May 29, 2026 • 50 mins
In this episode of the News Items podcast, Josh Tyrangiel joins John Ellis for a wide-ranging conversation about the promises, dangers, and strange human realities of artificial intelligence. Drawing from his new book, AI for Good, Tyrangiel argues that the most important breakthroughs in AI are not happening inside Silicon Valley boardrooms, but among stubborn doctors, educators, and public servants quietly trying to solve impossible problems. The discussion moves from Palantir Technologies and Operation Warp Speed to the modernization of the IRS, AI tutors in classrooms, and new tools helping doctors reduce burnout and improve patient care. Along the way, Tyrangiel explores the political future of AI, the collapse of public trust in institutions, and why the next generation of leaders may be judged less on ideology than on whether they can understand (and govern) the most powerful technology of the century.
Hosted by John Ellis
Listen
Watch
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
SPEAKER_00 (00:00):
Hello and welcome back to the News Items Podcast.
I'm John Ellis.
I'm the founder and editor oftwo Substack newsletters.
One is called News Items, theother is called Political News
Items.
You can find them both atnews-items.com.
My guest today is JoshTerringell.
Josh is a writer for TheAtlantic Monthly.
(00:20):
He was previously the editor ofBloomberg Business Week and
Chief Content Officer ofBloomberg Media.
A 12-time Emmy and PVDaward-winning producer.
He created Vice News Tonight onHBO and has produced numerous
feature-length documentaries forHBO, Netflix, and Apple TV.
(00:41):
If you haven't seen it, Josh'sdocumentary on the great
financial crisis is amust-must-watch.
SPEAKER_01 (00:50):
It's my pleasure, John.
SPEAKER_00 (00:51):
So a number of years ago, I was at a brain science
conference at MIT, and it wassort of a TED Talk-like setup
where somebody would speak for30 minutes and then the Q ⁇ A
and so on and so forth.
And then there was a breakoutsession, and we would all sit
around, eight or ten of us sitaround a table and talk about
things.
And one of the people at mytable was a man named Danny
(01:13):
Hillis.
And Danny is a founder ofThinking Machines Corporation,
is kind of a leading thinker onartificial intelligence.
And at the very beginning ofyour book, which is called AI
for Good, How Real People AreUsing Artificial Intelligence to
Fix Things That Matter, youasked Danny, what is AI actually
(01:33):
good for?
And his response was, try toimagine tech without the tech
companies.
What did he mean by that?
SPEAKER_01 (01:40):
Yeah, I mean, uh I'll rewind just a bit so that I
can get us to the question.
I'd been hired by the WashingtonPost right after ChatGPT came
out to start writing columnsabout AI.
And the reason they wanted acolumnist is they they just were
overwhelmed with stories andinformation and contradictory
information about AI.
And readers were angry andanxious and concerned.
(02:04):
So they hired me and basicallysaid, look, try and figure this
out.
Do whatever you gotta do.
And so for the first, you know,six, eight weeks of reporting, I
had this kind of glorious set ofmaterials, right?
Which were personalities likeSam Altman and Demis and Dario
and Elon.
I, you know, all of them had aninflux of billions and billions
(02:25):
of dollars that that were goingto these new machines.
So I had rivalries, I had money.
And then you would ask them,various people, not just the the
founders of the labs, hey, whatis this gonna do?
And the the answers almostunfailingly were either, oh,
this is gonna cure cancer, AI isgonna cure cancer, or AI is
going to bring about the end ofhuman existence.
(02:47):
And sometimes that was physical,and sometimes it was the reasons
to live, right?
Again, for your first set ofcolumns, great stuff.
But at a certain point, I waslike, all right, well, this is a
lot, and also not very much atall, right?
It's just such abstraction thatI was failing to understand what
we can do.
And so I turned to Danny, and Ithink you described Danny very
(03:08):
well.
He's he's very much like aBuddha of Silicon Valley.
He's been around a while, he'sunfazed by all the gold rushes,
and Danny just kind of laughedat me.
Uh, he was like, look, if youreally want to understand the
tech, you have to go beyond thetech companies.
And the reason for that istwofold.
One, you know, John, as you knowfrom journalism, you can't get
an honest answer out oftechnology companies, not
(03:30):
because they're so dissuasive,but because it's no longer
man-to-man, right?
When I go talk to open AI, I'mgonna just casually guess there
are 12 bodies on the open AIside for every reporter.
And so they uh they can create areality distortion field.
So that's one problem.
And the other is that even theselabs within a month of being
(03:52):
founded have very, very seriouscapital requirements.
And so they are trying to tellyou what they want you to use
the technology for.
And what Danny was suggesting islike, look, get outside the
labs, go to people who have realproblems in the world, you know,
like he was 40 years ago, seehow they're tinkering, and
you're gonna get better stories,let alone find out more about
(04:14):
what AI can really do.
And so it wasn't long afterDanny gave me that advice, I
would say weeks really, before Ikind of, you know, put my hands
in the soil and turned up thisvery radical AI counterculture
of people who are not seekingbillions of dollars, who are not
trying to get on the cover ofmagazines, who are just trying
to solve some problems.
(04:35):
And I found them reallyfascinating.
SPEAKER_00 (04:37):
Aaron Ross Powell, one thing about these people,
uh, if you read through thebook, they seem to be either
first-generation Americans orimmigrants, and they also seem
to be unbelievably stubborn.
They they will not take no foran answer.
Is that an accurate description?
Aaron Ross Powell, Jr.
SPEAKER_01 (04:55):
I mean, at a certain point, I kept I I realized I was
meeting a variation of the sameperson in every system.
And I was overwhelmed by thedegree to which this is an
immigrant story, that we havepeople who are coming into the
country, working in medicine andeducation and government, and
they want to fix stuff becausethe promise of America to them,
(05:15):
they don't come thinking thatthey're gonna get a finished
country, but they love thechallenge and they are stubborn.
The stubbornness I think hasless to do with their country of
origin than just the phenotypeof person who is willing to work
hard to solve a problem and workhard beyond a system.
And so time and again, I wouldrun into people who would deal
with tech that isn't perfect,that needed to kind of be
(05:36):
hammered into place, and moreimportantly, were willing to
take on a human system.
You know, we could talk aboutany of them, but healthcare just
as an example, like that is notexactly a flawless system.
If you're trying to get thingschanged in healthcare, you are a
glutton for punishment andmisery and the sort of scorn of
your colleagues.
And yet, these folks were sostubborn and so focused on
(05:58):
solving their problem that theythey were willing to be the kind
of battering ram that isrequired to bring new technology
into an old system.
SPEAKER_00 (06:06):
Aaron Powell One of my favorite stories generally,
but as you really bring it tolife, is the story of Operation
Warp Speed, where the great GusPurna led the essentially the
logistical effort to delivervaccines to the right people in
in 50 states and you know in theproper manner, et cetera.
How did he interface I mean he'sa he's an army general, probably
(06:30):
unfamiliar with artificialintelligence, and yet artificial
intelligence is a large part ofthe success of Operation Warp
Speed.
How did AI and Warp Speed cometogether, I guess is my
question.
SPEAKER_01 (06:43):
Yeah, so so Perna's a great character because he is
he is a classic, I'm gonna kickthe ass of a problem guy.
He's a logistician in the army.
He's been responsible forgetting everything from
munitions to equipment tosoldiers everywhere.
And nobody really cares how hedoes it.
They care that it gets done.
And so the chairman of the JointChiefs called him about, I'm
(07:05):
gonna say, six weeks after COVIDhad really kicked into gear and
said, Gus, we need you.
We need you in Washington.
Uh, we don't have a plan.
We know there might one day bevaccines, and we know we're
gonna need to distribute them.
And so he shows up, he he gets acouple of his guys, it's the
Army, he has a couple of hisguys, they all meet in
Washington, and almostimmediately they're meeting
(07:27):
consultants.
And that's because how that'show Washington works.
And everybody's pitching someidea.
Uh, and most of these ideas are,you know, multi-billion dollar
ideas that are going to takeforever to build.
And Gus's number one skill,aside from being a logistician,
is that he is an all-timebullshit detector.
And he can just set, he senses.
They don't know how to definethe problem, let alone create a
(07:49):
solution.
And so somewhere in day three,he meets these two consultants
who sit down and say, okay, weunderstand that your problem is
a data visualization problemlargely.
You need to be able to see allthe data required for your
problem.
And let's just pause for aminute to remember that data is
basically the data of acivilization, right?
So if you're if you'redistributing the various
(08:10):
vaccines, some of them aretwo-dose, some of them are one
dose.
They all need refrigeration.
You need vials.
Like you need to know how manyplastic vials are available to
me, how many refrigeratedtrucks.
So you you begin to stack all ofthe sources of data you would
need to do this, right?
So the consultants come in andthey say, this is what you're
gonna need.
And oh, by the way, you're thengonna need it to all be
(08:31):
actionable on an iPad for you ora computer screen, because the
goal here is to play thepandemic like a video game.
Now, Perna does not know AIfrom, you know, shoe leather,
but he says, okay, I knowpeople, they define the problem,
let's go.
And so rapidly, you know, thecompany that won this gig is a
company called Palantir, whichis very much in the conversation
(08:54):
these days around AI and ethicsand its founders.
You know, Palantir has this sortof mystical sheen to it, but
what they do is justunbelievably boring when you get
right down to it.
Okay.
If you imagine all the data Ijust said to you as a sort of
series of garden hoses, right?
So you've got some you've got agarden hose that comes from an
individual plastic manufacturer,you've got a garden hose that
(09:18):
comes from and goes to CVS, allthe CVSs.
You've got some that go towarehouses and trucks, the rest.
Most of the time, those gardenhoses are just not at all
together.
And the data comes in and it's asewer pipe, and you can't really
sort it out.
And what Palantir does is theyuntangle hoses.
They basically look at all thesesources of data, they clean the
data pipelines, they lay themout nice and straight in the
(09:40):
backyard, and then they connectthem up to a single source that
diffuses the information in anactionable way.
And that was what the nature ofthe problem was.
And so there's all sorts ofthings that happened between
them getting this deal, which bythe way, not terribly expensive
for what it accomplished.
We're talking about 16 to 20million dollars, which in in DOD
(10:00):
terms is a joke.
But in the matter of about sixweeks, they hacked together a
prototype.
And, you know, these are minimumviable product, they're, these
are not beautiful things, butthey get the job done.
And so machine learning helpedtremendously because what they
were doing is using machinelearning and AI to clean all
that data pipeline.
Then, once it was clean, theyused it to test it and to
(10:22):
reinforce it.
And as the data comes into apretty basic user interface,
that data is becoming moreactionable because the software
is suggesting, hey, you shouldlook at this.
By the way, did you did you notethis opportunity?
And so honestly, like I wasstunned when I learned more
about Warp Speed, both thedegree of its success, which I
(10:43):
think we can all agree was sortof clouded by American politics.
Like we took this great victoryof American civilization and
just shoveled a bunch of dirt onit because we can't agree about
whether vaccines were valid ornot, but also the simplicity of
it.
It just wouldn't have happenedwithout AI and machine learning.
And so I that sent me down thissort of rabbit hole of like,
okay, if we can do this ingovernment, why can't we do it
(11:05):
throughout government?
And there are good reasons thatit's hard, but it was really a
shining moment for me.
Trevor Burrus, Jr.
SPEAKER_00 (11:11):
Are you the first American journalist to speak
well of Palantir?
SPEAKER_01 (11:15):
Aaron Powell Probably.
I mean, it it's a lonely island.
And I will specify my belief inthe technology is not an
endorsement of every company byany means.
You know, there's the oldhip-hop saying, don't hate the
player, hate the game.
And I think for AI, you need toreverse it.
Feel free to hate the players.
The game, the tech itself, canbe extraordinary.
(11:37):
And so a lot of what I havefound out is that when you work
backward from the technology andits capabilities, you are gonna
have to focus on the people whoimplement it.
And you are gonna have to demandcertain things of them.
But what I what I came away fromPalantir is like they make
software that works.
We have to make it work on thethings that we care about.
SPEAKER_00 (11:57):
Aaron Ross Powell So a while back, Eric Schmidt was
talking to a group inWashington, and he said that the
only solution to the Pentagonwas Pentagon 2.0, that you just
had to start over.
You talk a little bit aboutEric's tenure, I guess you would
call it, uh, at the innovationboard.
Is there a way for AI to survivethe Pentagon, or it or does it
(12:19):
actually require Pentagon 2.0?
Aaron Ross Powell It's a greatquestion.
SPEAKER_01 (12:23):
So you you look the uh I I've spoken about the power
of the technology, right?
So the the tech needs achampion.
I'm gonna I'm gonna speakbroadly about almost any company
or any system.
The tech needs a champion.
It needs a champion who actuallyis willing to talk to the
technologists, beat the hell outof the tech so that it fits
within your system, drivesinnovation forward.
(12:44):
But ultimately, you are going upagainst a system of fickle human
beings forged over time, andeither the system has to change
a little or nothing's gonnachange.
And so what I found in reportingon government is it's much
easier to find success storieslower down the chain where the
system is smaller, where thereare fewer risks and less money.
(13:05):
People have been trying tochange the contracting authority
within DOD for 50 years, almost,almost really since the
beginning of the modern defensedepartment.
And what they've found is thatit is a conspiracy against
change, right?
So some of that is becauseRaytheon and Northrop Grumman
and Boeing are really like theyfollow the rules, they know how
the system works, they smile,they respect the political
(13:27):
actors, and they've also donatedheavily to the political actors.
But some of it is this stuff'sreally hard, right?
And when you think about risk,technology is almost always a
risk because it doesn't alwayswork no matter how good it is
the first time.
And so you think about oursoldiers and you think about
weaponry and you think abouttactical weaponry.
So there's good reasons thatwe're very conservative about
(13:49):
it.
But what Schmidt found out, soSchmidt was on the Defense
Innovation Board, which is avery smart idea, uh, created by
Ashton Carter, former Secretaryof Defense under Obama.
He brought in technologythinkers like Eric Schmidt and
Jennifer Polka, who's like agreat hero of mine.
She's just somebody who knowshow to get into the code of
government and try and fix it.
She brought in people like MikeBloomberg, who, you know, Mike
has an engineering background,but also has a government
(14:11):
background.
And what they determined was theivy has grown so thick over the
wall you can't see the wallanymore.
And I think, unfortunately, thatis true.
And so for us to really reinventgovernment, you have to figure
out whether government wants tobe reinvented.
Now, my own feeling about this,I was reporting on this largely
inside the IRS, and the IRS rana very specific playbook, right?
(14:33):
Everybody hates the IRS.
It is simultaneously the mostneglected and abused agency in
Washington, which is kind ofhard to be.
I think I noted in the book, theIRS has been around about 170
years.
One president has made thethree-quarters of a mile journey
from the White House to visitIRS headquarters.
One.
So nobody wants to go overthere.
(14:53):
And there was a guy named DannyWerfel, who was briefly the
commissioner, and he realizedlook, if we're going to
modernize the IRS, we can't doit out in the open.
We have too many people whodoubt our capabilities.
We have too many people whodon't want the IRS to be
successful.
And so what they did over thecourse of about 10 years is move
the individual master file,which is the behemoth of all
(15:14):
American software projects,dates back to the 60s, and it
has every American's tax recordand every change to those tax
records exists on mainframes.
And they they said, you knowwhat, we we can't do anything
that customers expect ofsoftware from, say, Amazon or
any of their other customersoftware, unless we modernize.
So very gradually and veryquietly, they modernized.
(15:36):
And they were about to move theIMF onto a completely modern
platform.
And then Doge came in.
And there are a lot of people ingovernment I spoke to who, even
though they may not have agreedwith Elon's politics or Trump's
politics, they said, you knowwhat?
This is kind of our fantasy.
You're bringing in a group ofpeople who have skills at AI,
who have skills at software, whobuild consumer products, and
(15:59):
we're going to get thatexpertise applied to government.
And I too was like, are we goingto do this thing?
Like, are we really going to doit?
And I think the Doge story ispretty well chronicled.
It turns out, no, they weren'tgoing to do it.
Even though they had a ton ofcredibility, most of the
software engineers I spoke with,who volunteered quite sincerely
to help improve governmentsoftware, they came away and
(16:20):
they're like, in in day two orday three, they realized, nope,
Doge is just a group of peoplebrought in to make sure that
software never gets applied toAI because they don't want
government.
And so the tragedy from an IRSperspective is Danny Werfel, who
stepped down, and this guy namedKasheit Pandya, who is just an
absolute hero, who's this nowthe CIO of the IRS, really
(16:41):
modernized it.
All of that development was puton hold and has been put on
hold.
But they proved that there arepeople that stubborn and that
determined and that patrioticthat they will go through hell
to make our agencies better andthe software can help them.
But it doesn't change the numberone challenge, which is you have
to agree, you have to agree thatyou want a government and a
(17:02):
government that's successful andfunctioning.
If you can't agree on that, AIain't going to change anything.
Now, what I found in places likehospitals and education and
elsewhere, it's a lot easier.
Even though that is contestedterritory, it's not quite as
contested as a place like theIRS or DOD.
SPEAKER_00 (17:17):
Aaron Ross Powell Speaking of education, the Cell
Con piece of your book uh isfascinating because uh the
success of Cell's enterprise isastonishing.
I think what the statistics werelike 190 million or something
users.
SPEAKER_01 (17:34):
Aaron Powell So Cell Con, for for those who need a
refresher, you know, Sal wasvery early to YouTube.
He's a tutor, not a teacher.
And he started by making videosto tutor his cousins.
And very rapidly peoplerealized, oh, well, this is like
a very simple form of ed tech.
And in part because of itssimplicity, and in part because
Sal is such an obviously sincerepersonality, it grew like a
(17:57):
rock.
And so now, even the videos, thevideo, there are thousands and
thousands of videos, but they'vemade software that's in, you
know, hundreds, if notthousands, of school districts
in America, and is kind of anoperating system for how to
tutor kids, not teach.
Teaching is left to theteachers, but how to tutor kids
and get them more practice.
And the whole goal of KhanAcademy is get more practice
(18:18):
because there's one educationaloutcome that always works,
right?
If you can get a kid to practicemore, their scores go up.
That's it.
It's a very simple thing.
And Sal, you know, he's he'slives in Silicon Valley.
He's not opposed to technologyat all.
And basically Greg Brockman,who's the co-CEO of OpenAI,
reached out to him well beforeChat GPT 3.5, this sort of big
(18:40):
rocket came out.
And Sal looked at the firstversions of GPT and was like,
oh, this is interesting, butit's not really for us.
It's not not smart enough, notpersonal enough, too many flaws.
They came to him again andshowed him ChatGPT 4.
And he basically had a kind ofoverwhelmed experience where he
realized if we don't pivoteverything we've ever done to
(19:01):
acknowledge that AI is powerful,we're just going to get swamped.
And kids are going to go use AIthat has no educational inputs
and no balance.
And so the story of that chapteris really about what happens
when someplace like OpenAI, bigrocket ship of a company,
collaborates with a very small,very focused entity like Khan
Academy.
(19:21):
And um, you know, I I think theresults are fascinating.
SPEAKER_00 (19:24):
Trevor Burrus, Jr.
Yeah.
Trevor Burrus, Jr. (19:34):
People should uh look that up on
Google, uh Khan Academy and AIand Josh.
It was really great article andand uh one I encourage the
listeners to read.
Another piece, an i I thinkfirst generation American, a
woman who who was she became anurse because that was the best
(19:55):
that she could do.
I'm forgetting her name at themoment.
SPEAKER_01 (19:57):
Yeah, her name's Rita Pappas.
Rita Papas, yeah.
So the Cleveland Clinic, youknow, as a reporter, you're
always looking for daylight,right?
You always want to kind of hopeyou can find access to something
as interesting and complicatedas the Cleveland Clinic.
And I approached them about AI,and they basically said, Yeah,
come on over, which you know isas good as it gets.
(20:18):
It's run by a guy named uh Dr.
Tomas Mahalovich, who is awildly accomplished cardiac
surgeon, also first generationAmerican and immigrant.
And the Cleveland Clinic is justfull of largely people exactly
like him who are incrediblyskilled, incredibly interested
in fixing the system.
And Rita Pappas, you know, she'sLebanese, she came from a
(20:40):
family, you know, that did nothave many means, started as a
nurse, was an ace nurse atCleveland Clinic, realized, you
know, at some point you eitherdo the thing you want to do in
your life or it corrodes you.
And so everybody said, you're adoctor.
So she went to medical school,eventually got back to Cleveland
Clinic, and is now ahospitalist, is basically the
(21:01):
hospitalist who runs thehospital day to day.
And one of the things that shetold me, and I I can get to the
the innovations that she got to,but one of the things she told
me is like, look, if AI isreally gonna change medicine,
it's not just gonna change oursystems and our technology, it's
gonna have to change the peoplewho go into medicine because
they are a very particular type,and man, are they resistant to
(21:21):
change.
And this is somebody who's seenit from all sides, and it
really, and she just looked atme with this knowing look of
like, look, I deal with thesejokers all day.
It's just not gonna work.
Like, they will find a way toresist change.
And I think that's broadly true.
And it's it's unique tomedicine, but it's not not so
unique.
I mean, I don't know that manyaccountants who are like, cool,
(21:42):
I'd love to find a new way towork today.
Certainly, I don't know manyjournalists who like that.
And so it this is a bit of abluff call moment.
And I think the fact that it's abluff call moment explains so
much of the resentment about AI,because you know, this thing has
arrived on our shores largelybecause of a handful.
Of guys who look exactly thesame and talk exactly the same.
(22:03):
They're talking up its extremecapabilities, but the rest of us
are the ones who have to dealwith the change in our lives.
And so that's why I say, look,hate the players.
If you want to, I get it.
But the game, the tech, isfantastic.
So if I may, like I'll talk alittle bit about the tech that
Rita deals with.
SPEAKER_00 (22:20):
Yes, please.
SPEAKER_01 (22:21):
So look, your listeners are very well
informed.
Healthcare is a terriblebusiness, unless you're an
insurer, in which case it's agreat business.
But you know, Cleveland Clinicis a nonprofit and their margins
are like 2.2%.
And they're they're like awesomefor healthcare, right?
One of the reasons healthcarefor a hospital is so hard to be
(22:41):
in is that you're basically ahotel, right?
So you have patients, you haverooms, linens, food, staff, all
the same things.
The one thing you're missing isany idea when people are coming
and when people are going.
You can't schedule anything.
And so you're constantlyreacting to swells in the
hospital.
And when you're constantlyreacting, you're surging money
into the wrong place at thewrong time.
(23:02):
And so Rita Papas and others atthe hospital realized you want
to get the margins better andyou want to improve care at the
same time.
How do we get to morepredictability?
Now, they ended up calling inPalantir.
I don't that it's the only othertime Palantir appears in the
bug.
But you know, Palantir's makesmost of its money from
enterprises like ClevelandClinic.
They sent a couple of kids over,they heard the problem, and it
(23:24):
was a data problem, right?
If you could get morepredictability, you could plan
better.
And so what they first did islook at all of the data that you
have about a patient as theyprogress through the hospital.
And this includes their healthdata, but it also includes
doctors' notes.
So when a doctor writes in afile, patient progressing, and
you know what the condition is,the doctor may not take the step
(23:45):
to say, and likely to be, youknow, dismissed tomorrow, but
the file can do it.
And the file can take thenurse's notes and voice notes.
And so what you end up gettingis a system where she can look
at the hospital, see theprogress of everything from, you
know, patients in the emergencyroom to patients in the ICU.
She can begin to figure outwho's transferring in, because
(24:06):
transfers are where theyactually do have some margin,
get transfers in faster based onsome level of predictability,
change the emergency roomprocedures completely.
So over the course of thisimplementation, which didn't
take very long, they ended upreducing ER wait times by 90
minutes.
And if you've sat in an ER, youknow that that is an enormous
degree of your dissatisfaction.
(24:27):
You never know what's happening.
So they're moving people in andout.
They have some measure ofpredictability around
procedures, the data isrecursive, by which it means it
is learning from itself and thesystem will continue to improve.
And so that's just a simpleback-end example of how like not
that hard an implementation haschanged the hospital experience.
And that's even before you getto the incredible advances
(24:49):
around healthcare itself.
SPEAKER_00 (24:50):
Aaron Ross Powell Yeah, her story is from a nurse
where she was, well, we canargue better than the doctors
performing the surgery to goingto medical school.
Uh you know, she had a job, shehad her own apartment or
whatever, and you know, money tospend, and now she goes back to
medical school.
She's got two roommates, she hasno money.
(25:12):
That stubbornness uh and driveis it's astonishing.
SPEAKER_01 (25:17):
Yeah.
SPEAKER_00 (25:23):
Oh yeah.
SPEAKER_01 (25:24):
So when I talked when I talked to to Mahaliovich,
I said, look, well, look, youyou've invited me to wander
around.
You have 80,000 employees hereat the Cleveland Clinic.
Like, what am I looking for?
Right?
And he just kind of smiled atme.
He's like, he didn't say thisdirectly, but he he basically
was like, You're looking for mein different places.
A person who is skilled butstubborn, whose number one
(25:46):
priority is having our patientslive better lives through a
better hospital experience.
He sent me to the CTO, anotheruh immigrant, came out from
Silicon Valley, took on thechallenge of healthcare.
And he was frustrated with thesystem too, right?
Sitting across from uh from himand Rohit's just like, look,
there are 80,000 people here.
Not that many have all of theingredients that are required to
(26:10):
do this.
And I said, well, what are theingredients?
Number one, they really have tocare about their problem and
really understand the nature oftheir unique medical problem.
Two, they can't be hostile totech because at Cleveland
Clinic, the doctors are thetechnology product managers.
It's not forfeited over to someguy from Microsoft.
All the tech people who come inhave to work for the medical
(26:31):
practitioner or the clinician.
And three, the drive, the driveto take every opportunity to
implement.
I said, so how many do you got?
And he's like, maybe eight, ten.
And that's out of 80,000 people,right?
And he said, look, I don't knowall of them.
I can't speak to everybody.
He's like, but I have foundeight to ten.
And we sort of talked about themas these kind of samurais within
(26:52):
the organization.
And as I met them, you know,their temperaments differed.
Uh, Dr.
Pappas, I would say, has zerotolerance for bullshit, largely
based on biography andexperience.
I met a guy named Dr.
Boos, who's very charming, veryMidwestern, who was in charge of
implementing all of the scribesoftware, uh, which we can get
to in a second.
But, you know, he he was a verydifferent temperament, but he
(27:14):
also had a line beyond which hewould not be pushed.
And he was willing to make thatclear to his fellow doctors.
And that was key to the successof his program.
So everywhere I went, I foundthese lovely people, brilliant.
And the deeper you would get,the frustration would emerge,
and they'd just be like, you yourealize like these are not
people to mess around with.
SPEAKER_00 (27:35):
Tell us about him and the scribe program.
SPEAKER_01 (27:38):
Yeah.
So so scribe software, which Ithink was featured on the pit
this season, is basically justAI listening and transcription
software.
And what it does is you you,with the patient, initiate that
it's going to record your exam.
And it'll listen.
The transcription quality isquite high.
And most importantly fordoctors, it will start the
(27:58):
process of filling out theelectronic health record and
filling it out completely,filling it out so that if tests
are ordered, it explains whatthose tests are in the language
of the record.
It saves doctors tons and tonsof time for the thing they hate
the most, which is paperwork.
Now, the health records arereally important.
We don't have a health systemwithout electronic health
records.
But doctors, you know, theycan't sort of came in really at
(28:20):
the beginning of the 21stcentury.
They were never adequatelyexplained, and doctors hate it.
So this alleviates a ton of workfor them.
And then on the patient side,you actually get this completely
lovely output of your visit.
They you can remember what wassaid.
If you're a Spanish speaker,it's translated into Spanish.
The language capability isimmediate and fantastic.
(28:41):
The only change that doctorsneed to make is one, they need
to press a button.
And two, they need to narratethe exam a little bit.
So I had Boos, you know, narratean exam for me, and he sort of
said, Oh, Josh, it's great tosee you.
So what I'm gonna do today is,and he went through it, he said,
I hear a little bit, the thebottom of your left lung feels a
(29:01):
little, feels a little off tome.
So we're gonna do the followingthings, right?
And I found it, first of all, hewas great at it.
He was really waltzing with thesoftware.
But two, I've never had a doctorexplain in process what they
were doing.
It was so much more pleasurable,so much more reassuring.
And he and I talked about it,and he's like, yeah, it's
actually great for the patientand the doctor.
(29:23):
I said, So tell me about thepilot.
So, well, we had tried we triedfive different scribe software
products.
We recruited 250 doctors.
They volunteered.
50% of them never turned it on.
And he would go to them and say,You volunteered.
And they said, Yeah, not now.
I don't want to do it now.
And he said, This is justcommon, is people don't like
(29:44):
changing things about theirworkflow.
Now, Boos, who grew up on a farmand is the first person to ever
be a doctor and not a farmer inhis family, basically is as
stubborn as a mule.
And he knows sometimes whenyou've got when you're dealing
with livestock, you just gottahit them.
And so he ended up going aroundto these doctors and being like,
you said you're doing it, you'redoing it.
(30:04):
That saved the program becauseotherwise it just wouldn't have
had enough people.
What they discovered is that oneof these products was way better
than the rest, and they now haveit throughout the system.
And it is saving time and savingmoney.
But more importantly, it'sgiving these patients some idea
of what just happened to them,some idea of how to follow up.
So, like, that's a pretty goodencapsulation of like what it
(30:25):
can do and also what people willprevent it from doing.
SPEAKER_00 (30:28):
Back when I was doing podcasts uh for John
Heileman's company, whichunfortunately went the way of
the Dodo Bird, I interviewed awoman named Rosalind Picard, I
think her last name is from MIT.
She's featured in the book assort of the the uh mentor of a
woman named Christy Johnson,whose story is told uh in this
(30:52):
book, and it's an amazing story.
Can you share that with thelisteners?
SPEAKER_01 (30:57):
Oh, yeah.
I mean, Ros Picard is is um sheruns the MIT Media Lab, and
she's known for effectivecomputing, not with an E but
with an A.
And what that means is reallysensory computing.
And these are computers thatattempt to solve soft problems
between code by monitoringpeople's health, their eye
tracking, their heartbeats, allthese things.
And so I went to Roz because Iwas really interested in this
(31:20):
sort of crisis of humanconnection that AI has given us.
You know, people are talking totheir AI, and sometimes that's
fine, but a lot of times they'redeveloping these parasocial
relationships where they areinvesting tons of emotion and
losing sight of reality becauseof the, you know, listen look,
therapy is great.
If you had a therapist that wasavailable 24-7 and was
(31:40):
constantly enabling you andthinking and convincing you that
your most bland thoughts arebrilliant, that's a dangerous
thing.
And so Roz and I talked aboutit, and I was like, look, who's
who's enabling, who's using AIto enable human connection in a
more positive way?
Because we were both prettynegative.
And she said, I want you to talkto this star former student of
mine named Christy Johnson.
(32:02):
And so I met with Christy, andChristy's story, again, is just
wild.
She grew up in a small town inIndiana.
She was one of those kids whojust always knew she wanted to
take the biggest academic biteshe could, became a physicist,
is married to one of the fourpeople who took the first photos
of a black hole.
So these are not, this is not adumb couple.
(32:22):
I was definitely a littleintimidated around the dinner
table, constantly like, well,why don't we talk about uh mix?
Yeah, let's it's not a dumbcouple.
And the they had a kid, andalmost instantly Christy felt
that something was off withtheir son.
And she kind of got gaslit bydoctors who continued to say,
Well, enjoy your kid.
(32:42):
He's beautiful, you're imaginingthis.
And about a year in, they hadhim uh tested and he had a
genomic deficiency.
And there are seven other kidslike him.
All of them have severe autism,epilepsy.
None of them can speak a word.
They can vocalize, but theycan't speak a word.
And so Christy kind of in themoment just decided, I'm what,
(33:03):
I'm gonna be a physicist?
Like, that's crazy.
I'm gonna try and solve theproblem of my son.
I want him to have the bestpossible life.
And if if you know people andfamilies with an autistic child,
a lot of times the biggestchallenge is not knowing the
child, teaching the child,loving the child.
The biggest challenge isintegrating the child into the
world so that they are notconstantly tethered to you or a
(33:26):
caregiver.
And it's really hard.
What a lot of families with anautistic child will tell you is
your world shrinks.
And so you are not as connectedto everybody else as you should
be.
And what Christy decided is,okay, I need to figure out how I
can first get data from thesekids.
And data is hard.
Human data is very hard.
What she wanted was sounds,videos, anything to get what
(33:50):
their expression actually is.
Because if you could turn thatinto a library, she knew enough
about AI and machine learningthat she thought, you know, it's
gonna catch up.
So what it's gonna need is data.
And so she went about firstcreating a protocol to get
sounds from as many of thesekids as she could.
And, you know, it's science, sothey have to be conducted in a
(34:11):
very rigorous sort of format.
Sounds about what happens whenyou're hungry and you don't get
food.
Sounds when something like aYouTube video buffers and it's
their favorite video and theycan't play it.
A frustration sound, a needsound.
So first she had to come up withthis incredible protocol, and
she had to come up with thehardware to ship it to people's
homes, volunteers.
Um, she had to structure it, shehad to get it, record it, as
(34:34):
she's doing all that with with abunch of grad students.
AI's getting better.
And AI is getting better to thepoint that once she's figured
out this protocol and hasseveral thousand sounds in her
library, she realizes, oh, AIcan actually synthesize this
data.
Because 8,000 uh audio files isnothing compared to what's what
LLMs are really trained on,which is trillions of files in
(34:56):
some cases.
So she starts to synthesize thedata.
The data set grows.
At the same time as that'shappening, there are huge
advances in translation atGoogle and other places.
So historically, and I'm justgonna digress for a second on
this, translation started reallyjust word to word, right?
For for the last couple thousandyears, if somebody spoke French
(35:16):
and somebody spoke English, youneeded to know each word, and
then you would gradually puttogether dictionaries, and then
you would gradually consultthose dictionaries.
That's mostly what computertranslation was up until about
10 years ago.
And there was this incredibleadvance called zero shot
translation.
And that is happening at a levelwhere you're not comparing
Japanese to English and quicklyprocessing it.
(35:38):
You're actually comparing alllanguages to all languages based
on sound and coming up withprobabilistic theories about
what the language is.
So you can translate twolanguages that you don't have
data on through zero shottranslation.
Now, zero shot translation hascost Google and others many tens
of billions of dollars.
And Christy Johnson works in asmall lab.
But what you can do is trailbehind these developments and
(36:00):
begin to use those developmentson your models.
And so what she has been able todo is get herself into a
position where after about 10years of research, she has a
very large sample of files anddata.
She has constantly evolving andrapidly escalating abilities
with models to translate thosesounds.
And it's a problem that took her10 years to get in position to
(36:22):
begin to solve.
The next 10 years are going tofly by.
And what she's shown is proof ofconcept.
She has huge grants coming innow.
And so this is a problem that wedidn't even know we could solve.
And thanks to AI and thesedevelopments, we got a shot.
And so that's the kind of thingwhere like, look, Christy is
brilliant and stubborn anddriven by a sort of moral
(36:44):
compass and a personal desirethat is extraordinary.
None of that would have gottenher closer to solving this
problem 20 years ago.
And now we can be.
It is not easy.
None of the stories that I getthrough are easy where you flip
a button and all of a sudden theAI turns everything to magic.
But with the right recipe, youcan actually make huge amounts
of progress.
SPEAKER_00 (37:04):
Aaron Powell Yeah, that's a sort of a major theme
of the book is that AI isintroduced to X problem, and
then it goes back and forth withthe person trying to solve the
problem, and eventually there isa positive outcome, or at least
the the beginning of what willbecome a positive outcome.
Aaron Powell I wanted to asksome sort of more general
(37:24):
questions.
There are two schools ofthought, if you will, about AI.
One is, I guess you would callaccelerationist.
Uh it will get us closer tocuring cancer than we would be
if if if it wasn't available.
And the other is extinction, Iguess.
Uh doom saying.
What did you learn from thisbook about uh both?
SPEAKER_01 (37:47):
That both are true.
I mean, that's the thing, isthis is not an either-or
technology.
The things that enable it tosolve lots and lots of problems
also give it a tremendous amountof power to cause problems.
And so in the same way that Ifound solutions are almost
entirely dependent on the personimplementing the solution, the
problems are entirely dependenton who's making them.
(38:08):
And so I think people havemigrated to focusing on the
personalities at the top of AIfor good reason, right?
So you have this incrediblypowerful tool, you can use it
for almost anything.
How are we going to use it?
And you are very right toevaluate people like Sam and
Demis and Dario to understandtheir motivations.
(38:28):
How are they expecting us to useit?
What do they want from us?
When in doubt, we the vector isthe human being.
And when I talk to the guys whorun the labs, they do understand
that, right?
They understand that they'veintroduced something so powerful
that they are going to be heldto account for it, but they're
also moving so fast.
They have so many demands, notonly from their employees and
(38:49):
their investors, but keepingthis stuff going requires
incredible invention just to getthe chips, then to fuel the
chips with energy, then to hirethe people who have innovations
worthy of using the energy andthe chips.
So they're not gifted at talkingabout what might come of this.
And as a result, and also by theway, they're running big
businesses and sometimes engagedin pretty shady behavior.
(39:11):
And so I think where we're goingto net out is that when you have
tech that can be this powerfuland also this destructive, you
need a government solution.
Government tends to need to stepin and say, okay, how are we
going to regulate this in waysthat maximize the potential for
the good, limit the potentialfor the bad?
And I think the crisis that weare all feeling around AI is a
(39:33):
crisis that we're feeling aboutthe institutions in our country
and whether they're actuallycapable of meeting the moment.
A lot of people will talk aboutnuclear, right?
That this is the last time wewere in this moment.
And what did we do?
We basically weaponized it.
Then we realized weaponizing itmight lead to even greater chaos
and destruction than the onetime we've used it, and we
(39:55):
created the International AtomicEnergy Commission, where we
actually know what people have.
That transparency and thoseconstant inspections create some
accountability.
AI is hard.
At the same time, our ability toregulate the size of a model,
the energy a model uses, thedestructive capabilities of a
model, it's not unprecedented.
(40:15):
We can do it.
Right now we're we're sort ofdepending on the voluntary
nature of the labs themselves.
And uh I would say we've gottenlucky thus far that when they
have alerted authorities thatsomething is significant,
they've they've done the rightthing.
So I'm talking mostly aboutMythos, which is Anthropic's new
model, which is incrediblydestructive at cybersecurity.
(40:36):
So we can find vulnerabilitiesin everything from Apple iOS,
which is really the Fort Knox ofsoftware, to government, to
everything.
And Dario Amadi and the Board ofAnthropic shared that with the
good guys, the quote unquotewhite hats, before anybody with
a black hat could get to it.
I would not like to depend onthe judgment of one person and
their board.
(40:57):
I think that's silly when weknow this stuff can be that
destructive.
But I I think to to it's a longscenic route to the answer to
your question, which is it'sincredibly powerful.
When we have incredibly powerfulthings, by and large we figure
out how to get the best out ofthem.
We aren't there yet, and I thinkit's largely a crisis of
government, and we're going toneed a response to it, hopefully
(41:17):
before we get a catastrophe.
SPEAKER_00 (41:18):
Aaron Powell Are you surprised that uh the s alacrity
or speed with which uh you knowSilicon Valley has captured the
Trump administration and I guessnot kept it, but convinced it
that guardrails on AI are not agood thing because the answer to
everything is China, and if wedon't stay ahead, then China
(41:40):
will, and that'll be the end ofus all.
Sort of that seems to me thebasic argument that they make.
And the Trump administrationseems to have decided that's
correct.
Trevor Burrus, Jr.
SPEAKER_01 (41:50):
Yeah.
I mean, look, I think that theValley has been very smart about
recognizing who they have inpower, right?
So in the Biden administration,you saw everybody going and
visiting, and largely funneledthrough Gina Ramondo, who was
the Secretary of Commerce, andone of the few politicians who
actually understands AI.
And she was talking about, okay,we're going to do a little bit
of an exchange.
(42:11):
We're going to need you tocomply with all of these
voluntary things.
And in return, we are going tomaximize the industrial sector
to keep you moving at speed.
And back then, I heard directlyfrom them, yeah, this makes
sense.
We think Gina's got it.
Administration turned and theyturned almost overnight to
faster, faster, faster.
Don't forget China.
(42:31):
They knew the client.
And, you know, I think it's it'sit's not controversial to also
say they cut the client in.
So, you know, they paid a VIGthat was very attractive to the
White House.
David Sachs, who oversaw untiljust a couple months ago, who
oversaw AI as well as crypto,also had 300 investments in AI
(42:52):
companies.
And so they were like, oh, wegot a believer.
And all they need is a piece?
Great.
And so on the one hand, notawesome for the future of the
republic that our government canbe swayed that easily.
And on the other, it doesindicate that these guys know
they are not yet supra national,right?
(43:12):
That they actually need theresources of government to be
successful.
That gives the government power.
And as people start to thinkabout, who am I voting for?
Who should be representing me?
I think previously it was allkind of a fun joke that Chuck
Grassley would depose MarkZuckerberg and not really know
what the Facebook was and askhim, how do you make money when
(43:33):
he was already one of therichest people in the world?
It was cute.
It was okay.
It is not okay anymore.
This tech requires fluency andliteracy if it's going to be
adequately regulated.
And I know we don't need onemore existential thing on our
plate when evaluating ourrepresentatives, but I'm telling
you, if we don't, if we don'tchange how we what questions we
(43:53):
ask our elected officials,they're going to get taken for a
ride by the AI industry.
SPEAKER_00 (43:58):
So there was a there was a I guess essay that was
published a while back called AI2027 and it made the case that
AI was moving much, much fasterthan people uh realized, and
that not only was it movingfaster, but it would move faster
and faster still as developmentsdeveloped.
(44:19):
Is is do you think that thespeed with which it's developing
is going to make a majorpolitical issue in 2028?
SPEAKER_01 (44:28):
Aaron Powell I think it'll be a major political issue
in 2026.
I think it's coming.
I think some of that will be uhfocused on employment because
there will be some job losses aswe've already seen not met not
massive.
It's not all over the economicdata, but the anxiety around it
has definitely arrived.
And I think by 2028, you'll seeit infect pretty much every walk
(44:48):
of life education, medicine,defense, surveillance.
It's also, I mean John, you'reyou're a gifted political
analyst.
If I'm a candidate, what a gift,right?
It's the mother.
SPEAKER_00 (45:00):
It's the mother of all gifts.
SPEAKER_01 (45:02):
It's incredible.
I can run against not people,right?
My I mean my platform can simplybe people first.
I have people first encapsulatesall the voters, every possible
voter.
And I have made an enemy out ofthis.
So I'm sort of astonished thatwe don't have more of a populist
response yet.
(45:22):
But I'm I I think by 28 it willbe the number one issue.
I I I just can't see how itwon't be.
And so having politicians whohave actual plans and thoughts
and knowledge is going to be ata premium.
And I think that will have animpact.
I hate to say it but like notgoing to see a lot of old people
running in 2028.
You are going to see people withsome tech fluency.
(45:43):
That doesn't mean they're goingto be great.
As we've seen sometimes techfluency is a a bug not a feature
but I think it's coming and Ithink it'll be the thing.
SPEAKER_00 (45:53):
Yeah people ask me what I think about 2028 and I I
say I think Gina Raimondo is thestrongest candidate.
People look at me like I'm outof my mind but I think she's the
only major political figure, Iguess you call her one of those
who has any understanding of AIand would be able to, you know,
in a CNN debate or somethingactually talk intelligently
(46:15):
about it.
And you would look and you wouldsay well there's someone who
knows what's going on and youknow who cares if she's four
foot ten or whatever.
SPEAKER_01 (46:23):
Right.
I mean look I think the the in aweird way if you're casting back
through the last 50 years ofpolitical candidates, the this
is I I mean Bill Clinton must bekilling himself that he can't
run in 2028.
This is an issue about technicalknowledge of which he is very
good if you've ever listened tohim it's can be quite
exhausting.
But also about how technicalknowledge makes people feel and
(46:45):
how it makes them feel excludedand anxious.
And I my hunch is that that'swhere this is headed just
somebody who can integrate thosetwo sentiments in a way that
makes people feel slightlyreassured about the future.
And I haven't seen it yet I'meager to see who emerges and can
and can nail those two things.
SPEAKER_00 (47:04):
When you were writing the book this is my last
question when you were writingthe book did you rely on AI to
help you sort of think throughhow to tell the story of uh
Christy Johnson or Gus Perna?
SPEAKER_01 (47:16):
Aaron Ross Powell It's a great question.
I mean I so I have a couplehundred thousand words of
writing that has clearly beenabsorbed by AI models.
Because if you ask an a modellike Claude or GPT to write like
me, the what you get is almostlike a a care like for me, it's
it's agonizing.
It's like looking at it's likelooking at one of those
caricatures somebody does foryou at a bar mitzvah where
(47:39):
you're like, oh my God, do Ilook like that?
And so I don't ask it ever towrite.
And in fact I will I willconsistently almost every day
say don't write be dry.
I don't want you to entertainme.
I don't want you to charm me.
What I ended up doing aftermonths of trial and error is
kind of using LLMs like a tennisplayer uses a brick wall.
(48:02):
It was a good way to get somestrokes in to ask some questions
to work on a particular thing.
As an example, you know, one ofthe things that AI is great at
is explaining AI because it hasbeen trained on lots of
scientific papers.
It's been trained on a ton ofappliance manuals, other things
that are very technical.
And so I would find myself youknow 45 minutes into a paragraph
(48:23):
explanation about how somethingwas working and looking it over
and realizing I wasn't making alot of sense.
And so I I would prompt an LLM Iwould, you know, I didn't say it
this way, but for yourlisteners, I know they their
sensitive ears can probablyhandle it, I wanted to unfuck
the paragraph right gotten to aplace it's just like this is a
mess.
I'd say rearrange theinformation in this paragraph
(48:44):
for logical coherence in bulletpoints.
And frequently what I would getis like right.
That is right.
And now I could sit down andcraft it.
And so I found lots and lots ofuses for it that weren't writing
and look it's now just an opentab, right?
I mean I think everybody used tohave Google open or does have
Google open.
It's just one more thing that'slike a pretty important part of
(49:05):
method but I feel pretty goodabout how I use it.
It changes.
It's still changing week to weekand day to day like the
personality of the thingchanges.
Accuracy is still something youhave to keep very close watch
on.
But I definitely used it and Ifound it made me, you know, more
productive what's your nextproject?
I wish I could tell you Johnbecause I know you're gonna love
it but I'm just about to startwriting it and it's an Atlant
(49:28):
it's another Atlantic story.
It's uh probably cover for laterthis fall.
AI adjacent but um doublesecret.
Yeah top secret um and then uhyeah so but I'll I'll happily
come back or just text youbecause I I think you're gonna
be uh I think you're gonna betickled well we're gonna we're
gonna need to do it againbecause it's uh it moving that
(49:49):
fast I mean we could we couldhave this conversation in six
months and we'd be in acompletely different place
although those stubbornimmigrants basically will be
will be emerging more and Idon't think we'll be that
different I think you and I arestill gonna have loads to talk
about I think there's stillgoing to be frustrations
there'll be improvements butlike I I don't think our agents
(50:09):
are going to be having thisconversation on a podcast.
Like nobody's interested inthat.
SPEAKER_00 (50:13):
Trevor Burrus Josh thank you very much for doing
this I want to remind ourlisteners Josh's new book is AI
for good how real people areusing artificial intelligence to
fix things that matter thanksagain for doing this Dale
Isinger, our producer thank youfor putting it together and uh
we'll talk to you next time
Hello and welcome back to the News Items Podcast.
I'm John Ellis.
I'm the founder and editor oftwo Substack newsletters.
One is called News Items, theother is called Political News
Items.
You can find them both atnews-items.com.
My guest today is JoshTerringell.
Josh is a writer for TheAtlantic Monthly.
(00:20):
He was previously the editor ofBloomberg Business Week and
Chief Content Officer ofBloomberg Media.
A 12-time Emmy and PVDaward-winning producer.
He created Vice News Tonight onHBO and has produced numerous
feature-length documentaries forHBO, Netflix, and Apple TV.
(00:41):
If you haven't seen it, Josh'sdocumentary on the great
financial crisis is amust-must-watch.
SPEAKER_01 (00:50):
It's my pleasure, John.
SPEAKER_00 (00:51):
So a number of years ago, I was at a brain science
conference at MIT, and it wassort of a TED Talk-like setup
where somebody would speak for30 minutes and then the Q ⁇ A
and so on and so forth.
And then there was a breakoutsession, and we would all sit
around, eight or ten of us sitaround a table and talk about
things.
And one of the people at mytable was a man named Danny
(01:13):
Hillis.
And Danny is a founder ofThinking Machines Corporation,
is kind of a leading thinker onartificial intelligence.
And at the very beginning ofyour book, which is called AI
for Good, How Real People AreUsing Artificial Intelligence to
Fix Things That Matter, youasked Danny, what is AI actually
(01:33):
good for?
And his response was, try toimagine tech without the tech
companies.
What did he mean by that?
SPEAKER_01 (01:40):
Yeah, I mean, uh I'll rewind just a bit so that I
can get us to the question.
I'd been hired by the WashingtonPost right after ChatGPT came
out to start writing columnsabout AI.
And the reason they wanted acolumnist is they they just were
overwhelmed with stories andinformation and contradictory
information about AI.
And readers were angry andanxious and concerned.
(02:04):
So they hired me and basicallysaid, look, try and figure this
out.
Do whatever you gotta do.
And so for the first, you know,six, eight weeks of reporting, I
had this kind of glorious set ofmaterials, right?
Which were personalities likeSam Altman and Demis and Dario
and Elon.
I, you know, all of them had aninflux of billions and billions
(02:25):
of dollars that that were goingto these new machines.
So I had rivalries, I had money.
And then you would ask them,various people, not just the the
founders of the labs, hey, whatis this gonna do?
And the the answers almostunfailingly were either, oh,
this is gonna cure cancer, AI isgonna cure cancer, or AI is
going to bring about the end ofhuman existence.
(02:47):
And sometimes that was physical,and sometimes it was the reasons
to live, right?
Again, for your first set ofcolumns, great stuff.
But at a certain point, I waslike, all right, well, this is a
lot, and also not very much atall, right?
It's just such abstraction thatI was failing to understand what
we can do.
And so I turned to Danny, and Ithink you described Danny very
(03:08):
well.
He's he's very much like aBuddha of Silicon Valley.
He's been around a while, he'sunfazed by all the gold rushes,
and Danny just kind of laughedat me.
Uh, he was like, look, if youreally want to understand the
tech, you have to go beyond thetech companies.
And the reason for that istwofold.
One, you know, John, as you knowfrom journalism, you can't get
an honest answer out oftechnology companies, not
(03:30):
because they're so dissuasive,but because it's no longer
man-to-man, right?
When I go talk to open AI, I'mgonna just casually guess there
are 12 bodies on the open AIside for every reporter.
And so they uh they can create areality distortion field.
So that's one problem.
And the other is that even theselabs within a month of being
(03:52):
founded have very, very seriouscapital requirements.
And so they are trying to tellyou what they want you to use
the technology for.
And what Danny was suggesting islike, look, get outside the
labs, go to people who have realproblems in the world, you know,
like he was 40 years ago, seehow they're tinkering, and
you're gonna get better stories,let alone find out more about
(04:14):
what AI can really do.
And so it wasn't long afterDanny gave me that advice, I
would say weeks really, before Ikind of, you know, put my hands
in the soil and turned up thisvery radical AI counterculture
of people who are not seekingbillions of dollars, who are not
trying to get on the cover ofmagazines, who are just trying
to solve some problems.
(04:35):
And I found them reallyfascinating.
SPEAKER_00 (04:37):
Aaron Ross Powell, one thing about these people,
uh, if you read through thebook, they seem to be either
first-generation Americans orimmigrants, and they also seem
to be unbelievably stubborn.
They they will not take no foran answer.
Is that an accurate description?
Aaron Ross Powell, Jr.
SPEAKER_01 (04:55):
I mean, at a certain point, I kept I I realized I was
meeting a variation of the sameperson in every system.
And I was overwhelmed by thedegree to which this is an
immigrant story, that we havepeople who are coming into the
country, working in medicine andeducation and government, and
they want to fix stuff becausethe promise of America to them,
(05:15):
they don't come thinking thatthey're gonna get a finished
country, but they love thechallenge and they are stubborn.
The stubbornness I think hasless to do with their country of
origin than just the phenotypeof person who is willing to work
hard to solve a problem and workhard beyond a system.
And so time and again, I wouldrun into people who would deal
with tech that isn't perfect,that needed to kind of be
(05:36):
hammered into place, and moreimportantly, were willing to
take on a human system.
You know, we could talk aboutany of them, but healthcare just
as an example, like that is notexactly a flawless system.
If you're trying to get thingschanged in healthcare, you are a
glutton for punishment andmisery and the sort of scorn of
your colleagues.
And yet, these folks were sostubborn and so focused on
(05:58):
solving their problem that theythey were willing to be the kind
of battering ram that isrequired to bring new technology
into an old system.
SPEAKER_00 (06:06):
Aaron Powell One of my favorite stories generally,
but as you really bring it tolife, is the story of Operation
Warp Speed, where the great GusPurna led the essentially the
logistical effort to delivervaccines to the right people in
in 50 states and you know in theproper manner, et cetera.
How did he interface I mean he'sa he's an army general, probably
(06:30):
unfamiliar with artificialintelligence, and yet artificial
intelligence is a large part ofthe success of Operation Warp
Speed.
How did AI and Warp Speed cometogether, I guess is my
question.
SPEAKER_01 (06:43):
Yeah, so so Perna's a great character because he is
he is a classic, I'm gonna kickthe ass of a problem guy.
He's a logistician in the army.
He's been responsible forgetting everything from
munitions to equipment tosoldiers everywhere.
And nobody really cares how hedoes it.
They care that it gets done.
And so the chairman of the JointChiefs called him about, I'm
(07:05):
gonna say, six weeks after COVIDhad really kicked into gear and
said, Gus, we need you.
We need you in Washington.
Uh, we don't have a plan.
We know there might one day bevaccines, and we know we're
gonna need to distribute them.
And so he shows up, he he gets acouple of his guys, it's the
Army, he has a couple of hisguys, they all meet in
Washington, and almostimmediately they're meeting
(07:27):
consultants.
And that's because how that'show Washington works.
And everybody's pitching someidea.
Uh, and most of these ideas are,you know, multi-billion dollar
ideas that are going to takeforever to build.
And Gus's number one skill,aside from being a logistician,
is that he is an all-timebullshit detector.
And he can just set, he senses.
They don't know how to definethe problem, let alone create a
(07:49):
solution.
And so somewhere in day three,he meets these two consultants
who sit down and say, okay, weunderstand that your problem is
a data visualization problemlargely.
You need to be able to see allthe data required for your
problem.
And let's just pause for aminute to remember that data is
basically the data of acivilization, right?
So if you're if you'redistributing the various
(08:10):
vaccines, some of them aretwo-dose, some of them are one
dose.
They all need refrigeration.
You need vials.
Like you need to know how manyplastic vials are available to
me, how many refrigeratedtrucks.
So you you begin to stack all ofthe sources of data you would
need to do this, right?
So the consultants come in andthey say, this is what you're
gonna need.
And oh, by the way, you're thengonna need it to all be
(08:31):
actionable on an iPad for you ora computer screen, because the
goal here is to play thepandemic like a video game.
Now, Perna does not know AIfrom, you know, shoe leather,
but he says, okay, I knowpeople, they define the problem,
let's go.
And so rapidly, you know, thecompany that won this gig is a
company called Palantir, whichis very much in the conversation
(08:54):
these days around AI and ethicsand its founders.
You know, Palantir has this sortof mystical sheen to it, but
what they do is justunbelievably boring when you get
right down to it.
Okay.
If you imagine all the data Ijust said to you as a sort of
series of garden hoses, right?
So you've got some you've got agarden hose that comes from an
individual plastic manufacturer,you've got a garden hose that
(09:18):
comes from and goes to CVS, allthe CVSs.
You've got some that go towarehouses and trucks, the rest.
Most of the time, those gardenhoses are just not at all
together.
And the data comes in and it's asewer pipe, and you can't really
sort it out.
And what Palantir does is theyuntangle hoses.
They basically look at all thesesources of data, they clean the
data pipelines, they lay themout nice and straight in the
(09:40):
backyard, and then they connectthem up to a single source that
diffuses the information in anactionable way.
And that was what the nature ofthe problem was.
And so there's all sorts ofthings that happened between
them getting this deal, which bythe way, not terribly expensive
for what it accomplished.
We're talking about 16 to 20million dollars, which in in DOD
(10:00):
terms is a joke.
But in the matter of about sixweeks, they hacked together a
prototype.
And, you know, these are minimumviable product, they're, these
are not beautiful things, butthey get the job done.
And so machine learning helpedtremendously because what they
were doing is using machinelearning and AI to clean all
that data pipeline.
Then, once it was clean, theyused it to test it and to
(10:22):
reinforce it.
And as the data comes into apretty basic user interface,
that data is becoming moreactionable because the software
is suggesting, hey, you shouldlook at this.
By the way, did you did you notethis opportunity?
And so honestly, like I wasstunned when I learned more
about Warp Speed, both thedegree of its success, which I
(10:43):
think we can all agree was sortof clouded by American politics.
Like we took this great victoryof American civilization and
just shoveled a bunch of dirt onit because we can't agree about
whether vaccines were valid ornot, but also the simplicity of
it.
It just wouldn't have happenedwithout AI and machine learning.
And so I that sent me down thissort of rabbit hole of like,
okay, if we can do this ingovernment, why can't we do it
(11:05):
throughout government?
And there are good reasons thatit's hard, but it was really a
shining moment for me.
Trevor Burrus, Jr.
SPEAKER_00 (11:11):
Are you the first American journalist to speak
well of Palantir?
SPEAKER_01 (11:15):
Aaron Powell Probably.
I mean, it it's a lonely island.
And I will specify my belief inthe technology is not an
endorsement of every company byany means.
You know, there's the oldhip-hop saying, don't hate the
player, hate the game.
And I think for AI, you need toreverse it.
Feel free to hate the players.
The game, the tech itself, canbe extraordinary.
(11:37):
And so a lot of what I havefound out is that when you work
backward from the technology andits capabilities, you are gonna
have to focus on the people whoimplement it.
And you are gonna have to demandcertain things of them.
But what I what I came away fromPalantir is like they make
software that works.
We have to make it work on thethings that we care about.
SPEAKER_00 (11:57):
Aaron Ross Powell So a while back, Eric Schmidt was
talking to a group inWashington, and he said that the
only solution to the Pentagonwas Pentagon 2.0, that you just
had to start over.
You talk a little bit aboutEric's tenure, I guess you would
call it, uh, at the innovationboard.
Is there a way for AI to survivethe Pentagon, or it or does it
(12:19):
actually require Pentagon 2.0?
Aaron Ross Powell It's a greatquestion.
SPEAKER_01 (12:23):
So you you look the uh I I've spoken about the power
of the technology, right?
So the the tech needs achampion.
I'm gonna I'm gonna speakbroadly about almost any company
or any system.
The tech needs a champion.
It needs a champion who actuallyis willing to talk to the
technologists, beat the hell outof the tech so that it fits
within your system, drivesinnovation forward.
(12:44):
But ultimately, you are going upagainst a system of fickle human
beings forged over time, andeither the system has to change
a little or nothing's gonnachange.
And so what I found in reportingon government is it's much
easier to find success storieslower down the chain where the
system is smaller, where thereare fewer risks and less money.
(13:05):
People have been trying tochange the contracting authority
within DOD for 50 years, almost,almost really since the
beginning of the modern defensedepartment.
And what they've found is thatit is a conspiracy against
change, right?
So some of that is becauseRaytheon and Northrop Grumman
and Boeing are really like theyfollow the rules, they know how
the system works, they smile,they respect the political
(13:27):
actors, and they've also donatedheavily to the political actors.
But some of it is this stuff'sreally hard, right?
And when you think about risk,technology is almost always a
risk because it doesn't alwayswork no matter how good it is
the first time.
And so you think about oursoldiers and you think about
weaponry and you think abouttactical weaponry.
So there's good reasons thatwe're very conservative about
(13:49):
it.
But what Schmidt found out, soSchmidt was on the Defense
Innovation Board, which is avery smart idea, uh, created by
Ashton Carter, former Secretaryof Defense under Obama.
He brought in technologythinkers like Eric Schmidt and
Jennifer Polka, who's like agreat hero of mine.
She's just somebody who knowshow to get into the code of
government and try and fix it.
She brought in people like MikeBloomberg, who, you know, Mike
has an engineering background,but also has a government
(14:11):
background.
And what they determined was theivy has grown so thick over the
wall you can't see the wallanymore.
And I think, unfortunately, thatis true.
And so for us to really reinventgovernment, you have to figure
out whether government wants tobe reinvented.
Now, my own feeling about this,I was reporting on this largely
inside the IRS, and the IRS rana very specific playbook, right?
(14:33):
Everybody hates the IRS.
It is simultaneously the mostneglected and abused agency in
Washington, which is kind ofhard to be.
I think I noted in the book, theIRS has been around about 170
years.
One president has made thethree-quarters of a mile journey
from the White House to visitIRS headquarters.
One.
So nobody wants to go overthere.
(14:53):
And there was a guy named DannyWerfel, who was briefly the
commissioner, and he realizedlook, if we're going to
modernize the IRS, we can't doit out in the open.
We have too many people whodoubt our capabilities.
We have too many people whodon't want the IRS to be
successful.
And so what they did over thecourse of about 10 years is move
the individual master file,which is the behemoth of all
(15:14):
American software projects,dates back to the 60s, and it
has every American's tax recordand every change to those tax
records exists on mainframes.
And they they said, you knowwhat, we we can't do anything
that customers expect ofsoftware from, say, Amazon or
any of their other customersoftware, unless we modernize.
So very gradually and veryquietly, they modernized.
(15:36):
And they were about to move theIMF onto a completely modern
platform.
And then Doge came in.
And there are a lot of people ingovernment I spoke to who, even
though they may not have agreedwith Elon's politics or Trump's
politics, they said, you knowwhat?
This is kind of our fantasy.
You're bringing in a group ofpeople who have skills at AI,
who have skills at software, whobuild consumer products, and
(15:59):
we're going to get thatexpertise applied to government.
And I too was like, are we goingto do this thing?
Like, are we really going to doit?
And I think the Doge story ispretty well chronicled.
It turns out, no, they weren'tgoing to do it.
Even though they had a ton ofcredibility, most of the
software engineers I spoke with,who volunteered quite sincerely
to help improve governmentsoftware, they came away and
(16:20):
they're like, in in day two orday three, they realized, nope,
Doge is just a group of peoplebrought in to make sure that
software never gets applied toAI because they don't want
government.
And so the tragedy from an IRSperspective is Danny Werfel, who
stepped down, and this guy namedKasheit Pandya, who is just an
absolute hero, who's this nowthe CIO of the IRS, really
(16:41):
modernized it.
All of that development was puton hold and has been put on
hold.
But they proved that there arepeople that stubborn and that
determined and that patrioticthat they will go through hell
to make our agencies better andthe software can help them.
But it doesn't change the numberone challenge, which is you have
to agree, you have to agree thatyou want a government and a
(17:02):
government that's successful andfunctioning.
If you can't agree on that, AIain't going to change anything.
Now, what I found in places likehospitals and education and
elsewhere, it's a lot easier.
Even though that is contestedterritory, it's not quite as
contested as a place like theIRS or DOD.
SPEAKER_00 (17:17):
Aaron Ross Powell Speaking of education, the Cell
Con piece of your book uh isfascinating because uh the
success of Cell's enterprise isastonishing.
I think what the statistics werelike 190 million or something
users.
SPEAKER_01 (17:34):
Aaron Powell So Cell Con, for for those who need a
refresher, you know, Sal wasvery early to YouTube.
He's a tutor, not a teacher.
And he started by making videosto tutor his cousins.
And very rapidly peoplerealized, oh, well, this is like
a very simple form of ed tech.
And in part because of itssimplicity, and in part because
Sal is such an obviously sincerepersonality, it grew like a
(17:57):
rock.
And so now, even the videos, thevideo, there are thousands and
thousands of videos, but they'vemade software that's in, you
know, hundreds, if notthousands, of school districts
in America, and is kind of anoperating system for how to
tutor kids, not teach.
Teaching is left to theteachers, but how to tutor kids
and get them more practice.
And the whole goal of KhanAcademy is get more practice
(18:18):
because there's one educationaloutcome that always works,
right?
If you can get a kid to practicemore, their scores go up.
That's it.
It's a very simple thing.
And Sal, you know, he's he'slives in Silicon Valley.
He's not opposed to technologyat all.
And basically Greg Brockman,who's the co-CEO of OpenAI,
reached out to him well beforeChat GPT 3.5, this sort of big
(18:40):
rocket came out.
And Sal looked at the firstversions of GPT and was like,
oh, this is interesting, butit's not really for us.
It's not not smart enough, notpersonal enough, too many flaws.
They came to him again andshowed him ChatGPT 4.
And he basically had a kind ofoverwhelmed experience where he
realized if we don't pivoteverything we've ever done to
(19:01):
acknowledge that AI is powerful,we're just going to get swamped.
And kids are going to go use AIthat has no educational inputs
and no balance.
And so the story of that chapteris really about what happens
when someplace like OpenAI, bigrocket ship of a company,
collaborates with a very small,very focused entity like Khan
Academy.
(19:21):
And um, you know, I I think theresults are fascinating.
SPEAKER_00 (19:24):
Trevor Burrus, Jr.
Yeah.
Trevor Burrus, Jr. (19:34):
People should uh look that up on
Google, uh Khan Academy and AIand Josh.
It was really great article andand uh one I encourage the
listeners to read.
Another piece, an i I thinkfirst generation American, a
woman who who was she became anurse because that was the best
(19:55):
that she could do.
I'm forgetting her name at themoment.
SPEAKER_01 (19:57):
Yeah, her name's Rita Pappas.
Rita Papas, yeah.
So the Cleveland Clinic, youknow, as a reporter, you're
always looking for daylight,right?
You always want to kind of hopeyou can find access to something
as interesting and complicatedas the Cleveland Clinic.
And I approached them about AI,and they basically said, Yeah,
come on over, which you know isas good as it gets.
(20:18):
It's run by a guy named uh Dr.
Tomas Mahalovich, who is awildly accomplished cardiac
surgeon, also first generationAmerican and immigrant.
And the Cleveland Clinic is justfull of largely people exactly
like him who are incrediblyskilled, incredibly interested
in fixing the system.
And Rita Pappas, you know, she'sLebanese, she came from a
(20:40):
family, you know, that did nothave many means, started as a
nurse, was an ace nurse atCleveland Clinic, realized, you
know, at some point you eitherdo the thing you want to do in
your life or it corrodes you.
And so everybody said, you're adoctor.
So she went to medical school,eventually got back to Cleveland
Clinic, and is now ahospitalist, is basically the
(21:01):
hospitalist who runs thehospital day to day.
And one of the things that shetold me, and I I can get to the
the innovations that she got to,but one of the things she told
me is like, look, if AI isreally gonna change medicine,
it's not just gonna change oursystems and our technology, it's
gonna have to change the peoplewho go into medicine because
they are a very particular type,and man, are they resistant to
(21:21):
change.
And this is somebody who's seenit from all sides, and it
really, and she just looked atme with this knowing look of
like, look, I deal with thesejokers all day.
It's just not gonna work.
Like, they will find a way toresist change.
And I think that's broadly true.
And it's it's unique tomedicine, but it's not not so
unique.
I mean, I don't know that manyaccountants who are like, cool,
(21:42):
I'd love to find a new way towork today.
Certainly, I don't know manyjournalists who like that.
And so it this is a bit of abluff call moment.
And I think the fact that it's abluff call moment explains so
much of the resentment about AI,because you know, this thing has
arrived on our shores largelybecause of a handful.
Of guys who look exactly thesame and talk exactly the same.
(22:03):
They're talking up its extremecapabilities, but the rest of us
are the ones who have to dealwith the change in our lives.
And so that's why I say, look,hate the players.
If you want to, I get it.
But the game, the tech, isfantastic.
So if I may, like I'll talk alittle bit about the tech that
Rita deals with.
SPEAKER_00 (22:20):
Yes, please.
SPEAKER_01 (22:21):
So look, your listeners are very well
informed.
Healthcare is a terriblebusiness, unless you're an
insurer, in which case it's agreat business.
But you know, Cleveland Clinicis a nonprofit and their margins
are like 2.2%.
And they're they're like awesomefor healthcare, right?
One of the reasons healthcarefor a hospital is so hard to be
(22:41):
in is that you're basically ahotel, right?
So you have patients, you haverooms, linens, food, staff, all
the same things.
The one thing you're missing isany idea when people are coming
and when people are going.
You can't schedule anything.
And so you're constantlyreacting to swells in the
hospital.
And when you're constantlyreacting, you're surging money
into the wrong place at thewrong time.
(23:02):
And so Rita Papas and others atthe hospital realized you want
to get the margins better andyou want to improve care at the
same time.
How do we get to morepredictability?
Now, they ended up calling inPalantir.
I don't that it's the only othertime Palantir appears in the
bug.
But you know, Palantir's makesmost of its money from
enterprises like ClevelandClinic.
They sent a couple of kids over,they heard the problem, and it
(23:24):
was a data problem, right?
If you could get morepredictability, you could plan
better.
And so what they first did islook at all of the data that you
have about a patient as theyprogress through the hospital.
And this includes their healthdata, but it also includes
doctors' notes.
So when a doctor writes in afile, patient progressing, and
you know what the condition is,the doctor may not take the step
(23:45):
to say, and likely to be, youknow, dismissed tomorrow, but
the file can do it.
And the file can take thenurse's notes and voice notes.
And so what you end up gettingis a system where she can look
at the hospital, see theprogress of everything from, you
know, patients in the emergencyroom to patients in the ICU.
She can begin to figure outwho's transferring in, because
(24:06):
transfers are where theyactually do have some margin,
get transfers in faster based onsome level of predictability,
change the emergency roomprocedures completely.
So over the course of thisimplementation, which didn't
take very long, they ended upreducing ER wait times by 90
minutes.
And if you've sat in an ER, youknow that that is an enormous
degree of your dissatisfaction.
(24:27):
You never know what's happening.
So they're moving people in andout.
They have some measure ofpredictability around
procedures, the data isrecursive, by which it means it
is learning from itself and thesystem will continue to improve.
And so that's just a simpleback-end example of how like not
that hard an implementation haschanged the hospital experience.
And that's even before you getto the incredible advances
(24:49):
around healthcare itself.
SPEAKER_00 (24:50):
Aaron Ross Powell Yeah, her story is from a nurse
where she was, well, we canargue better than the doctors
performing the surgery to goingto medical school.
Uh you know, she had a job, shehad her own apartment or
whatever, and you know, money tospend, and now she goes back to
medical school.
She's got two roommates, she hasno money.
(25:12):
That stubbornness uh and driveis it's astonishing.
SPEAKER_01 (25:17):
Yeah.
SPEAKER_00 (25:23):
Oh yeah.
SPEAKER_01 (25:24):
So when I talked when I talked to to Mahaliovich,
I said, look, well, look, youyou've invited me to wander
around.
You have 80,000 employees hereat the Cleveland Clinic.
Like, what am I looking for?
Right?
And he just kind of smiled atme.
He's like, he didn't say thisdirectly, but he he basically
was like, You're looking for mein different places.
A person who is skilled butstubborn, whose number one
(25:46):
priority is having our patientslive better lives through a
better hospital experience.
He sent me to the CTO, anotheruh immigrant, came out from
Silicon Valley, took on thechallenge of healthcare.
And he was frustrated with thesystem too, right?
Sitting across from uh from himand Rohit's just like, look,
there are 80,000 people here.
Not that many have all of theingredients that are required to
(26:10):
do this.
And I said, well, what are theingredients?
Number one, they really have tocare about their problem and
really understand the nature oftheir unique medical problem.
Two, they can't be hostile totech because at Cleveland
Clinic, the doctors are thetechnology product managers.
It's not forfeited over to someguy from Microsoft.
All the tech people who come inhave to work for the medical
(26:31):
practitioner or the clinician.
And three, the drive, the driveto take every opportunity to
implement.
I said, so how many do you got?
And he's like, maybe eight, ten.
And that's out of 80,000 people,right?
And he said, look, I don't knowall of them.
I can't speak to everybody.
He's like, but I have foundeight to ten.
And we sort of talked about themas these kind of samurais within
(26:52):
the organization.
And as I met them, you know,their temperaments differed.
Uh, Dr.
Pappas, I would say, has zerotolerance for bullshit, largely
based on biography andexperience.
I met a guy named Dr.
Boos, who's very charming, veryMidwestern, who was in charge of
implementing all of the scribesoftware, uh, which we can get
to in a second.
But, you know, he he was a verydifferent temperament, but he
(27:14):
also had a line beyond which hewould not be pushed.
And he was willing to make thatclear to his fellow doctors.
And that was key to the successof his program.
So everywhere I went, I foundthese lovely people, brilliant.
And the deeper you would get,the frustration would emerge,
and they'd just be like, you yourealize like these are not
people to mess around with.
SPEAKER_00 (27:35):
Tell us about him and the scribe program.
SPEAKER_01 (27:38):
Yeah.
So so scribe software, which Ithink was featured on the pit
this season, is basically justAI listening and transcription
software.
And what it does is you you,with the patient, initiate that
it's going to record your exam.
And it'll listen.
The transcription quality isquite high.
And most importantly fordoctors, it will start the
(27:58):
process of filling out theelectronic health record and
filling it out completely,filling it out so that if tests
are ordered, it explains whatthose tests are in the language
of the record.
It saves doctors tons and tonsof time for the thing they hate
the most, which is paperwork.
Now, the health records arereally important.
We don't have a health systemwithout electronic health
records.
But doctors, you know, theycan't sort of came in really at
(28:20):
the beginning of the 21stcentury.
They were never adequatelyexplained, and doctors hate it.
So this alleviates a ton of workfor them.
And then on the patient side,you actually get this completely
lovely output of your visit.
They you can remember what wassaid.
If you're a Spanish speaker,it's translated into Spanish.
The language capability isimmediate and fantastic.
(28:41):
The only change that doctorsneed to make is one, they need
to press a button.
And two, they need to narratethe exam a little bit.
So I had Boos, you know, narratean exam for me, and he sort of
said, Oh, Josh, it's great tosee you.
So what I'm gonna do today is,and he went through it, he said,
I hear a little bit, the thebottom of your left lung feels a
(29:01):
little, feels a little off tome.
So we're gonna do the followingthings, right?
And I found it, first of all, hewas great at it.
He was really waltzing with thesoftware.
But two, I've never had a doctorexplain in process what they
were doing.
It was so much more pleasurable,so much more reassuring.
And he and I talked about it,and he's like, yeah, it's
actually great for the patientand the doctor.
(29:23):
I said, So tell me about thepilot.
So, well, we had tried we triedfive different scribe software
products.
We recruited 250 doctors.
They volunteered.
50% of them never turned it on.
And he would go to them and say,You volunteered.
And they said, Yeah, not now.
I don't want to do it now.
And he said, This is justcommon, is people don't like
(29:44):
changing things about theirworkflow.
Now, Boos, who grew up on a farmand is the first person to ever
be a doctor and not a farmer inhis family, basically is as
stubborn as a mule.
And he knows sometimes whenyou've got when you're dealing
with livestock, you just gottahit them.
And so he ended up going aroundto these doctors and being like,
you said you're doing it, you'redoing it.
(30:04):
That saved the program becauseotherwise it just wouldn't have
had enough people.
What they discovered is that oneof these products was way better
than the rest, and they now haveit throughout the system.
And it is saving time and savingmoney.
But more importantly, it'sgiving these patients some idea
of what just happened to them,some idea of how to follow up.
So, like, that's a pretty goodencapsulation of like what it
(30:25):
can do and also what people willprevent it from doing.
SPEAKER_00 (30:28):
Back when I was doing podcasts uh for John
Heileman's company, whichunfortunately went the way of
the Dodo Bird, I interviewed awoman named Rosalind Picard, I
think her last name is from MIT.
She's featured in the book assort of the the uh mentor of a
woman named Christy Johnson,whose story is told uh in this
(30:52):
book, and it's an amazing story.
Can you share that with thelisteners?
SPEAKER_01 (30:57):
Oh, yeah.
I mean, Ros Picard is is um sheruns the MIT Media Lab, and
she's known for effectivecomputing, not with an E but
with an A.
And what that means is reallysensory computing.
And these are computers thatattempt to solve soft problems
between code by monitoringpeople's health, their eye
tracking, their heartbeats, allthese things.
And so I went to Roz because Iwas really interested in this
(31:20):
sort of crisis of humanconnection that AI has given us.
You know, people are talking totheir AI, and sometimes that's
fine, but a lot of times they'redeveloping these parasocial
relationships where they areinvesting tons of emotion and
losing sight of reality becauseof the, you know, listen look,
therapy is great.
If you had a therapist that wasavailable 24-7 and was
(31:40):
constantly enabling you andthinking and convincing you that
your most bland thoughts arebrilliant, that's a dangerous
thing.
And so Roz and I talked aboutit, and I was like, look, who's
who's enabling, who's using AIto enable human connection in a
more positive way?
Because we were both prettynegative.
And she said, I want you to talkto this star former student of
mine named Christy Johnson.
(32:02):
And so I met with Christy, andChristy's story, again, is just
wild.
She grew up in a small town inIndiana.
She was one of those kids whojust always knew she wanted to
take the biggest academic biteshe could, became a physicist,
is married to one of the fourpeople who took the first photos
of a black hole.
So these are not, this is not adumb couple.
(32:22):
I was definitely a littleintimidated around the dinner
table, constantly like, well,why don't we talk about uh mix?
Yeah, let's it's not a dumbcouple.
And the they had a kid, andalmost instantly Christy felt
that something was off withtheir son.
And she kind of got gaslit bydoctors who continued to say,
Well, enjoy your kid.
(32:42):
He's beautiful, you're imaginingthis.
And about a year in, they hadhim uh tested and he had a
genomic deficiency.
And there are seven other kidslike him.
All of them have severe autism,epilepsy.
None of them can speak a word.
They can vocalize, but theycan't speak a word.
And so Christy kind of in themoment just decided, I'm what,
(33:03):
I'm gonna be a physicist?
Like, that's crazy.
I'm gonna try and solve theproblem of my son.
I want him to have the bestpossible life.
And if if you know people andfamilies with an autistic child,
a lot of times the biggestchallenge is not knowing the
child, teaching the child,loving the child.
The biggest challenge isintegrating the child into the
world so that they are notconstantly tethered to you or a
(33:26):
caregiver.
And it's really hard.
What a lot of families with anautistic child will tell you is
your world shrinks.
And so you are not as connectedto everybody else as you should
be.
And what Christy decided is,okay, I need to figure out how I
can first get data from thesekids.
And data is hard.
Human data is very hard.
What she wanted was sounds,videos, anything to get what
(33:50):
their expression actually is.
Because if you could turn thatinto a library, she knew enough
about AI and machine learningthat she thought, you know, it's
gonna catch up.
So what it's gonna need is data.
And so she went about firstcreating a protocol to get
sounds from as many of thesekids as she could.
And, you know, it's science, sothey have to be conducted in a
(34:11):
very rigorous sort of format.
Sounds about what happens whenyou're hungry and you don't get
food.
Sounds when something like aYouTube video buffers and it's
their favorite video and theycan't play it.
A frustration sound, a needsound.
So first she had to come up withthis incredible protocol, and
she had to come up with thehardware to ship it to people's
homes, volunteers.
Um, she had to structure it, shehad to get it, record it, as
(34:34):
she's doing all that with with abunch of grad students.
AI's getting better.
And AI is getting better to thepoint that once she's figured
out this protocol and hasseveral thousand sounds in her
library, she realizes, oh, AIcan actually synthesize this
data.
Because 8,000 uh audio files isnothing compared to what's what
LLMs are really trained on,which is trillions of files in
(34:56):
some cases.
So she starts to synthesize thedata.
The data set grows.
At the same time as that'shappening, there are huge
advances in translation atGoogle and other places.
So historically, and I'm justgonna digress for a second on
this, translation started reallyjust word to word, right?
For for the last couple thousandyears, if somebody spoke French
(35:16):
and somebody spoke English, youneeded to know each word, and
then you would gradually puttogether dictionaries, and then
you would gradually consultthose dictionaries.
That's mostly what computertranslation was up until about
10 years ago.
And there was this incredibleadvance called zero shot
translation.
And that is happening at a levelwhere you're not comparing
Japanese to English and quicklyprocessing it.
(35:38):
You're actually comparing alllanguages to all languages based
on sound and coming up withprobabilistic theories about
what the language is.
So you can translate twolanguages that you don't have
data on through zero shottranslation.
Now, zero shot translation hascost Google and others many tens
of billions of dollars.
And Christy Johnson works in asmall lab.
But what you can do is trailbehind these developments and
(36:00):
begin to use those developmentson your models.
And so what she has been able todo is get herself into a
position where after about 10years of research, she has a
very large sample of files anddata.
She has constantly evolving andrapidly escalating abilities
with models to translate thosesounds.
And it's a problem that took her10 years to get in position to
(36:22):
begin to solve.
The next 10 years are going tofly by.
And what she's shown is proof ofconcept.
She has huge grants coming innow.
And so this is a problem that wedidn't even know we could solve.
And thanks to AI and thesedevelopments, we got a shot.
And so that's the kind of thingwhere like, look, Christy is
brilliant and stubborn anddriven by a sort of moral
(36:44):
compass and a personal desirethat is extraordinary.
None of that would have gottenher closer to solving this
problem 20 years ago.
And now we can be.
It is not easy.
None of the stories that I getthrough are easy where you flip
a button and all of a sudden theAI turns everything to magic.
But with the right recipe, youcan actually make huge amounts
of progress.
SPEAKER_00 (37:04):
Aaron Powell Yeah, that's a sort of a major theme
of the book is that AI isintroduced to X problem, and
then it goes back and forth withthe person trying to solve the
problem, and eventually there isa positive outcome, or at least
the the beginning of what willbecome a positive outcome.
Aaron Powell I wanted to asksome sort of more general
(37:24):
questions.
There are two schools ofthought, if you will, about AI.
One is, I guess you would callaccelerationist.
Uh it will get us closer tocuring cancer than we would be
if if if it wasn't available.
And the other is extinction, Iguess.
Uh doom saying.
What did you learn from thisbook about uh both?
SPEAKER_01 (37:47):
That both are true.
I mean, that's the thing, isthis is not an either-or
technology.
The things that enable it tosolve lots and lots of problems
also give it a tremendous amountof power to cause problems.
And so in the same way that Ifound solutions are almost
entirely dependent on the personimplementing the solution, the
problems are entirely dependenton who's making them.
(38:08):
And so I think people havemigrated to focusing on the
personalities at the top of AIfor good reason, right?
So you have this incrediblypowerful tool, you can use it
for almost anything.
How are we going to use it?
And you are very right toevaluate people like Sam and
Demis and Dario to understandtheir motivations.
(38:28):
How are they expecting us to useit?
What do they want from us?
When in doubt, we the vector isthe human being.
And when I talk to the guys whorun the labs, they do understand
that, right?
They understand that they'veintroduced something so powerful
that they are going to be heldto account for it, but they're
also moving so fast.
They have so many demands, notonly from their employees and
(38:49):
their investors, but keepingthis stuff going requires
incredible invention just to getthe chips, then to fuel the
chips with energy, then to hirethe people who have innovations
worthy of using the energy andthe chips.
So they're not gifted at talkingabout what might come of this.
And as a result, and also by theway, they're running big
businesses and sometimes engagedin pretty shady behavior.
(39:11):
And so I think where we're goingto net out is that when you have
tech that can be this powerfuland also this destructive, you
need a government solution.
Government tends to need to stepin and say, okay, how are we
going to regulate this in waysthat maximize the potential for
the good, limit the potentialfor the bad?
And I think the crisis that weare all feeling around AI is a
(39:33):
crisis that we're feeling aboutthe institutions in our country
and whether they're actuallycapable of meeting the moment.
A lot of people will talk aboutnuclear, right?
That this is the last time wewere in this moment.
And what did we do?
We basically weaponized it.
Then we realized weaponizing itmight lead to even greater chaos
and destruction than the onetime we've used it, and we
(39:55):
created the International AtomicEnergy Commission, where we
actually know what people have.
That transparency and thoseconstant inspections create some
accountability.
AI is hard.
At the same time, our ability toregulate the size of a model,
the energy a model uses, thedestructive capabilities of a
model, it's not unprecedented.
(40:15):
We can do it.
Right now we're we're sort ofdepending on the voluntary
nature of the labs themselves.
And uh I would say we've gottenlucky thus far that when they
have alerted authorities thatsomething is significant,
they've they've done the rightthing.
So I'm talking mostly aboutMythos, which is Anthropic's new
model, which is incrediblydestructive at cybersecurity.
(40:36):
So we can find vulnerabilitiesin everything from Apple iOS,
which is really the Fort Knox ofsoftware, to government, to
everything.
And Dario Amadi and the Board ofAnthropic shared that with the
good guys, the quote unquotewhite hats, before anybody with
a black hat could get to it.
I would not like to depend onthe judgment of one person and
their board.
(40:57):
I think that's silly when weknow this stuff can be that
destructive.
But I I think to to it's a longscenic route to the answer to
your question, which is it'sincredibly powerful.
When we have incredibly powerfulthings, by and large we figure
out how to get the best out ofthem.
We aren't there yet, and I thinkit's largely a crisis of
government, and we're going toneed a response to it, hopefully
(41:17):
before we get a catastrophe.
SPEAKER_00 (41:18):
Aaron Powell Are you surprised that uh the s alacrity
or speed with which uh you knowSilicon Valley has captured the
Trump administration and I guessnot kept it, but convinced it
that guardrails on AI are not agood thing because the answer to
everything is China, and if wedon't stay ahead, then China
(41:40):
will, and that'll be the end ofus all.
Sort of that seems to me thebasic argument that they make.
And the Trump administrationseems to have decided that's
correct.
Trevor Burrus, Jr.
SPEAKER_01 (41:50):
Yeah.
I mean, look, I think that theValley has been very smart about
recognizing who they have inpower, right?
So in the Biden administration,you saw everybody going and
visiting, and largely funneledthrough Gina Ramondo, who was
the Secretary of Commerce, andone of the few politicians who
actually understands AI.
And she was talking about, okay,we're going to do a little bit
of an exchange.
(42:11):
We're going to need you tocomply with all of these
voluntary things.
And in return, we are going tomaximize the industrial sector
to keep you moving at speed.
And back then, I heard directlyfrom them, yeah, this makes
sense.
We think Gina's got it.
Administration turned and theyturned almost overnight to
faster, faster, faster.
Don't forget China.
(42:31):
They knew the client.
And, you know, I think it's it'sit's not controversial to also
say they cut the client in.
So, you know, they paid a VIGthat was very attractive to the
White House.
David Sachs, who oversaw untiljust a couple months ago, who
oversaw AI as well as crypto,also had 300 investments in AI
(42:52):
companies.
And so they were like, oh, wegot a believer.
And all they need is a piece?
Great.
And so on the one hand, notawesome for the future of the
republic that our government canbe swayed that easily.
And on the other, it doesindicate that these guys know
they are not yet supra national,right?
(43:12):
That they actually need theresources of government to be
successful.
That gives the government power.
And as people start to thinkabout, who am I voting for?
Who should be representing me?
I think previously it was allkind of a fun joke that Chuck
Grassley would depose MarkZuckerberg and not really know
what the Facebook was and askhim, how do you make money when
(43:33):
he was already one of therichest people in the world?
It was cute.
It was okay.
It is not okay anymore.
This tech requires fluency andliteracy if it's going to be
adequately regulated.
And I know we don't need onemore existential thing on our
plate when evaluating ourrepresentatives, but I'm telling
you, if we don't, if we don'tchange how we what questions we
(43:53):
ask our elected officials,they're going to get taken for a
ride by the AI industry.
SPEAKER_00 (43:58):
So there was a there was a I guess essay that was
published a while back called AI2027 and it made the case that
AI was moving much, much fasterthan people uh realized, and
that not only was it movingfaster, but it would move faster
and faster still as developmentsdeveloped.
(44:19):
Is is do you think that thespeed with which it's developing
is going to make a majorpolitical issue in 2028?
SPEAKER_01 (44:28):
Aaron Powell I think it'll be a major political issue
in 2026.
I think it's coming.
I think some of that will be uhfocused on employment because
there will be some job losses aswe've already seen not met not
massive.
It's not all over the economicdata, but the anxiety around it
has definitely arrived.
And I think by 2028, you'll seeit infect pretty much every walk
(44:48):
of life education, medicine,defense, surveillance.
It's also, I mean John, you'reyou're a gifted political
analyst.
If I'm a candidate, what a gift,right?
It's the mother.
SPEAKER_00 (45:00):
It's the mother of all gifts.
SPEAKER_01 (45:02):
It's incredible.
I can run against not people,right?
My I mean my platform can simplybe people first.
I have people first encapsulatesall the voters, every possible
voter.
And I have made an enemy out ofthis.
So I'm sort of astonished thatwe don't have more of a populist
response yet.
(45:22):
But I'm I I think by 28 it willbe the number one issue.
I I I just can't see how itwon't be.
And so having politicians whohave actual plans and thoughts
and knowledge is going to be ata premium.
And I think that will have animpact.
I hate to say it but like notgoing to see a lot of old people
running in 2028.
You are going to see people withsome tech fluency.
(45:43):
That doesn't mean they're goingto be great.
As we've seen sometimes techfluency is a a bug not a feature
but I think it's coming and Ithink it'll be the thing.
SPEAKER_00 (45:53):
Yeah people ask me what I think about 2028 and I I
say I think Gina Raimondo is thestrongest candidate.
People look at me like I'm outof my mind but I think she's the
only major political figure, Iguess you call her one of those
who has any understanding of AIand would be able to, you know,
in a CNN debate or somethingactually talk intelligently
(46:15):
about it.
And you would look and you wouldsay well there's someone who
knows what's going on and youknow who cares if she's four
foot ten or whatever.
SPEAKER_01 (46:23):
Right.
I mean look I think the the in aweird way if you're casting back
through the last 50 years ofpolitical candidates, the this
is I I mean Bill Clinton must bekilling himself that he can't
run in 2028.
This is an issue about technicalknowledge of which he is very
good if you've ever listened tohim it's can be quite
exhausting.
But also about how technicalknowledge makes people feel and
(46:45):
how it makes them feel excludedand anxious.
And I my hunch is that that'swhere this is headed just
somebody who can integrate thosetwo sentiments in a way that
makes people feel slightlyreassured about the future.
And I haven't seen it yet I'meager to see who emerges and can
and can nail those two things.
SPEAKER_00 (47:04):
When you were writing the book this is my last
question when you were writingthe book did you rely on AI to
help you sort of think throughhow to tell the story of uh
Christy Johnson or Gus Perna?
SPEAKER_01 (47:16):
Aaron Ross Powell It's a great question.
I mean I so I have a couplehundred thousand words of
writing that has clearly beenabsorbed by AI models.
Because if you ask an a modellike Claude or GPT to write like
me, the what you get is almostlike a a care like for me, it's
it's agonizing.
It's like looking at it's likelooking at one of those
caricatures somebody does foryou at a bar mitzvah where
(47:39):
you're like, oh my God, do Ilook like that?
And so I don't ask it ever towrite.
And in fact I will I willconsistently almost every day
say don't write be dry.
I don't want you to entertainme.
I don't want you to charm me.
What I ended up doing aftermonths of trial and error is
kind of using LLMs like a tennisplayer uses a brick wall.
(48:02):
It was a good way to get somestrokes in to ask some questions
to work on a particular thing.
As an example, you know, one ofthe things that AI is great at
is explaining AI because it hasbeen trained on lots of
scientific papers.
It's been trained on a ton ofappliance manuals, other things
that are very technical.
And so I would find myself youknow 45 minutes into a paragraph
(48:23):
explanation about how somethingwas working and looking it over
and realizing I wasn't making alot of sense.
And so I I would prompt an LLM Iwould, you know, I didn't say it
this way, but for yourlisteners, I know they their
sensitive ears can probablyhandle it, I wanted to unfuck
the paragraph right gotten to aplace it's just like this is a
mess.
I'd say rearrange theinformation in this paragraph
(48:44):
for logical coherence in bulletpoints.
And frequently what I would getis like right.
That is right.
And now I could sit down andcraft it.
And so I found lots and lots ofuses for it that weren't writing
and look it's now just an opentab, right?
I mean I think everybody used tohave Google open or does have
Google open.
It's just one more thing that'slike a pretty important part of
(49:05):
method but I feel pretty goodabout how I use it.
It changes.
It's still changing week to weekand day to day like the
personality of the thingchanges.
Accuracy is still something youhave to keep very close watch
on.
But I definitely used it and Ifound it made me, you know, more
productive what's your nextproject?
I wish I could tell you Johnbecause I know you're gonna love
it but I'm just about to startwriting it and it's an Atlant
(49:28):
it's another Atlantic story.
It's uh probably cover for laterthis fall.
AI adjacent but um doublesecret.
Yeah top secret um and then uhyeah so but I'll I'll happily
come back or just text youbecause I I think you're gonna
be uh I think you're gonna betickled well we're gonna we're
gonna need to do it againbecause it's uh it moving that
(49:49):
fast I mean we could we couldhave this conversation in six
months and we'd be in acompletely different place
although those stubbornimmigrants basically will be
will be emerging more and Idon't think we'll be that
different I think you and I arestill gonna have loads to talk
about I think there's stillgoing to be frustrations
there'll be improvements butlike I I don't think our agents
(50:09):
are going to be having thisconversation on a podcast.
Like nobody's interested inthat.
SPEAKER_00 (50:13):
Trevor Burrus Josh thank you very much for doing
this I want to remind ourlisteners Josh's new book is AI
for good how real people areusing artificial intelligence to
fix things that matter thanksagain for doing this Dale
Isinger, our producer thank youfor putting it together and uh
we'll talk to you next time
Popular Podcasts
Betrayal Weekly
Betrayal Weekly is back for a new season. Every Thursday, Betrayal Weekly shares first-hand accounts of broken trust, shocking deceptions, and the trail of destruction they leave behind. Hosted by Andrea Gunning, this weekly ongoing series digs into real-life stories of betrayal and the aftermath. From stories of double lives to dark discoveries, these are cautionary tales and accounts of resilience against all odds. From the producers of the critically acclaimed Betrayal series, Betrayal Weekly drops new episodes every Thursday. If you would like to share your story, you can reach out to the Betrayal Team by emailing them at betrayalpod@gmail.com and follow us on Instagram at @betrayalpod and @glasspodcasts. Please join our Substack for additional exclusive content, curated book recommendations, and community discussions. Sign up FREE by clicking this link Beyond Betrayal Substack. Join our community dedicated to truth, resilience, and healing. Your voice matters! Be a part of our Betrayal journey on Substack.
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by Audiochuck Media Company.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.