Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:04):
Welcome to Tech Stuff, a production from iHeartRadio. Hey therein
Welcome to Tech Stuff. I'm your host, Jonathan Strickland. I'm
an executive producer at iHeart Podcasts. And how the tech
are you. It's time for the tech news for the
week ending July nineteenth, twenty twenty four. And the big
(00:26):
story that I have this week is how a product
update for software called Falcon Sensor led to worldwide operating
system failures, which in turn necessitated that important businesses businesses
like airlines, for example, had to shut down or at
least delay operations while folks tried to fix the problem.
(00:48):
The heart of the matter lies with a cybersecurity company
called CrowdStrike. So CrowdStrike pushed out a content update to
customers for the Falcon Sensor product. And Microsoft is such
a customer, and this update went out to Windows virtual machines,
many of which handle mission critical operations for big customers,
(01:11):
including some major companies all over the world. The update
caught the machines into a recovery boot loop, which means
the machines never really loaded in. They would get to
a certain point of booting up and then they would
reboot and they would just keep doing this, and that's
when the failures began. So here in the United States,
Delta Airlines, American Airlines, and United Airlines all grounded their
(01:35):
flights because their systems effectively went down. In other parts
of the world, banks were affected, which severely slowed down
operations there. One stock trader called it the mother of
all global market outages. In the UK, Sky News went
off the air they were unable to broadcast. In Australia,
emergency call centers reported disruptions due to this issue, which
(01:58):
is pretty darn scary. Everyone was plagued by the dreaded
blue screen of death. George Kurt's CEO of CrowdStrike, took
to x formerly known as Twitter, and posted quote, CrowdStrike
is actively working with customers impacted by a defect found
in a single content update for Windows hosts, Mac and
(02:19):
Linux hosts are not impacted. This is not a security
incident or cyber attack. The issue has been identified, isolated,
and a fix has been deployed. End quote. Now, deploying
a fix is great, However, it does not magically stop
the already affected machines from being in that reboot loop cycle.
So to do that it admins will actually have to
(02:42):
stop the reboot cycle. They'll have to boot into essentially
safe mode and then navigate to the CrowdStrike file directory
and delete the appropriate file manually. And depending on the
it situation for each business, that could be a fairly
easy thing to do or a not so easy thing
to do. It's definitely a time consuming thing to do,
(03:04):
but in the meantime, a lot of companies are unable
to function properly, some can't really function at all, and
this mistake really points out how reliant the world is
on a limited number of service providers, and I would
argue it also reinforces to hackers the potential impact of
targeting the supply chain because it lets you hit a
huge number of targets simultaneously and disrupt them. Though again
(03:27):
in this case, we're not talking about a hacker attack.
This was just a mistake that CrowdStrike made that has
affected a lot of people all around the world. And
I bet there's going to be some uncomfortable meetings about
it at CrowdStrike headquarters today, So yeah, yikes. Another obviously
huge story this week was the assassination attempt on the
(03:47):
life of former president and now GOP nominee Donald Trump.
But this is not a politics or crime podcast, so
why would I bring it up. Well, that's because the
FBI took possess of the would be assassin's phone and
within a couple of days announced that they had successfully
accessed the phone's contents, which raised a lot of questions
(04:09):
in the tech community, as there is this ongoing concern
in tech about the extent to which your privacy and
security can be insured by the hardware that you use.
So if someone were to get hold of your device,
would they be able to access it without your direct aid?
And how did the FBI manage this? While the phone
(04:30):
in question is a Samsung device and it was locked
when the FBI received it. Bloomberg reported that the FBI
initially couldn't crack the phone's security, but then they worked
with an Israeli digital intelligence company called Celebrate to get access.
Earlier this week, I saw online speculation that the at
that time unnamed Israeli firm was actually the NSO Group.
(04:55):
That's the group that's best known for producing the spyware
on steroids product called Pegasus. But no, these are two
separate cyber intelligence companies. According to Bloomberg, Celebrate used new
software that's actually still in development, it's not offered as
a product yet, and cracked the phone security and gave
FBI access to it. A piece in nine to five,
(05:17):
Mac points out that a data leak from within Celebrate
revealed its software is currently unable to crack iPhone security
at least on any devices running iOS seventeen point four
or later. So this just shows a gap between their
ability to crack security in iOS platforms versus Android platforms.
(05:39):
And yeah, this is like an ongoing discussion in the
tech community. Is you know which operating system provides better security?
Can you expect to keep your stuff on lock even
if it were to fall into the hands of, say
the FBI. Like, there are a lot of conversations in
tech that follow that, which is why I covered the story.
(05:59):
On Thursday, open ai unveiled a slimmed down AI Large
Language Model, So I guess a not so large language
model or LM, and it's called fittingly enough GPT four
to oh mini oh in this case is the lowercase
letter O not the number zero or the numeral zero
or however you want to call it. I'm not a mathematician. Further,
(06:21):
this model is slated to take over from GPT three
point five Turbo for the purposes of powering chat GPT,
so it will be the language model behind the chatbot.
It's currently available for consumer chat GPT users. It will
later spread to enterprise customers next week. For now, it's
similar to earlier large language models, but open ai says
(06:45):
this one will eventually be able to analyze and interpret
images as well as generate its own so not just text,
plus it will also be able to analyze audio. It
will not, however, have access to the latest breaking news.
It will have a knowledge cutoff of October twenty twenty three.
So what's the benefit of moving to a smaller large
(07:05):
language model. Well, the big one is cost. They are
less expensive to operate and thus less expensive for customers.
They're still fairly robust, though they aren't as capable of
performing in depth analysis as their bigger siblings. They're good
for smaller and more niche oriented tasks, so the sort
(07:25):
of stuff you would probably encounter in your typical apps
that tap into generative AI. Because your apps usually aren't
do everything apps, they usually are keyed to a specific
subset of features. So, for example, let's say you've made
an app that is essentially a dynamic to do list. Well,
you won't need access to a massive, large language model
(07:47):
to handle any AI component of this implementation, right, because
that would be overkill. Like, you don't need something that's
great at everything, or even passable at everything. You need
something that's good at, you know, a very small set
of tasks. But on the flip side, if you did
need some heavy lifting on the AI front, smaller models
(08:08):
aren't necessarily going to be as good of a fit
because they just won't be capable of performing that in
depth analysis. You would end up spending more time and
possibly more money to achieve the results you actually wanted,
So there are trade offs. Plus, you are also still
beholden to whatever material was used to train the model
in the first place. That doesn't matter about the size.
(08:29):
In fact, like whether it's a small model or a
large model, the training material needs to be good. If
the training material is garbage, that's what you're going to
get on the other end, right, garbage in garbage out.
I talked about this previously this week on tech Stuff
about how bad training material leads to bad AI performance,
and there's a real fear that AI models of the
(08:50):
future are going to be training off of AI generated
content from today, and that will lead to what is
called model collapse. That's a real thing. Listen to earlier
episodes this week to learn more about that. Ours Technica's
Ashley Bellinger has an article titled Elon Musk's X may
(09:11):
succeed in blocking California content moderation law on appeal and
I would like to touch on this. The matter at
hand concerns a law called AB five to eight seven
in California, and it is concerned with the terms of
service for social media companies. Specifically, the bill says that
social media companies must make available their terms of service.
They must define clearly what is and is not permitted
(09:35):
under those terms, and then create a record of any
and all cases in which users or their content are
subjected to action from the company for violation of those terms,
such as when a company deletes a post or bans
a user, as well as a record of exactly what
violation occurred. X's law team argue that the law violates
(09:55):
the company's First Amendment rights to conduct content moderation. That's
forcing them to make a comment on specific controversial subjects,
and that is against the First Amendment, and you know,
to definitively state why someone or something was banned, which
also sends the message that stuff that is allowed to
stay on the platform is permitted. This is actually a
(10:16):
movement to create backlash against X, and the State of
California is determined to just stir stuff up, so to speak,
by making X essentially say, well, yeah, what this person
said was awful, but it's not against our terms of service,
or we don't like what this person said, so we
banned them. Bellinger points out that the appeals court seems
at least partly sympathetic to X, and I can understand
(10:37):
why I might not agree with stuff going on over
at X, and in fact, I disagree a lot with
stuff going on over at X. But you do get
into a pretty thorny issue when a company has to
first define what is and isn't acceptable. As one judge
pointed out, x's argument seems to be that the State
of California is trying to compel X to express company
views on these controversial subjects that range from stuff like
(10:59):
hates each to foreign political interference, and that compelling someone
to speak does seem like it is running a foul
of first Amendment protections. So does this mean the appeals
court will ultimately find in favor of X. Possibly, but
it's not a done deal yet. Okay, I got a
few more news stories to cover before we get to that.
Let's take a quick break. We're back, so and less
(11:30):
serious but still irritating news. Netflix is increasing the cost
of subscribers who want an AD free experience on the
streaming platform by getting rid of the existing lowest tier
in that in that range, so the monthly fee for
the lowest AD free tier, which limited users to viewing
(11:51):
Netflix on just one device at a time, as well
as viewing content and a maximum resolution of seven hundred
and twenty P. That has been twelve dollars a month,
But now that tier is being ended phased out, and
current customers who are at that level will get shifted
to a different tier, less expensive per month, but it
is AD supported. If they want to move to an
(12:12):
AD free version, they'll have to go up to fifteen
dollars forty nine cents per month to do it. Now.
This is technically a different tier, a different plan. It
supports up to two devices and also a resolution of
ten ADP plus. Customers at this tier are allowed to
download content to a local device, So like if you're
going on a plane or something and you're not going
(12:33):
to have access to Wi Fi, you download the episode
or whatever tier phone and you can watch it then. So,
according to Sharon Harding of Ours Technica, Netflix's strategy is
to get more people to move to this lowest lower tier,
this AD supported tier, because ads are more lucrative as
a revenue generator than subscriptions alone. So their hope is
(12:54):
not to convince people to spend more per month and
move to the new lowest AD free tier, but rather
be content with the fact that they're also going to
have to watch some ads, because that is a more
lucrative revenue generation model. Of course, Netflix isn't the only
streaming service that has recently bumped up prices. That's been
(13:15):
true across the board. We've actually seen, you know, some
streaming platforms bump up prices a couple of times in
the span of a year. A heck, I just got
an email earlier this week about how my Peacock subscription
is getting more expensive and I'm wondering should I even
bother keeping it? I mean, I only use it to
watch like things super rarely, but I think this also
(13:35):
has to be expected because obviously, early on in the
streaming days, the game was all about getting as large
a subscriber base as you possibly could, even if that
meant you were losing money in the process. Like you
had these very attractive Intrey offers, you had a lot
of original material. But you can't operate like this forever.
You know, you do have to find ways to make
(13:57):
a profit or at least to keep shareholders happy, since
that's really what it's all about these days, and that
means you can't just give stuff away for very low
amounts of money. You have to figure out a method
to cover the costs of operation at the very least,
and preferably make a profit. Just Weatherbed of The Verge
(14:17):
has a piece about how Google is cleaning house on
the app front. The company says it will be cracking
down on apps that don't live up to Google's play
Store standards. So for apps that have extremely limited functionality
or content and thus are seen as low value apps,
they could find themselves on the chopping block. So let's
say it's an app and all it is is like
a wallpaper. That's it. That's the whole app. This is
(14:39):
one little thing of wallpaper. That's the sort of thing
that Google says will not survive this change. Google already
has policies in place for apps that are downright malicious
or that introduce security vulnerabilities to hardware, which is a
no brainer. But this new rule says, not only must
your app be safe, it also has to be you know,
kind of good. The new rule take effect at the
(15:01):
end of August. There's a famous saying that is often
attributed to Dustyyevsky, and it goes something like this, the
degree of civilization in a society can be judged by
entering its prisons. Now, before I go any further, I
should add that Princeton University's Russian literature professor Ilyavinitsky has
written extensively about how this quote is not in fact
(15:23):
from Dustyevsky, and actually it's in conflict with what Dustyevsky's
actually actual words on the subject of prisons have said.
It's a great read, by the way, I love reading
experts who are saying, yeah, everyone says this quote came
from this place, but it really doesn't. But wherever the
quote does come from, it has been repeated and reinforced
by lots of others, including like Nelson Mandela, And while
(15:46):
the origins of the quote are perhaps unknown and maybe
even unknowable, I do think the intent behind the quote
has merit no matter who it was that said at first.
Which is a long way to go to introduce this
next story, which is that the FCC has no now
closed a loophole that existed here in the United States
that allowed the various prison telephone service operators in this
(16:08):
country to place really high telephone rates on prisoners who
wanted to make calls, specifically intra state calls like within
a state. And most prisoners, like their families, live in
the same state that the prison that they're in is in,
so most of the calls they want to make are
within the state. And for a long time, those rates
(16:29):
were pretty darn high. And like a fifteen minute phone
call might cost a prisoner eleven dollars thirty five cents. Well,
that doesn't sound like a lot maybe to you, But
then you've got to remember these prisoners are making pennies
on the dollar in prison. Like there's a whole conversation
to be had about what ultimately amounts to slave labor
(16:51):
and the prison system of the United States. But anyway,
now that eleven dollars thirty five cents is going to
be closer to ninety cents at least that's what it
would be for very large prisons, like prisons that have
at least one thousand inmates, if not more. This is
significant because a previous challenge to the FCC's attempts to
curb telephone rates in prisons was defeated when the companies
(17:15):
that are actually providing these services argued that the FCC
lacked the authority to impose price caps in the first place.
But President Biden signed into law a bill that grants
such powers to the FCC. So now the FCC is saying, well,
Congress has granted us that authority. Now, so now you
have to reiin in your prices, and we've already analyzed
(17:36):
the numbers. We can prove that, based upon the information
provided by you, you'll still make a profit and cover
all costs of operations even with these lower price caps.
So there, so that's a good story. I think that
shows how if you don't have regulations, there's no check
and balance system for most vulnerable populations, right like powerful populations,
(17:58):
ones that have a lot of sway they can push back,
and that's substantial. But when you're talking about prisoners, who
are the most vulnerable in society arguably or at least
among the most vulnerable populations in society, they have no
political clout. So obviously any company that decides that they're
going to price gouge, there's no checks and balances there.
(18:21):
So now the FCC is saying, well, we've been grinded
that authority. You can't do it anymore. We'll see where
it goes from here. To round up for recommended reading
this week, I've got a couple of space related pieces
from ours Tetnica, both by Stephen Clark, and I think
these are worth your time. So first up is his
Rocket Report, which is a weekly roundup of space news,
and it's incredible, Like he I don't know where he
(18:43):
gets the time to do all this, because he does
a great job. But you can learn about how a
space startup recently lost at CEO due to inappropriate relationships
with an employee of the company. You can learn about
how SpaceX's Falcon nine rockets are temporarily grounded after a
launch vehicle. The second stage of a launch vehicle failed
(19:04):
to deliver twenty Starlink Internet satellites into their proper orbit
last week. That being said, the Falcon nine actually remains
the most reliable launch vehicle in history and it has
a ninety nine point seven percent success rate. This was
the first failure, so it's an amazing piece of technology
and the FAA is likely to lift that grounded status
(19:25):
within the next day or so. There are also lots
more stories in that roundup, and next up is another
piece by Stephen Clark in Ours Technica. It's titled NASA
built a moon rover, but can't afford to get it
to the launch pad, and it's the sad tale of Viper,
a project that already saw four hundred and fifty million
dollars of investment, but sounds like the various delays in
(19:45):
the project and the cost increases reached a point where
NASA had to make a decision they had to either
cut the project or they would have to cancel a
bunch of other projects in order to find money in
the budget to get it across the finish line. And
last is a piece in The Guardian by Charlotte thal
titled doom Scrolling Linked to existential Anxiety distrust suspicion, and
(20:08):
despair study finds. I think Thal did a really good
job pointing out that the study had some limitations and
it does seem to confirm a lot of common sense
ideas about the nature of doom scrolling, like it's going
to just sort of be like, yeah, of course. But
she also points out that due to the scope and
nature of the study, it is perhaps viewed as a
(20:28):
good reason to research the matter more thoroughly, rather than
to draw conclusions from the study itself. In other words,
be careful not to fall victim to confirmation bias. And
that's it for the tech News for the week ending
July nineteenth, twenty twenty four. I hope you are all well,
and I'll talk to you again really soon. Tech Stuff
(20:53):
is an iHeart Radio production. For more podcasts from iHeartRadio,
visit the iHeartRadio app, Apple podcas casts, or wherever you listen
to your favorite shows.