June 13, 2025 • 30 mins
What’s inside AI’s black box? This week in the News Roundup, Oz unpacks the uncomfortable truth that even the people building today’s AI models often can’t explain how they work — or why they behave the way they do. But that hasn’t stopped tech companies from pushing colleges and universities to embrace chatbots. On TechSupport, 404 Media’s Jason Koebler explains the strange world of airport body scans and the future of the TSA.
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:12):
Welcome to Tech Stuff, a production of iHeart Podcasts and Kaleidoscope.
I'ma's Voloshan and carac prices out this week, so I'll
be bringing you the headlines of the week, including AI's
black box problem and open Eyes push to infiltrate college campuses.
And on today's Tech Support segment, we'll talk to four
or four Media's Jason Kebler about the weird world of
(00:35):
body scans and the future of TSA security.
Speaker 2 (00:39):
Airplanes and airports in general are interesting because it feels
like new technologies are like rolled out there before they're
rolled out more broadly into society.
Speaker 1 (00:50):
All of that On the weekend Tech It's a Spooky Day, Friday,
June thirteenth. As I mentioned, Carras out this week, so
we're going to dive right into the headlines, and a
couple of big stories have been on my mind, both
highlighting that as AI becomes more and more ubiquitous and
(01:11):
as a rush to deploy it, we still don't fully
understand how these systems actually work. AI models are such
a black box that even the developers who build them
can't always explain or predict the behavior of their models,
And this week, Axios released a detailed roundup of how
leaders are simultaneously admitting this lack of knowledge while also
(01:33):
pushing for rapid deployment and implementation. Here's an example. Anthropic,
while testing its latest model, Clawed four, gave the model
access to corporate emails, setting up this fictional scenario where
Claude knew that one it was going to be shut
down and two that the engineer responsible for shutting it
(01:54):
down was having an affair. In eighty four percent of
cases when this test was run, the model attempted to
blackmail the engineer, supposedly to preserve itself. And here's the thing.
Anthropic doesn't know how the new model chose blackmail as
a tactic or how to prevent it from doing so
(02:15):
in future, but Anthropic released the model. Now there is
an accepted term to refer to this problem I that
we don't yet understand the why of AI's behavior, and
that term is interpretability. CEOs of AI companies like open
Ai Samultman and Anthropics Dario Amode have openly acknowledged that
(02:36):
quote interpretability is an important problem to solve and that
our lack of understanding about how these models work can
pose significant risk in April, A. Mooday wrote that quote
people outside the field are often surprised and alarmed to
learn that we do not understand how our AI creations work.
They are right to be concerned. This lack of understanding
(02:58):
is essentially unprecedented in the history of technology. The question
to me is is there enough investment and attention going
towards this problem versus the race forever stronger performance between
privately funded companies. AI is being developed and implemented without
meaningful guardrails, and that's something that US government is actively encouraging.
(03:19):
In fact, President Trump's Big Beautiful Bill includes a provision
for a decade long ban on states attempting to regulate AI,
either through new regulations or through enforcing existing regulations. Now
the bill has been passed by the House and is
being heard in the Senate. Changes are very possible because
even Republican senators like Josh Hawley of Missouri and Marshall
(03:43):
Blackburn of Tennessee have pushed back on this ban. Now,
for those hoping humans will maintain control over AI, there
is some potentially encouraging news coming out of Apple This week.
The company's machine learning researchers publish an academic paper with
very buzzy title, The Illusion of Thinking. The paper studies
(04:05):
reasoning models like open Aiyes three and deep seeks are one,
which are models designed to problem solve. What the Apple
paper found is that when these reasoning models are presented
with complex logic problems, the models fail to solve the problems.
Mashable reports that in the study, reasoning models were given
classic logic puzzles like jumping checker pieces into empty spaces,
(04:27):
or the river crossing problem, the one involving a fox,
a chicken, and a bag of grain. This is a
pretty simple test of a human's ability to problem solve,
because once you figure out the rules, it's actually pretty
easy to continue solving these problems even when they get
more complex. But these reasoning models start to fail at
a certain point. So the big question is is this
(04:49):
simply a bump in the road or is it indicative
of a larger problem in how these reasoning models actually work.
Which brings me to my next headline from the New
York Times, Welcome to campus, here's your chat GPT. Because
if critical thinking and logical reasoning skills are where humans
still have an edge over AI, it's pretty reasonable to me.
(05:11):
There are concerns about the rise of AI on college campuses.
Apparently Open Ai has a plan to integrate chatbots into
every facet of student life, and they're calling the goal
of this AI integrated higher ed quote AI Native Universities.
A suite of premium services called chat gpt Edu is
(05:31):
being sold to universities for faculty and students to use,
and the company is promising AI tutors and chatbots that
can do everything from conduct mock job interviews to quiz
you before exams. Schools are getting in on the action.
California State University is making chatchipt available to more than
four hundred and sixty thousand students. Duke University made its
(05:54):
own AI powered platform called Duke gpt, and in June
they began offering unlimited Chatchipt access to students and faculty.
Even if a campus isn't striking a deal with open Ai.
There's a marketing campaign targeting the handful and I assume
it is a handful of students who aren't yet using chatchypt.
(06:16):
Many find this all deeply troubling, especially as there's new
data and research to suggest that so called cognitive offloading
I letting a chatbot write your first draft or think
for you makes you measurably less able to problem solve yourself.
On the other hand, recent grads are facing historically tough
(06:37):
job market, especially in fields where AI is starting to
automate tasks like writing code. So these schools are trying
to boost their students' prospects by providing them with AI
tools and with the skills to make the most of
the tools. So is this strange moment where we don't
really understand how our lambs make their decisions or how
using them affects our brains. But they're here, We're using
(07:00):
them anyway, and we and they are adapting in real time.
As The Times puts it, quote open ayes pushed to
AIIFI college education amounts to a national experiment on millions
of students. I've got a couple more headlines for you,
(07:22):
and one starts with an old idea brought to life
with modern technology. About half of all people on Earth
experience some form of water scarcity. But could that change
if you drink the ocean. Desalination, the process of removing
salt to create clean drinking water, has been possible for decades,
but it's the most expensive way to create clean drinking water. However,
(07:45):
The Wall Street Journal recently spoke to three companies that
are experimenting with an alternative method, a potential solution we've
known about for years, but that is now being made
possible by the improved functionality and lower prices of deep
sea roboots, under sea power cables, and other technologies. Basically,
the old way of desalinating water involved pumping seawater to
(08:07):
the surface and then boiling it to create clean water.
The new way involves an underwater membrane that uses the
ocean's natural pressure to filter out salt before clean water
is then pumped to the surface. It's a method that
could save significant energy and money. Sometimes innovation relies on
simple upgrades, other times it's all about competition. President Donald
(08:33):
Trump's one hundred and seventy five billion dollar plan for
a quote Golden Dome defense system has triggered a race
between tech companies and defense groups. According to Financial Times,
the idea is to create a space based system that
can detect and destroy foreign weapons like missiles at launch.
(08:53):
The Trump administration is called for quote non traditional contractors
to help create a system. Cue the competition between ten
tech companies like Paneteer and Microsoft, and established defense contractors
like Lockheed Martin. This all comes at a time when
big tech companies like Meta are stepping into the world
of developing military technology, and startup defense companies like Anderil
(09:16):
are raising money at huge valuations ahead of plans to
go public. The Missile Defense Agency plans to award ten
year contracts in an open competitive process, and so far
the agency has gotten over five hundred responses to their
request for information. After the break, we'll hear from four
(09:39):
or four Media's Jason Kebler about the dreaded airport pat
down and its future high tech makeover. So for our
next segment, we're going to dive into one of the
(09:59):
more annoying parts travel, especially for anyone who flies frequently,
airport security and the role technology.
Speaker 3 (10:06):
Plays in it.
Speaker 1 (10:07):
In response to nine to eleven, the Transportation Security Administration,
or TSA was born. Ever since, TSA has been part
of the flying experience here in the US, ensuring that
passengers are safe and highly agitated, one plastic bin at
a time. If you're not one of about thirty percent
of American fliers with TSA pre check, you know the
(10:28):
drill put your personal items in a plastic bin, take
off your shoes, put your bins through the conveyor belt
where they will go under a scanner, walk through a
body scanner, and then if you're unlucky or you've forgot
to take a receipt out of your pocket, you might
be pulled aside for an intimate pat down, which can
be pretty awkward. Over the years, there have been multiple
upgrades to TSA technology, from more discerning luggage scanners to
(10:52):
full body scans in place of metal detectors. But the
latest idea from the Department of Homeland Security, well, I
certainly don't see it coming. And here to tell us
how virtual reality could forever alter the dreaded pat down
process is four or four Media's Jason Kebler. Jason, thanks
for coming back to tech stuff.
Speaker 3 (11:10):
Hey, thanks for having me. So.
Speaker 1 (11:12):
I'm always quite fascinated by airport so I spent a
lot of time in them, and so when I saw
your most recent piece with the headline TSA working on
heptic tech to feel your body in virtual reality, I
had to know more. What inspired the piece and what
did you learn?
Speaker 2 (11:28):
Yeah, I mean I think that since nine to eleven,
TSA has been looking for ways to quote unquote improve.
Speaker 3 (11:36):
The security process.
Speaker 2 (11:37):
It's like one day you'll just show up at the
airport and they'll have all new machines and it will
be like a totally different process. And I think that
airplane security and airport security is something that the United
States really likes to spend money on, I mean, for
sort of obvious reasons.
Speaker 3 (11:53):
After something like nine to eleven. There's sort of the.
Speaker 2 (11:56):
Security theater of it all where it's like, please take
off your shoes, We're going to go through your bags.
Sometimes you can bring liquids, sometimes you can't. It really
feels like a roll of the dice. And so TSA
has been basically researching this technology to give their agents
virtual reality goggles and then haptic feedback gloves, and so
(12:18):
haptic feedback is like where you wear a glove and
you can literally like feel in virtual reality. And so
they've really designed this incredibly complex and seemingly like overkill
a piece of technology where instead of doing a pat down,
which notoriously can be very invasive, like that certainly is
(12:41):
a problem, they're saying, well, people don't like those, so
we're going to do a virtual reality pat down by
using these like advanced sensors to detect the outlines of
someone's body and see if they have like sharp objects
that they're trying to smuggle in, and then the TSA
officer will be able to see that in their virtual
(13:02):
reality and also like feel that in their haptic gloves,
but without actually touching you. And so I don't know,
I don't know how you feel about it. I saw
it and I was like, this is this seems like overkill?
And in some ways, it's like the clothes, right, there's that,
and then there's also like nominally, like if someone is
touching you, you can say I don't like that, like
(13:23):
stop doing that. But you know, in this case, they're
literally creating like a three D scan of your body,
and in some ways it's like possibly more invasive. I
don't know, it's it's just like very weird technology. I
think it's kind of uncharted territory.
Speaker 1 (13:40):
What does it mean to feel when we're talking about
haptic tech?
Speaker 2 (13:43):
Yeah, so I've never done it myself. Like this technology
does exist in rudimentary form, but as I understand it,
it's like you're wearing a glove that has electronics in it,
and like let's say you move your hand around, presumably
some air pockets or some thing in there will activate
so that your hand is feeling like a force back
(14:05):
against it. That would simulate you actually touching something. I'm
trying to think of like analogies for what that might be. Like,
I think maybe like a massage chair, Like imagine a
massage chair that.
Speaker 1 (14:19):
You wear, or a video game controller or a steering
wheel that kind of rattles around and like moves your
hands and stuff.
Speaker 3 (14:26):
Yeah, yeah, that's absolutely right.
Speaker 2 (14:28):
So it's that and then the sort of like input
data would come from all these sensors that would like
project an image onto your hand, and then your hand
would be connected to this VR system so it would
track your hand around in virtual reality. And then you know,
(14:48):
basically like put some physics in there where it says
like okay, you're touching a table, you're touching a person,
here's what it should feel like. And that technology does
exist where like handtrack in virtual reality, object tracking in
virtual reality, that sort of thing. So really like the
haptics is the new thing here and sort of making
(15:10):
that happen in real time where a sensor array is
grabbing all of that as you walk through security.
Speaker 1 (15:18):
Use the word wild in your story, which I think
is an appropriate one. You know this obviously a podcast.
Can you give a visual of what this will look like?
I mean the TSA agent wearing a headset, wearing these levels,
because there's also some quite amusing cartoons or like visualizations
that you're able to source in your reporting of what
TSA imagine this will look like.
Speaker 2 (15:38):
Yeah, so I want to stress that a lot of
this comes from patent applications, which are you know, kind
of like it's very very early. We don't know that
if this technology is going to be deployed, whether it
will be deployed, what it will look like when it's deployed.
But in patent schematics and drawings, it's like they sort
(15:59):
of imagine and hey, this is what we think it
could look like. And so in the information sheet, they
have a TSA officer who has a computer strapped to
his face like a VR goggles, and he's holding up
a gloved hand, and then they have this diagram of
(16:19):
a baseball, like they drew a baseball on a table,
and then they have this, uh, they call it a glove,
but it's really like a hand computer like it's like
they took a touchpad and put it on someone's hand,
strapped it their hand, and that is the haptic feedback glove.
Speaker 3 (16:40):
And so.
Speaker 2 (16:42):
It reminded me of that toy I had when I
was a kid with all those little pins, the like
metal pins where you could put your hand or face and.
Speaker 3 (16:50):
It would do an outline.
Speaker 2 (16:52):
Yeah, and so that's kind of what it's like, a
digital recreation of that. They also have a diagram where
a person is trying to smuggle in scissors around their chest,
and so they have like a picture of scissors on
someone's chest, and then they have a picture of their
belt buckle and the hands are like right up on
(17:13):
their waist and the diagram makes sure to say that
it says, quote scan imagery obscured it due to proximity
to private body zone. And yeah, so this technology would
like somehow obscure I guess, your private areas because they're
(17:34):
conscious of your privacy.
Speaker 1 (17:36):
But the idea is like you will stand facing a
TSA agent. They'll be wearing a virtual reality headset, running
their hands up and down your body but not actually
touching you, and then both feeling on their hands and
seeing in VR objects that maybe underneath your clothes.
Speaker 3 (17:55):
That's what they're proposing.
Speaker 2 (17:57):
And it certainly sounds like the TSA officer could be
right next to you, but they also make it sound
like they could be like in another room. For example,
you'd walk through some sort of sensor system that would
like scan your body in real time, and then the
officer either could be there and would be like, oh,
I'm not touching you. I'm not touching you, but they
(18:18):
could also be in another room doing this sort of
without you even knowing.
Speaker 1 (18:23):
So how did you hear about this? And do you
have the sense this is something that would actually be
seeing in airports anytime soon?
Speaker 3 (18:30):
Yeah, So it's interesting. The Department of Homeland.
Speaker 2 (18:33):
Security, which oversees TSA, actually published a two page information
sheet about this on its website. And I mean, I
guess this is something that they do periodically where they're like,
we have new research, come check it out, but they
publish it like pretty deep on their website. And frankly,
we have like a lot of nerds who read our
articles and are constantly scanning for things like this and
(18:56):
know that it's something that we would care about, you know.
We try to write about futuristic and weird technology, and
so one of our readers sent it to us. I
checked it out, and from that information sheet it said,
like want to learn more, go check out these patents.
Whether we see this in practice or not, I'm not sure.
(19:17):
It seems pretty early, like I haven't seen video of
this working, like a prototype or anything like that. It
seems like it's more of a concept that they're working
on at the moment and our researching. I also think
that honestly, like since nine to eleven, TSA has kind
of dialed in the security screening process. I don't think
(19:39):
it's anything that anyone enjoys. But we're not waiting for
two hours at the security line anymore like we were
kind of immediately after nine to eleven. As unpleasant as
it can be, it's usually pretty quick. And so I
don't know if we'll ever see this, but it is
tech that they're researching.
Speaker 1 (19:56):
And where does the money come from?
Speaker 2 (19:58):
Is this?
Speaker 1 (19:58):
Like do we know how long it? You know how
much they've spent on it? Like what's the kind of
the structural backdrop of this type of project.
Speaker 2 (20:05):
Yeah, So the Department of Homeland Security has an Office
of Science and Technology, but they don't say here's how
much money we spent on this specific virtual reality haptic
feedback remote sensing.
Speaker 1 (20:18):
Is the Nintendo we of apod security basically, right.
Speaker 2 (20:23):
Yeah, but I think it is important to sort of
realize that we're in an era where a lot of
government funding is being slashed for new technologies. And I'm
not saying that DHS shouldn't be researching new technology, Like
a lot of new technology comes out of agencies like
the Department of Homeland Security, like the CIA, like the NSSA.
(20:45):
But at the same time, like this feels like something
that's a solution in search of a problem, I would say,
And so yeah, it's something to consider. Like we're slashing
budgets for science across the government, but we're not really
slashing budgets for the Department of Homeland Security.
Speaker 3 (21:02):
Were increasing that budget if anything.
Speaker 1 (21:05):
Yeah, I mean I did see on your reporting that
I think they filed the patents for this technology in
twenty twenty two, so I guess under the Biden.
Speaker 3 (21:12):
Administration they've been working on it for a while.
Speaker 1 (21:15):
I mean, this is not like a recent fancy, so
that's kind of interesting. I Mean, some people say the
whole VR is a solution in search of a problem, right,
But like, what's the wider arc of leading edge tech
for the miniature industrial complex.
Speaker 2 (21:32):
Yeah, it's interesting because VR and the metaverse, which you know,
associated technology, has been a massive flop.
Speaker 3 (21:41):
Like it was the.
Speaker 2 (21:43):
AI before there was AI, it was crypto before it
was crypto. Like that, there was tons and tons of
hype about VR being just around the corner and people
using it for all sorts of things, whether to work,
to play, to game. But interestingly, like one of the
places where VR has actually been useful has been on
the job training and for things like post traumatic stress disorder,
(22:06):
exposure therapy for soldiers and things like that, and so
the military actually has found some uses for virtual reality.
At the same time, DHS has been really interested in
using virtual reality to kind of like see through things
(22:27):
for lack of a better term, Like there was this
project that Customs and Border Patrol tried to do where
they wanted to use virtual reality goggles in a way
similar to this to see through boxes in order to
determine whether like counterfeit goods were being brought into the country.
As far as I know, they're not doing anything like that.
That was a project that has been around for many
(22:50):
years and was never deployed. There's been university research that
was associated with the Department of Homeland Security where they
wanted to use VR at the border to see terrorists,
is what they said, but they never explained. Yeah, they
never explained like how that would work. But I do
think that there was a period that is still going
(23:11):
on but has largely been taken over by AI, where
you would take any process that the government would do
or that any business would do, and say, well, how
can we add VR to this Because it's sort of
the hyped new technology. We can seem like we're forward looking,
and we can also maybe get some money to research something.
Speaker 3 (23:30):
That makes sense.
Speaker 1 (23:31):
Are there any other interesting airport tech stories that you've
come across recently?
Speaker 2 (23:37):
I need to do more reporting on this, but facial
recognition is very very common at airports now. Global Entry,
which is a customs and border patrol system where you
don't need to show your passport if you've been pre
vetted and are an American citizen, now uses facial recognition
where you just get off the plane and they detect
who you are and they say, welcome back to the
(23:57):
United States. A lot of air lines are using facial
recognition to just boord planes so you don't need to
show a boarding pass. And then TSA is also using
facial recognition at the screenings, and these are all things
that you can opt in or opt out of, but
it's becoming a lot more commonplace. There was no real
(24:18):
big announcement where it was like, hey, we're going to
be using.
Speaker 3 (24:21):
Facial recognition all over the airport. It was just sort
of there one day, And like, I think there needs to.
Speaker 2 (24:27):
Be more reporting on sort of where the technology came from,
like where the initial photos that identify you are coming from,
whether those images are being retained, that sort of thing.
But this is all being done sort of in the
name of convenience and streamlining the process. And I mean,
I'll admit it is a lot faster to do some
(24:49):
of these things with facial recognition. But then you start
worrying about like is that information being shared, how is
it being shared, who's it going to, what privacy guardrails
are there, And that's not something we know a lot
about unfortunately.
Speaker 1 (25:03):
Well, so the airport is a kind of interesting tech
testing ground or petri dish because like the contract you
sign implicitly or explicitly when you go to the airport
is basically that you surrender all of your rights without
any questions asked. Right, So, once you're like going through security,
you kind of accepted the premise that the Apple can
do whatever it wants to you, right, and that includes
(25:25):
rolling out technologies that you may technically consent to, but
in reality like very hard to withhold your consent from exactly.
Speaker 2 (25:32):
I mean, the only way to withhold your consent in
some cases is to just not fly. And I think
also airplanes and airports in general are interesting places where
technology is rolled out. Because it's very expensive to fly,
you sort of have this self selecting group of people
who go through a security process. They're not bringing guns,
(25:53):
they're not bringing knives. Like it's very safe. It's like
incredibly safe to be in an airport. And therefore you
have like a kind of Petri dish of like transient
people who are who are only there for a few hours.
You can test new technologies on them. People are bored,
and so they might say like, oh, here's like this
new VR game or this new new thing that I
(26:14):
can try out while I'm waiting for my flight. So
like I've seen a lot of like new gaming tech
at at airports and things like that. And I'm not
saying that it's all surveillance. I'm just saying that, like
it feels like new technologies are like rolled out there
before they're rolled out more broadly into society.
Speaker 1 (26:33):
You know, we've reading these stories about how Newark Airport
is running on floppy disks and there are these regular
outages where air traffic control can't even see the planes.
I think that's why I find it's one so fascinating,
because it's like it's a perfect place where you have
technology being used as a solution to a place where
there is no problem, and then huge real problems that
(26:54):
no one's interested in solutions for seeming Yeah.
Speaker 2 (26:57):
Yeah, and air traffic control problem is one of just
like standard human labor practices where it's like we're not
training enough air traffic controllers, we're not paying them enough,
they're very stressed. It's an incredibly important job where these
stakes couldn't possibly be higher, and so you have this
like natural human burnout. And then you also have this
(27:19):
situation where like handing over something like that to artificial
intelligence or to like machine vision, or like you could
see that being more efficient, but the technology clearly is
not ready and a mistake is a life and death situation,
and so you have like maybe some technological solutions that
(27:41):
could come through there, but you're sort of budding up
against as you say, like these really like old fashioned
problems of funding and treating workers correctly and you know,
the pipeline of training them and that sort of thing.
And so, I mean, I think that's a great observation.
Speaker 1 (28:00):
Mail Re reported your story without crediting you, of course,
but has this story traveled far and wide and what
the response been.
Speaker 2 (28:08):
It's been a lot of people saying this reminds me
of teledildonics, which is technology that has been created to
allow people in long distance relationships to have cyber sex
while actually feeling it. And I mean it relies on
a lot of these same technologies that we've been talking about.
(28:28):
And so I mean, we actually do see this time
and time again where sex tech and porn industry is
quite ahead of where society is going. And you know
that technology has been being worked on for.
Speaker 3 (28:43):
Years by.
Speaker 2 (28:46):
Let's say enterprising di I wires and now you have
like similar technology being looked at by Department of Homeland Security,
and so that's been a lot of the response so far.
Speaker 1 (29:01):
Jason, appreciate taking the time this week.
Speaker 3 (29:03):
Yeah, thank you so much. This is fine.
Speaker 1 (29:18):
That's it for this week for tech Stuff. I'm os Voloshin.
This episode was produced by Eliza Dennis and Victoria Domingez.
It was executive produced by me Kara Price and Kate
Osborne for Kaleidoscope and Katrina Norvel for iHeart Podcasts but
he phrased as I Engineer. Jack Insley mixed this episode
and Kyle Murdoch wrote our theme song. Join us next
(29:40):
Wednesday for an episode all about biometric data, from shopping
with the palm of your hand to donning multiple wearables.
How much should we really care about giving away all
that personal data? Please rate, review and reach out to
us at tech Stuff podcast at gmail dot com with
your feedback, with your story ideas, with whatever you want
(30:00):
to tell us, because we love hearing from you.
Welcome to Tech Stuff, a production of iHeart Podcasts and Kaleidoscope.
I'ma's Voloshan and carac prices out this week, so I'll
be bringing you the headlines of the week, including AI's
black box problem and open Eyes push to infiltrate college campuses.
And on today's Tech Support segment, we'll talk to four
or four Media's Jason Kebler about the weird world of
(00:35):
body scans and the future of TSA security.
Speaker 2 (00:39):
Airplanes and airports in general are interesting because it feels
like new technologies are like rolled out there before they're
rolled out more broadly into society.
Speaker 1 (00:50):
All of that On the weekend Tech It's a Spooky Day, Friday,
June thirteenth. As I mentioned, Carras out this week, so
we're going to dive right into the headlines, and a
couple of big stories have been on my mind, both
highlighting that as AI becomes more and more ubiquitous and
(01:11):
as a rush to deploy it, we still don't fully
understand how these systems actually work. AI models are such
a black box that even the developers who build them
can't always explain or predict the behavior of their models,
And this week, Axios released a detailed roundup of how
leaders are simultaneously admitting this lack of knowledge while also
(01:33):
pushing for rapid deployment and implementation. Here's an example. Anthropic,
while testing its latest model, Clawed four, gave the model
access to corporate emails, setting up this fictional scenario where
Claude knew that one it was going to be shut
down and two that the engineer responsible for shutting it
(01:54):
down was having an affair. In eighty four percent of
cases when this test was run, the model attempted to
blackmail the engineer, supposedly to preserve itself. And here's the thing.
Anthropic doesn't know how the new model chose blackmail as
a tactic or how to prevent it from doing so
(02:15):
in future, but Anthropic released the model. Now there is
an accepted term to refer to this problem I that
we don't yet understand the why of AI's behavior, and
that term is interpretability. CEOs of AI companies like open
Ai Samultman and Anthropics Dario Amode have openly acknowledged that
(02:36):
quote interpretability is an important problem to solve and that
our lack of understanding about how these models work can
pose significant risk in April, A. Mooday wrote that quote
people outside the field are often surprised and alarmed to
learn that we do not understand how our AI creations work.
They are right to be concerned. This lack of understanding
(02:58):
is essentially unprecedented in the history of technology. The question
to me is is there enough investment and attention going
towards this problem versus the race forever stronger performance between
privately funded companies. AI is being developed and implemented without
meaningful guardrails, and that's something that US government is actively encouraging.
(03:19):
In fact, President Trump's Big Beautiful Bill includes a provision
for a decade long ban on states attempting to regulate AI,
either through new regulations or through enforcing existing regulations. Now
the bill has been passed by the House and is
being heard in the Senate. Changes are very possible because
even Republican senators like Josh Hawley of Missouri and Marshall
(03:43):
Blackburn of Tennessee have pushed back on this ban. Now,
for those hoping humans will maintain control over AI, there
is some potentially encouraging news coming out of Apple This week.
The company's machine learning researchers publish an academic paper with
very buzzy title, The Illusion of Thinking. The paper studies
(04:05):
reasoning models like open Aiyes three and deep seeks are one,
which are models designed to problem solve. What the Apple
paper found is that when these reasoning models are presented
with complex logic problems, the models fail to solve the problems.
Mashable reports that in the study, reasoning models were given
classic logic puzzles like jumping checker pieces into empty spaces,
(04:27):
or the river crossing problem, the one involving a fox,
a chicken, and a bag of grain. This is a
pretty simple test of a human's ability to problem solve,
because once you figure out the rules, it's actually pretty
easy to continue solving these problems even when they get
more complex. But these reasoning models start to fail at
a certain point. So the big question is is this
(04:49):
simply a bump in the road or is it indicative
of a larger problem in how these reasoning models actually work.
Which brings me to my next headline from the New
York Times, Welcome to campus, here's your chat GPT. Because
if critical thinking and logical reasoning skills are where humans
still have an edge over AI, it's pretty reasonable to me.
(05:11):
There are concerns about the rise of AI on college campuses.
Apparently Open Ai has a plan to integrate chatbots into
every facet of student life, and they're calling the goal
of this AI integrated higher ed quote AI Native Universities.
A suite of premium services called chat gpt Edu is
(05:31):
being sold to universities for faculty and students to use,
and the company is promising AI tutors and chatbots that
can do everything from conduct mock job interviews to quiz
you before exams. Schools are getting in on the action.
California State University is making chatchipt available to more than
four hundred and sixty thousand students. Duke University made its
(05:54):
own AI powered platform called Duke gpt, and in June
they began offering unlimited Chatchipt access to students and faculty.
Even if a campus isn't striking a deal with open Ai.
There's a marketing campaign targeting the handful and I assume
it is a handful of students who aren't yet using chatchypt.
(06:16):
Many find this all deeply troubling, especially as there's new
data and research to suggest that so called cognitive offloading
I letting a chatbot write your first draft or think
for you makes you measurably less able to problem solve yourself.
On the other hand, recent grads are facing historically tough
(06:37):
job market, especially in fields where AI is starting to
automate tasks like writing code. So these schools are trying
to boost their students' prospects by providing them with AI
tools and with the skills to make the most of
the tools. So is this strange moment where we don't
really understand how our lambs make their decisions or how
using them affects our brains. But they're here, We're using
(07:00):
them anyway, and we and they are adapting in real time.
As The Times puts it, quote open ayes pushed to
AIIFI college education amounts to a national experiment on millions
of students. I've got a couple more headlines for you,
(07:22):
and one starts with an old idea brought to life
with modern technology. About half of all people on Earth
experience some form of water scarcity. But could that change
if you drink the ocean. Desalination, the process of removing
salt to create clean drinking water, has been possible for decades,
but it's the most expensive way to create clean drinking water. However,
(07:45):
The Wall Street Journal recently spoke to three companies that
are experimenting with an alternative method, a potential solution we've
known about for years, but that is now being made
possible by the improved functionality and lower prices of deep
sea roboots, under sea power cables, and other technologies. Basically,
the old way of desalinating water involved pumping seawater to
(08:07):
the surface and then boiling it to create clean water.
The new way involves an underwater membrane that uses the
ocean's natural pressure to filter out salt before clean water
is then pumped to the surface. It's a method that
could save significant energy and money. Sometimes innovation relies on
simple upgrades, other times it's all about competition. President Donald
(08:33):
Trump's one hundred and seventy five billion dollar plan for
a quote Golden Dome defense system has triggered a race
between tech companies and defense groups. According to Financial Times,
the idea is to create a space based system that
can detect and destroy foreign weapons like missiles at launch.
(08:53):
The Trump administration is called for quote non traditional contractors
to help create a system. Cue the competition between ten
tech companies like Paneteer and Microsoft, and established defense contractors
like Lockheed Martin. This all comes at a time when
big tech companies like Meta are stepping into the world
of developing military technology, and startup defense companies like Anderil
(09:16):
are raising money at huge valuations ahead of plans to
go public. The Missile Defense Agency plans to award ten
year contracts in an open competitive process, and so far
the agency has gotten over five hundred responses to their
request for information. After the break, we'll hear from four
(09:39):
or four Media's Jason Kebler about the dreaded airport pat
down and its future high tech makeover. So for our
next segment, we're going to dive into one of the
(09:59):
more annoying parts travel, especially for anyone who flies frequently,
airport security and the role technology.
Speaker 3 (10:06):
Plays in it.
Speaker 1 (10:07):
In response to nine to eleven, the Transportation Security Administration,
or TSA was born. Ever since, TSA has been part
of the flying experience here in the US, ensuring that
passengers are safe and highly agitated, one plastic bin at
a time. If you're not one of about thirty percent
of American fliers with TSA pre check, you know the
(10:28):
drill put your personal items in a plastic bin, take
off your shoes, put your bins through the conveyor belt
where they will go under a scanner, walk through a
body scanner, and then if you're unlucky or you've forgot
to take a receipt out of your pocket, you might
be pulled aside for an intimate pat down, which can
be pretty awkward. Over the years, there have been multiple
upgrades to TSA technology, from more discerning luggage scanners to
(10:52):
full body scans in place of metal detectors. But the
latest idea from the Department of Homeland Security, well, I
certainly don't see it coming. And here to tell us
how virtual reality could forever alter the dreaded pat down
process is four or four Media's Jason Kebler. Jason, thanks
for coming back to tech stuff.
Speaker 3 (11:10):
Hey, thanks for having me. So.
Speaker 1 (11:12):
I'm always quite fascinated by airport so I spent a
lot of time in them, and so when I saw
your most recent piece with the headline TSA working on
heptic tech to feel your body in virtual reality, I
had to know more. What inspired the piece and what
did you learn?
Speaker 2 (11:28):
Yeah, I mean I think that since nine to eleven,
TSA has been looking for ways to quote unquote improve.
Speaker 3 (11:36):
The security process.
Speaker 2 (11:37):
It's like one day you'll just show up at the
airport and they'll have all new machines and it will
be like a totally different process. And I think that
airplane security and airport security is something that the United
States really likes to spend money on, I mean, for
sort of obvious reasons.
Speaker 3 (11:53):
After something like nine to eleven. There's sort of the.
Speaker 2 (11:56):
Security theater of it all where it's like, please take
off your shoes, We're going to go through your bags.
Sometimes you can bring liquids, sometimes you can't. It really
feels like a roll of the dice. And so TSA
has been basically researching this technology to give their agents
virtual reality goggles and then haptic feedback gloves, and so
(12:18):
haptic feedback is like where you wear a glove and
you can literally like feel in virtual reality. And so
they've really designed this incredibly complex and seemingly like overkill
a piece of technology where instead of doing a pat down,
which notoriously can be very invasive, like that certainly is
(12:41):
a problem, they're saying, well, people don't like those, so
we're going to do a virtual reality pat down by
using these like advanced sensors to detect the outlines of
someone's body and see if they have like sharp objects
that they're trying to smuggle in, and then the TSA
officer will be able to see that in their virtual
(13:02):
reality and also like feel that in their haptic gloves,
but without actually touching you. And so I don't know,
I don't know how you feel about it. I saw
it and I was like, this is this seems like overkill?
And in some ways, it's like the clothes, right, there's that,
and then there's also like nominally, like if someone is
touching you, you can say I don't like that, like
(13:23):
stop doing that. But you know, in this case, they're
literally creating like a three D scan of your body,
and in some ways it's like possibly more invasive. I
don't know, it's it's just like very weird technology. I
think it's kind of uncharted territory.
Speaker 1 (13:40):
What does it mean to feel when we're talking about
haptic tech?
Speaker 2 (13:43):
Yeah, so I've never done it myself. Like this technology
does exist in rudimentary form, but as I understand it,
it's like you're wearing a glove that has electronics in it,
and like let's say you move your hand around, presumably
some air pockets or some thing in there will activate
so that your hand is feeling like a force back
(14:05):
against it. That would simulate you actually touching something. I'm
trying to think of like analogies for what that might be. Like,
I think maybe like a massage chair, Like imagine a
massage chair that.
Speaker 1 (14:19):
You wear, or a video game controller or a steering
wheel that kind of rattles around and like moves your
hands and stuff.
Speaker 3 (14:26):
Yeah, yeah, that's absolutely right.
Speaker 2 (14:28):
So it's that and then the sort of like input
data would come from all these sensors that would like
project an image onto your hand, and then your hand
would be connected to this VR system so it would
track your hand around in virtual reality. And then you know,
(14:48):
basically like put some physics in there where it says
like okay, you're touching a table, you're touching a person,
here's what it should feel like. And that technology does
exist where like handtrack in virtual reality, object tracking in
virtual reality, that sort of thing. So really like the
haptics is the new thing here and sort of making
(15:10):
that happen in real time where a sensor array is
grabbing all of that as you walk through security.
Speaker 1 (15:18):
Use the word wild in your story, which I think
is an appropriate one. You know this obviously a podcast.
Can you give a visual of what this will look like?
I mean the TSA agent wearing a headset, wearing these levels,
because there's also some quite amusing cartoons or like visualizations
that you're able to source in your reporting of what
TSA imagine this will look like.
Speaker 2 (15:38):
Yeah, so I want to stress that a lot of
this comes from patent applications, which are you know, kind
of like it's very very early. We don't know that
if this technology is going to be deployed, whether it
will be deployed, what it will look like when it's deployed.
But in patent schematics and drawings, it's like they sort
(15:59):
of imagine and hey, this is what we think it
could look like. And so in the information sheet, they
have a TSA officer who has a computer strapped to
his face like a VR goggles, and he's holding up
a gloved hand, and then they have this diagram of
(16:19):
a baseball, like they drew a baseball on a table,
and then they have this, uh, they call it a glove,
but it's really like a hand computer like it's like
they took a touchpad and put it on someone's hand,
strapped it their hand, and that is the haptic feedback glove.
Speaker 3 (16:40):
And so.
Speaker 2 (16:42):
It reminded me of that toy I had when I
was a kid with all those little pins, the like
metal pins where you could put your hand or face and.
Speaker 3 (16:50):
It would do an outline.
Speaker 2 (16:52):
Yeah, and so that's kind of what it's like, a
digital recreation of that. They also have a diagram where
a person is trying to smuggle in scissors around their chest,
and so they have like a picture of scissors on
someone's chest, and then they have a picture of their
belt buckle and the hands are like right up on
(17:13):
their waist and the diagram makes sure to say that
it says, quote scan imagery obscured it due to proximity
to private body zone. And yeah, so this technology would
like somehow obscure I guess, your private areas because they're
(17:34):
conscious of your privacy.
Speaker 1 (17:36):
But the idea is like you will stand facing a
TSA agent. They'll be wearing a virtual reality headset, running
their hands up and down your body but not actually
touching you, and then both feeling on their hands and
seeing in VR objects that maybe underneath your clothes.
Speaker 3 (17:55):
That's what they're proposing.
Speaker 2 (17:57):
And it certainly sounds like the TSA officer could be
right next to you, but they also make it sound
like they could be like in another room. For example,
you'd walk through some sort of sensor system that would
like scan your body in real time, and then the
officer either could be there and would be like, oh,
I'm not touching you. I'm not touching you, but they
(18:18):
could also be in another room doing this sort of
without you even knowing.
Speaker 1 (18:23):
So how did you hear about this? And do you
have the sense this is something that would actually be
seeing in airports anytime soon?
Speaker 3 (18:30):
Yeah, So it's interesting. The Department of Homeland.
Speaker 2 (18:33):
Security, which oversees TSA, actually published a two page information
sheet about this on its website. And I mean, I
guess this is something that they do periodically where they're like,
we have new research, come check it out, but they
publish it like pretty deep on their website. And frankly,
we have like a lot of nerds who read our
articles and are constantly scanning for things like this and
(18:56):
know that it's something that we would care about, you know.
We try to write about futuristic and weird technology, and
so one of our readers sent it to us. I
checked it out, and from that information sheet it said,
like want to learn more, go check out these patents.
Whether we see this in practice or not, I'm not sure.
(19:17):
It seems pretty early, like I haven't seen video of
this working, like a prototype or anything like that. It
seems like it's more of a concept that they're working
on at the moment and our researching. I also think
that honestly, like since nine to eleven, TSA has kind
of dialed in the security screening process. I don't think
(19:39):
it's anything that anyone enjoys. But we're not waiting for
two hours at the security line anymore like we were
kind of immediately after nine to eleven. As unpleasant as
it can be, it's usually pretty quick. And so I
don't know if we'll ever see this, but it is
tech that they're researching.
Speaker 1 (19:56):
And where does the money come from?
Speaker 2 (19:58):
Is this?
Speaker 1 (19:58):
Like do we know how long it? You know how
much they've spent on it? Like what's the kind of
the structural backdrop of this type of project.
Speaker 2 (20:05):
Yeah, So the Department of Homeland Security has an Office
of Science and Technology, but they don't say here's how
much money we spent on this specific virtual reality haptic
feedback remote sensing.
Speaker 1 (20:18):
Is the Nintendo we of apod security basically, right.
Speaker 2 (20:23):
Yeah, but I think it is important to sort of
realize that we're in an era where a lot of
government funding is being slashed for new technologies. And I'm
not saying that DHS shouldn't be researching new technology, Like
a lot of new technology comes out of agencies like
the Department of Homeland Security, like the CIA, like the NSSA.
(20:45):
But at the same time, like this feels like something
that's a solution in search of a problem, I would say,
And so yeah, it's something to consider. Like we're slashing
budgets for science across the government, but we're not really
slashing budgets for the Department of Homeland Security.
Speaker 3 (21:02):
Were increasing that budget if anything.
Speaker 1 (21:05):
Yeah, I mean I did see on your reporting that
I think they filed the patents for this technology in
twenty twenty two, so I guess under the Biden.
Speaker 3 (21:12):
Administration they've been working on it for a while.
Speaker 1 (21:15):
I mean, this is not like a recent fancy, so
that's kind of interesting. I Mean, some people say the
whole VR is a solution in search of a problem, right,
But like, what's the wider arc of leading edge tech
for the miniature industrial complex.
Speaker 2 (21:32):
Yeah, it's interesting because VR and the metaverse, which you know,
associated technology, has been a massive flop.
Speaker 3 (21:41):
Like it was the.
Speaker 2 (21:43):
AI before there was AI, it was crypto before it
was crypto. Like that, there was tons and tons of
hype about VR being just around the corner and people
using it for all sorts of things, whether to work,
to play, to game. But interestingly, like one of the
places where VR has actually been useful has been on
the job training and for things like post traumatic stress disorder,
(22:06):
exposure therapy for soldiers and things like that, and so
the military actually has found some uses for virtual reality.
At the same time, DHS has been really interested in
using virtual reality to kind of like see through things
(22:27):
for lack of a better term, Like there was this
project that Customs and Border Patrol tried to do where
they wanted to use virtual reality goggles in a way
similar to this to see through boxes in order to
determine whether like counterfeit goods were being brought into the country.
As far as I know, they're not doing anything like that.
That was a project that has been around for many
(22:50):
years and was never deployed. There's been university research that
was associated with the Department of Homeland Security where they
wanted to use VR at the border to see terrorists,
is what they said, but they never explained. Yeah, they
never explained like how that would work. But I do
think that there was a period that is still going
(23:11):
on but has largely been taken over by AI, where
you would take any process that the government would do
or that any business would do, and say, well, how
can we add VR to this Because it's sort of
the hyped new technology. We can seem like we're forward looking,
and we can also maybe get some money to research something.
Speaker 3 (23:30):
That makes sense.
Speaker 1 (23:31):
Are there any other interesting airport tech stories that you've
come across recently?
Speaker 2 (23:37):
I need to do more reporting on this, but facial
recognition is very very common at airports now. Global Entry,
which is a customs and border patrol system where you
don't need to show your passport if you've been pre
vetted and are an American citizen, now uses facial recognition
where you just get off the plane and they detect
who you are and they say, welcome back to the
(23:57):
United States. A lot of air lines are using facial
recognition to just boord planes so you don't need to
show a boarding pass. And then TSA is also using
facial recognition at the screenings, and these are all things
that you can opt in or opt out of, but
it's becoming a lot more commonplace. There was no real
(24:18):
big announcement where it was like, hey, we're going to
be using.
Speaker 3 (24:21):
Facial recognition all over the airport. It was just sort
of there one day, And like, I think there needs to.
Speaker 2 (24:27):
Be more reporting on sort of where the technology came from,
like where the initial photos that identify you are coming from,
whether those images are being retained, that sort of thing.
But this is all being done sort of in the
name of convenience and streamlining the process. And I mean,
I'll admit it is a lot faster to do some
(24:49):
of these things with facial recognition. But then you start
worrying about like is that information being shared, how is
it being shared, who's it going to, what privacy guardrails
are there, And that's not something we know a lot
about unfortunately.
Speaker 1 (25:03):
Well, so the airport is a kind of interesting tech
testing ground or petri dish because like the contract you
sign implicitly or explicitly when you go to the airport
is basically that you surrender all of your rights without
any questions asked. Right, So, once you're like going through security,
you kind of accepted the premise that the Apple can
do whatever it wants to you, right, and that includes
(25:25):
rolling out technologies that you may technically consent to, but
in reality like very hard to withhold your consent from exactly.
Speaker 2 (25:32):
I mean, the only way to withhold your consent in
some cases is to just not fly. And I think
also airplanes and airports in general are interesting places where
technology is rolled out. Because it's very expensive to fly,
you sort of have this self selecting group of people
who go through a security process. They're not bringing guns,
(25:53):
they're not bringing knives. Like it's very safe. It's like
incredibly safe to be in an airport. And therefore you
have like a kind of Petri dish of like transient
people who are who are only there for a few hours.
You can test new technologies on them. People are bored,
and so they might say like, oh, here's like this
new VR game or this new new thing that I
(26:14):
can try out while I'm waiting for my flight. So
like I've seen a lot of like new gaming tech
at at airports and things like that. And I'm not
saying that it's all surveillance. I'm just saying that, like
it feels like new technologies are like rolled out there
before they're rolled out more broadly into society.
Speaker 1 (26:33):
You know, we've reading these stories about how Newark Airport
is running on floppy disks and there are these regular
outages where air traffic control can't even see the planes.
I think that's why I find it's one so fascinating,
because it's like it's a perfect place where you have
technology being used as a solution to a place where
there is no problem, and then huge real problems that
(26:54):
no one's interested in solutions for seeming Yeah.
Speaker 2 (26:57):
Yeah, and air traffic control problem is one of just
like standard human labor practices where it's like we're not
training enough air traffic controllers, we're not paying them enough,
they're very stressed. It's an incredibly important job where these
stakes couldn't possibly be higher, and so you have this
like natural human burnout. And then you also have this
(27:19):
situation where like handing over something like that to artificial
intelligence or to like machine vision, or like you could
see that being more efficient, but the technology clearly is
not ready and a mistake is a life and death situation,
and so you have like maybe some technological solutions that
(27:41):
could come through there, but you're sort of budding up
against as you say, like these really like old fashioned
problems of funding and treating workers correctly and you know,
the pipeline of training them and that sort of thing.
And so, I mean, I think that's a great observation.
Speaker 1 (28:00):
Mail Re reported your story without crediting you, of course,
but has this story traveled far and wide and what
the response been.
Speaker 2 (28:08):
It's been a lot of people saying this reminds me
of teledildonics, which is technology that has been created to
allow people in long distance relationships to have cyber sex
while actually feeling it. And I mean it relies on
a lot of these same technologies that we've been talking about.
(28:28):
And so I mean, we actually do see this time
and time again where sex tech and porn industry is
quite ahead of where society is going. And you know
that technology has been being worked on for.
Speaker 3 (28:43):
Years by.
Speaker 2 (28:46):
Let's say enterprising di I wires and now you have
like similar technology being looked at by Department of Homeland Security,
and so that's been a lot of the response so far.
Speaker 1 (29:01):
Jason, appreciate taking the time this week.
Speaker 3 (29:03):
Yeah, thank you so much. This is fine.
Speaker 1 (29:18):
That's it for this week for tech Stuff. I'm os Voloshin.
This episode was produced by Eliza Dennis and Victoria Domingez.
It was executive produced by me Kara Price and Kate
Osborne for Kaleidoscope and Katrina Norvel for iHeart Podcasts but
he phrased as I Engineer. Jack Insley mixed this episode
and Kyle Murdoch wrote our theme song. Join us next
(29:40):
Wednesday for an episode all about biometric data, from shopping
with the palm of your hand to donning multiple wearables.
How much should we really care about giving away all
that personal data? Please rate, review and reach out to
us at tech Stuff podcast at gmail dot com with
your feedback, with your story ideas, with whatever you want
(30:00):
to tell us, because we love hearing from you.
TechStuff News
Hosts And Creators
Oz Woloshyn
Karah Preiss
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
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.
Ridiculous History
History is beautiful, brutal and, often, ridiculous. Join Ben Bowlin and Noel Brown as they dive into some of the weirdest stories from across the span of human civilization in Ridiculous History, a podcast by iHeartRadio.