 didn’tactually have internal storage.
They use attached storage.
But according to Wikipedia
generally it would be between 80 Megs and 250 Meg.
So we're going to use 250 megs.
The high end.
And compare that to your modern smartphone
which has 1 TB Terabyte.
(11:27):
So so how do wethink about that
in terms of somethingthat we can visualize.
Because the raw number, it’s four
again 4,000x times,
which is hard to groksomething 4,000x times bigger.
So I thought, okay, well, let's think of it in terms of documents
and storage space.
So it figures outthat it
(11:48):
one two-file or excuse me,
two-drawer file cabinet, which is
I have lots in my garage still
used to have them at the office all the time.
So just a standard2-drawer file cabinet,
at a calculation of50 kilobits per page, in a PDF
can hold about 5,000 pages.
So let’s just call it 5,000 pagesfor quick and dirty math.
(12:09):
And that's about a footof shelf space.
Times that by 4,000x.
Okay.
That's about 4,400 ft
of bookshelf space
or 0.85 miles almost a mile.
So instead of a footworth of bookshelf space,
(12:30):
you now have a mile of bookshelf space.
That's the difference
between how muchcan be stored
on a 250 Meg drive,compared to a 1 Terabyte drive.
Okay, so stay with me.
Let's talk about powerconsumption.
The Cray-1 supercomputer
(12:50):
used 115 kW
115 kW
As a comparison,
the average homein the US today
uses about 30 kW per day.
30 kW per day.
So
it's four times more[per day]
(13:13):
than you use in your home [in a month]
Modern smartphone,it says, according to ChatGPT,
when it's doing typical tasks,
it uses between 2 Watts and 4 Watts,
not 115 kW,2 and 4 Watts
So when it's working hard,it could spike to 8 to 12 Watts
So it's literally
(13:34):
0.000008
times more efficient
to do remember, 4,000x more times more [computing] power
These are just crazy numbers.
And really hard to kind of
get your hands around, get your head around.
But then here's the kicker.
On top of it all.
(13:55):
The Cray-1 in 1977cost $7.9 million.
So that's approximately$40 million
in today's money,
$40 million
$40 million.
What do you payfor your smartphone?
For a nice top end
top end Samsung or Apple?
(14:16):
A $1,000, $1,200, maybe $1,500?
It’s crazy
for doing 4,000x times more work
than what would cost $40 million
based on where the Cray-1 was years ago.
So a $1,000 versus$40 Million
(14:38):
for 4,000x times the computational power
than what you had there.
And then justthe last little piece,
that we measure in computingis bandwidth.
And so back in the day,
in the late 70s,
state of the artwas what was called a T1,
which was about 1.4 mbps,
(14:59):
which all my friends in the business say
was the best performanceon its best day
compared to today.
Most of the phonestoday are all 5G,
and 5G has a range,but most of the the top end ones.
And again, we've been talkingabout the top end phones
can do 10 Gbps.
So how do you comparethe bandwidth, the throughput
(15:21):
of 1.4 Mbps to 10 Gbps in 5G
and the comparison?
And I really struggled with ChatGPTto find a good comp.
So finally got to the pointwhere I said
if a garden hose at 5/8” inches40 psi (pounds per square inch)
at a typical house
puts out about 17 gallonsper minute.
(15:43):
So if that represented a T1
state of the art back in the late 70s,
what type of hoses
would you need to represent
the throughput of 5G?
I tried to find like a single
big giant hose,like a construction,
not like a construction hose
but like a construction pipestruggled with lots of things.
(16:04):
So finally we got to.
Okay, what if we.
What if we compare it tofire hoses?
So fire hose is about 2.5” in diameter
and those go at about 100 psi.
And so a fire hose at 2.5” inches and 100 psi
puts out 250 gallons per minute.
So we're comparing
so that's 250(gallons per minute)
(16:25):
The garden hose has 17.(gallons per minute)
To get it up tothe size of a 5G.
It's 6,000x times more
6,000x times more.
So it would take 440 fire hoses
to be 5G
compared to one garden hose for T1
(16:50):
That is so much more throughput.
And then again, here's the kickerit's all divided by price.
Back in the 70s
it would cost you about $1,500 a month.
Today, if you account for inflation
at $1,500 a month, turns into about $7,800 a month.
(17:11):
So imagine if your phone billjust for your connectivity,
was $7,800 a month
for only 1/6,000(0.0001667) of the throughput.
So again,I had kind of ChatGPT help me
write like a little summarydocument.
It’s crazy.
(17:31):
So comparing your modernsmartphone to a Cray 1 Supercomputer
you get about 4,000x times the working areain terms of the RAM
you get 31,000x times more horsepower.in terms of the flops
you get 4,000x times more storage[in terms of the storage]
using .00001 the power,
(17:55):
and it costs youfor the unit
$1,000 compared to $40 million.
And for yourmonthly service fee
call it a $100 compared to, close to $8,000,
for the Cray-1.
So I don't know if that helpsillustrate what's happening.
And again, as I've said many, many times,
(18:16):
today's the slowest dayof technological change
in [the rest of]your life.
I think theChatGPT event
two years ago,two Novembers ago
is really kind of exponential curvesthrown in our face
that we cannow via our phone and 5G,
talk to a supercomputerthat lives in cloud,
that understands natural language
(18:37):
better than they ever have.
And will talk back to you
and answer in a human voice
and give you the, information.
And it's only getting fasterand faster every day.
So I hope this was helpful.
I don't know if it's helpful.
I struggle with this.
I think
humans ability to grok
(18:57):
exponential curvesis a big part of,
the challengethat we have in technology,
as I've talked aboutin other videos, you know,
there's kind of human biologydevelopment,
evolution, which is pretty slow.
There's been civilizations’evolution over the last,
you know, many thousandsof years, which has also been
(19:19):
relatively slow.
It wasn't that long agothat you probably,
you know, presumed in your careerthat you'd have the same career
that your folks had,
which was the samethat their folks had
which was the samethat their folks had,
which was farmingfor most people,
not that many generations ago.
And you would presumeyour kids would do the same.
That's not theit's not the case, but
(19:39):
even civilizations’ changespeed is nothing
compared to theexponential speed
that we're seeing in,in computing.
And again, another really illustrative story
is the tale of the,
of the kid and the chess [board]and the grain of rice [or wheat]
(20:00):
It's an old fable
where some kidbeat the king in chess.
And the king said‘How could I pay you back?’
And he said,
I just want,
one grain of rice,
on the first
for the first dayof your payback,
which is on the first squareof the chessboard,
(20:21):
the second day,I want you to double it.
So two grains of rice,
third day, double it.
Four grains of rice.
Fourth day double it.
Eight grainsof rice
for the 64 days.
For the 64 squares on the chessboard.
And if you do the math
and there's a lot of great videosout there talking about it
(20:43):
by the time you get to the 64th day,
you basically would be coveringthe entire country of India
deep in grainsof rice.
There’s not that much rice.
So exponential curves grow fast.
And that's just doubling that’s not
timesing it by ten. So,
it's a tough thing to, to grok.
(21:03):
And it's a good thingto think about.
Hopefully thisis interesting.
Maybe it's not interesting.
I'll throw some visuals upeven the visuals
trying to get ChatGPT to help meto do the visuals,
to do the comparisonbetween one garden hose flow
versus however many fire hoses I said
it can't do it, you know,to have a mile's worth of
(21:24):
of file cabinetsor bookshelf space.
These are really difficultconcepts to grasp,
and we're not really built for it,which is why
there's so many advancesin technology on so many fronts.
And they just, sometimes aresurprising how fast they move.
And as I've always said to,
you know, you only knowabout the technological advances
(21:45):
in the industriesin which you play,
but it's happening across
all industries concurrently,
and you just don'tsee it all the time
because it's not stuffthat you're that aware of. So,
interesting times ahead.
Today's the slowest dayof technological change
for the rest of our lives.
It's only goingto get faster.
Hope thiswas helpful.
Hope it was entertaining.
(22:05):
I've been wanting to do thisfor a long time
as Keith Townsend said one time when we were together,
you know, he had a post
that he struggledwith trying to get it down.
I've struggled with this.
I've struggled with trying
to get the concepts togetherthat are illustrative,
but I don't wantto miss the window
and not get the post up,which is
something Keith and I talked about,which is the worst of all worlds.
(22:26):
So hopefully you enjoyed it.
Hopefully it helped.
I think it's interesting.
Thanks againfor watching.
Thanks for tuning in on YouTube.
Thanks for listening inon the podcast.
Appreciate,appreciate you being there.
And reach out, say ‘Hi’
Signing off fromthe home studio.
Jeff Frick.See you next time. Take care.
Hey, Jeff Frick herecoming to you from the future.
You know, I'mfinalizing the edits on this
latest episodeon exponential curves
and the fact that we havea supercomputer in our pocket
and trying to visualizeor create some comparisons
of things that are easier to grok,
so that we canreally understand
what's happening here.
And as I'm going through it,there's a lot of numbers,
(00:20):
there's a lot of math.
But ultimately I just wantedto get to some things
that we can kind of graspor visualize
to try to quantify better.
How much technologyhas changed.
And what's importantis really the concept
of how much, it'sgoing to change in the future.
So you'll heara lot of numbers.
Try not to get toowrapped up in the numbers.
(00:42):
I'll put all the mathin the notes
so you can goand run your own numbers.
But the conceptsto really focus on,
you know, we've gone from
a Cray supercomputer in 1973to our mobile phones
and our mobile phones are
thousands,if not tens of thousands
of timesmore powerful
for fractions of the cost,fractions of the weight,
(01:04):
fractions of the energy consumption.
So if you think aboutwhere we were at 1973
and where we are today, and
now you start to think aboutwhere are we going to be
two years from now, three years from now,
five years from now, ten years from now
on, these exponential curves,
it really begs the question,what will we be carrying
around in our pocketsor in our brains
(01:25):
or in our heads, or howeverwe're going to be connected
to this technology down the road,
you know, ChatGPT,
came onto the scene two Novembers ago.
They're alreadytalking about ChatGPT.
I think it's [ChatGPT] five
that’s coming out soonis a hundred times more powerful.
So it's two years, 100 times.
So 10x twice, to get here.
(01:47):
So I really
the most importantpart of this episode
is trying to visualize
how fast the stuff is moving
and try to think aboutwhere it's going to be going.
We didn't even get into
some of the interesting thingsthat Ray Kurzweil talks about
like mortality escape velocity,
digital twins when applied tobiology and people.
(02:08):
I'll put a bunch of his linksin the notes below as well,
so you can watch some ofhis interviews and videos,
because it’s a really interesting concept,and I think it's important
that we always kind ofthink of where we are today
as kind of where it's going to end,
and it's notat all.
You look backin any technology
and just a few years back,you know, what we were doing
(02:29):
seems almost juvenile sometimes.
And then you think aboutthe implications,
implicationsin something like,
medical technologywhere, you know,
artificial knees noware an outpatient procedure.
Artificial hips
are practicallyan outpatient procedure.
Stents and pacemakersand all these things
(02:50):
that we kind of take for granted
were amazing technological developmentsin the not too distant past.
So what does the future hold?
So keep that in mindas we go through
some of the numbersand some of the illustrations.
Don't get hung upon the math.
Run the math yourself.
Try to think of someinteresting ways to compare
what the power of computing
was 50 years ago,
(03:11):
compared to what the powerof computing is today,
and try to think aboutwhat is it going to be
five years from now,
10 years from now, 20 years from now,
50 years from now,
a very different future.
So with that,
I'll sendyou back to the past,
where we recordedthis episode,
which I've been tryingto get out for so, so long.
So I'm just going to hit publish,
tribute to my buddyKeith Townsend
(03:33):
and get it out there.
Hopefully you enjoy it.
Don't get too hung upin the numbers and the math.
Really try to focus on the illustrations
and if you got better ones,I'd love to hear it.
Thanks for listening.
Thanks for watching.
[clap, clap, clap]
OK
in three
two
one
Hey welcome back everybody.
(03:54):
Jeff Frick here
coming to youfrom the home studio
for another episode of‘Turn the Lens’
Going solo today
I got Ray Kurzweil herein kind of virtual spirit, virtual sense.
But he's not here in person,unfortunately.
But he recently came outwith his latest book
called ‘The Singularity Is Nearer’[2024 Viking]
And it's pretty wild.
And he's out doing the book tour.
I had the good fortune
(04:15):
of seeing him hereat one of our local bookstores.
If you get the chance,I would strongly suggest you do.
There's another great interview
where he talksabout a lot of the concepts.
If you don't have the,the chance to see him,
I'll put a link to thatin the description below.
But this whole conceptof exponential curves
is really hardfor us as humans to grok.
And I've know I'vementioned it before,
(04:36):
but I thought, you know,
let's go through some specificsand really see if we can,
give you some
tools to see
to make it more,
tangible I guess
digestible,recognizable,
of what exactly is going on here.
So Ray always shows this,
(04:59):
graph showing the scale
and the power of computation,
over time, looks like from 1935
to the present day.
And it's up into the right
and you say, ‘Wow’
looks like a pretty steady curve.
And the reason it doesn'treally express
the exponential-’ness’ clearly
(05:19):
Clearly is because it's price.
It's performance over price.
So this continues
to grow at a rapid rate.
And if you look to the leftand you see the scale,
you go from 0.0000001.
It's like five zeros and a oneas a fraction
to the top, which is not thousands,
(05:42):
not millions,not billions, but trillions.
So you go from,you know, a decimal place
with six placesto a trillion at the top.
So how do you even startto understand that?
What I wanted to dois just make
a real simple compareagainst a couple of devices
that most of ushave probably heard of,
(06:03):
if we're not very familiar with them.
And that is the Cray-1 supercomputer,
which came outin the late 70s,
and a modern smartphone,
like my Galaxy S24.
So we'll go throughsome of the basic
computing measurement parameters,
try to convert it into somethingthat's easier to grasp,
(06:24):
and really use thatto try to illustrate
how far things have gone.
Or, excuse me, how far things have come since the,
since the late 70s,which seems like a long time ago
50 some odd years ago,thereabouts.
But it's reallynot that long.
And what's importantis that the time going forward
were only on a, a steeper part of the curve.
(06:46):
So let's go throughsome of these,
some of these things.
So first off let's talk about
computing horsepower.
And that is measured
in something called ‘flops’ which is
floating point operations per second,which is doing the math.
The computer's doing the math.
And a Cray in the late 70s, a Cray-1
(07:07):
And I'll put all these numbersin the notes below.
Could do 160 Megaflops,
a modern smartphone
does 5 Teraflops
and a Teraflop is a million Megaflops.
So you go from 160 Megato 5 Teraflops.
(07:28):
So how much moreis that?
And it's more.
It's 31,000 times more.
31,000 times more between a Cray-1 supercomputer, and your phone.
So again, it's just like hard to even understand what that means.
So what I decided to do is
(07:49):
let's look at that in terms ofhow long would it take to compute
all the credit card transactions
that happen in the world today?[in a single day]
So right now,according to ChatGPT,
there's about 1.98, close to 2 billion
credit card transactionsthat happen around the globe
every day.
So let’s just call it 2 billionin round numbers.
(08:11):
So at 160 Megaflops,it would take approximately 3.5 hours
to process all those transactions.
So basically thinkold school batch processing.
All the credit cardsrun all day long.
You want to, run it throughthe machines that you can,
put all the money where it's supposed to go,
(08:32):
that would take about3 hours and 26 minutes.
With the state of the art machine
in the late 70s.
Using your modern smartphoneat 5 Teraflops,
that same calculationwould take.
How long do you think?
Again, the comp is3 hours and 26 minutes.
(08:54):
Modern smartphone.
It would take 0.39 seconds,
less than half a second
to do the equivalent work.
3.5 hoursto less than half a second.
That's how much smarteryour smartphone is
than the stateof the art computing.
Not that long ago.
So that's flops as a proxyfor computing power.
(09:17):
Let's look at working areaor RAM
And RAMI always like to think of
as kind of the tabletop of your computer.
How much stuff can you haveopen at one time?
You know, how much things can it be working on?
And the old, Cray-1 had4 MB of RAM, 4 megs.
(09:38):
Your modern smartphone again these are top end smartphones
can have 12 GB to 16 GB gigs,
so let's just go with 16 gigs.
So how much bigger is 16 GB gigs than 4 MB megs?
Well, it's 4,000x times.
4,000x times.
So if we think about thatin terms of workspace,
(10:00):
let's just say your modern desk.
And I just went and looked at a catalogfor a modern standing desk.
It's 72 in by 30 in
is about 15ft² square feetof working area,
which is a pretty good sized desk.
So if that represented 4 MB megs,
how much desk space would you need
(10:21):
to represent 16 GB gigs?[Gigabytes]
So compared to one deskon a Cray-1
how much does spaceis that equivalent to
in your modern smartphone at 16 GB gigs?
And it's basically,
a football field.
It's like 6,000 ft² square feetcompared to 15 ft² square feet
(10:45):
or, excuse me, 60,000 ft² Square feet
compared to 15 ft² square feet
So it's basicallycomparing a desk
working areato a football field,
including the end zones,which come in at about
60,000 ft²square feet.
Huge increase 4,000x times.
(11:05):
Okay.
Let's talk about storage.
It turns out the Cray-1 didn’tactually have internal storage.
They use attached storage.
But according to Wikipedia
generally it would be between 80 Megs and 250 Meg.
So we're going to use 250 megs.
The high end.
And compare that to your modern smartphone
which has 1 TB Terabyte.
(11:27):
So so how do wethink about that
in terms of somethingthat we can visualize.
Because the raw number, it’s four
again 4,000x times,
which is hard to groksomething 4,000x times bigger.
So I thought, okay, well, let's think of it in terms of documents
and storage space.
So it figures outthat it
(11:48):
one two-file or excuse me,
two-drawer file cabinet, which is
I have lots in my garage still
used to have them at the office all the time.
So just a standard2-drawer file cabinet,
at a calculation of50 kilobits per page, in a PDF
can hold about 5,000 pages.
So let’s just call it 5,000 pagesfor quick and dirty math.
(12:09):
And that's about a footof shelf space.
Times that by 4,000x.
Okay.
That's about 4,400 ft
of bookshelf space
or 0.85 miles almost a mile.
So instead of a footworth of bookshelf space,
(12:30):
you now have a mile of bookshelf space.
That's the difference
between how muchcan be stored
on a 250 Meg drive,compared to a 1 Terabyte drive.
Okay, so stay with me.
Let's talk about powerconsumption.
The Cray-1 supercomputer
(12:50):
used 115 kW
115 kW
As a comparison,
the average homein the US today
uses about 30 kW per day.
30 kW per day.
So
it's four times more[per day]
(13:13):
than you use in your home [in a month]
Modern smartphone,it says, according to ChatGPT,
when it's doing typical tasks,
it uses between 2 Watts and 4 Watts,
not 115 kW,2 and 4 Watts
So when it's working hard,it could spike to 8 to 12 Watts
So it's literally
(13:34):
0.000008
times more efficient
to do remember, 4,000x more times more [computing] power
These are just crazy numbers.
And really hard to kind of
get your hands around, get your head around.
But then here's the kicker.
On top of it all.
(13:55):
The Cray-1 in 1977cost $7.9 million.
So that's approximately$40 million
in today's money,
$40 million
$40 million.
What do you payfor your smartphone?
For a nice top end
top end Samsung or Apple?
(14:16):
A $1,000, $1,200, maybe $1,500?
It’s crazy
for doing 4,000x times more work
than what would cost $40 million
based on where the Cray-1 was years ago.
So a $1,000 versus$40 Million
(14:38):
for 4,000x times the computational power
than what you had there.
And then justthe last little piece,
that we measure in computingis bandwidth.
And so back in the day,
in the late 70s,
state of the artwas what was called a T1,
which was about 1.4 mbps,
(14:59):
which all my friends in the business say
was the best performanceon its best day
compared to today.
Most of the phonestoday are all 5G,
and 5G has a range,but most of the the top end ones.
And again, we've been talkingabout the top end phones
can do 10 Gbps.
So how do you comparethe bandwidth, the throughput
(15:21):
of 1.4 Mbps to 10 Gbps in 5G
and the comparison?
And I really struggled with ChatGPTto find a good comp.
So finally got to the pointwhere I said
if a garden hose at 5/8” inches40 psi (pounds per square inch)
at a typical house
puts out about 17 gallonsper minute.
(15:43):
So if that represented a T1
state of the art back in the late 70s,
what type of hoses
would you need to represent
the throughput of 5G?
I tried to find like a single
big giant hose,like a construction,
not like a construction hose
but like a construction pipestruggled with lots of things.
(16:04):
So finally we got to.
Okay, what if we.
What if we compare it tofire hoses?
So fire hose is about 2.5” in diameter
and those go at about 100 psi.
And so a fire hose at 2.5” inches and 100 psi
puts out 250 gallons per minute.
So we're comparing
so that's 250(gallons per minute)
(16:25):
The garden hose has 17.(gallons per minute)
To get it up tothe size of a 5G.
It's 6,000x times more
6,000x times more.
So it would take 440 fire hoses
to be 5G
compared to one garden hose for T1
(16:50):
That is so much more throughput.
And then again, here's the kickerit's all divided by price.
Back in the 70s
it would cost you about $1,500 a month.
Today, if you account for inflation
at $1,500 a month, turns into about $7,800 a month.
(17:11):
So imagine if your phone billjust for your connectivity,
was $7,800 a month
for only 1/6,000(0.0001667) of the throughput.
So again,I had kind of ChatGPT help me
write like a little summarydocument.
It’s crazy.
(17:31):
So comparing your modernsmartphone to a Cray 1 Supercomputer
you get about 4,000x times the working areain terms of the RAM
you get 31,000x times more horsepower.in terms of the flops
you get 4,000x times more storage[in terms of the storage]
using .00001 the power,
(17:55):
and it costs youfor the unit
$1,000 compared to $40 million.
And for yourmonthly service fee
call it a $100 compared to, close to $8,000,
for the Cray-1.
So I don't know if that helpsillustrate what's happening.
And again, as I've said many, many times,
(18:16):
today's the slowest dayof technological change
in [the rest of]your life.
I think theChatGPT event
two years ago,two Novembers ago
is really kind of exponential curvesthrown in our face
that we cannow via our phone and 5G,
talk to a supercomputerthat lives in cloud,
that understands natural language
(18:37):
better than they ever have.
And will talk back to you
and answer in a human voice
and give you the, information.
And it's only getting fasterand faster every day.
So I hope this was helpful.
I don't know if it's helpful.
I struggle with this.
I think
humans ability to grok
(18:57):
exponential curvesis a big part of,
the challengethat we have in technology,
as I've talked aboutin other videos, you know,
there's kind of human biologydevelopment,
evolution, which is pretty slow.
There's been civilizations’evolution over the last,
you know, many thousandsof years, which has also been
(19:19):
relatively slow.
It wasn't that long agothat you probably,
you know, presumed in your careerthat you'd have the same career
that your folks had,
which was the samethat their folks had
which was the samethat their folks had,
which was farmingfor most people,
not that many generations ago.
And you would presumeyour kids would do the same.
That's not theit's not the case, but
(19:39):
even civilizations’ changespeed is nothing
compared to theexponential speed
that we're seeing in,in computing.
And again, another really illustrative story
is the tale of the,
of the kid and the chess [board]and the grain of rice [or wheat]
(20:00):
It's an old fable
where some kidbeat the king in chess.
And the king said‘How could I pay you back?’
And he said,
I just want,
one grain of rice,
on the first
for the first dayof your payback,
which is on the first squareof the chessboard,
(20:21):
the second day,I want you to double it.
So two grains of rice,
third day, double it.
Four grains of rice.
Fourth day double it.
Eight grainsof rice
for the 64 days.
For the 64 squares on the chessboard.
And if you do the math
and there's a lot of great videosout there talking about it
(20:43):
by the time you get to the 64th day,
you basically would be coveringthe entire country of India
deep in grainsof rice.
There’s not that much rice.
So exponential curves grow fast.
And that's just doubling that’s not
timesing it by ten. So,
it's a tough thing to, to grok.
(21:03):
And it's a good thingto think about.
Hopefully thisis interesting.
Maybe it's not interesting.
I'll throw some visuals upeven the visuals
trying to get ChatGPT to help meto do the visuals,
to do the comparisonbetween one garden hose flow
versus however many fire hoses I said
it can't do it, you know,to have a mile's worth of
(21:24):
of file cabinetsor bookshelf space.
These are really difficultconcepts to grasp,
and we're not really built for it,which is why
there's so many advancesin technology on so many fronts.
And they just, sometimes aresurprising how fast they move.
And as I've always said to,
you know, you only knowabout the technological advances
(21:45):
in the industriesin which you play,
but it's happening across
all industries concurrently,
and you just don'tsee it all the time
because it's not stuffthat you're that aware of. So,
interesting times ahead.
Today's the slowest dayof technological change
for the rest of our lives.
It's only goingto get faster.
Hope thiswas helpful.
Hope it was entertaining.
(22:05):
I've been wanting to do thisfor a long time
as Keith Townsend said one time when we were together,
you know, he had a post
that he struggledwith trying to get it down.
I've struggled with this.
I've struggled with trying
to get the concepts togetherthat are illustrative,
but I don't wantto miss the window
and not get the post up,which is
something Keith and I talked about,which is the worst of all worlds.
(22:26):
So hopefully you enjoyed it.
Hopefully it helped.
I think it's interesting.
Thanks againfor watching.
Thanks for tuning in on YouTube.
Thanks for listening inon the podcast.
Appreciate,appreciate you being there.
And reach out, say ‘Hi’
Signing off fromthe home studio.
Jeff Frick.See you next time. Take care.
Popular Podcasts
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.