All Episodes

February 23, 2024 19 mins

In what I can only describe as an overcorrection, Google's Gemini AI has tried to promote diversity in situations where it's not appropriate, such as in an image portraying Nazi soldiers. Plus, Bluesky gives more control to users, Cruise looks to resume operations in a different state after California revoked its license, and researchers find a way to double processing power without upgrading hardware.

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 (00:04):
Welcome to tech Stuff, a production from iHeartRadio. Hey there,
and welcome to tech Stuff. I'm your host, Jonathan Strickland.
I'm an executive producer with iHeart Podcasts and how the
tech are you. It's time for some tech news that
caught my eye this week. We are on February twenty third,

(00:25):
twenty twenty four. Let's get to it now. Yesterday was
a bit of a doozy for cellular carriers here in
the United States. Disruptions and service plagued AT and T customers,
and reportedly some folks who patronize other carriers had similar problems.
It was a massive outage and it lasted for several hours.
Those with Wi Fi capable phones could still make calls

(00:48):
over Wi Fi networks, but once out of the electromagnetic
embrace of the router, they found themselves without servants. Speculation
range from a technical glitch to a possible cyber attack
as being the cause of the issue. That last possibility
got the authorities involved, as more than seventy thousand dollars
reports flooded in from cities like Houston and Dallas and Texas,

(01:09):
Los Angeles. Out in California, Chicago and Illinois and my
hometown of Atlanta, Georgia. As for those other carriers, T
Mobile and Cricket Wireless reported some of their customers had
also mentioned a loss of service, and curiously, it seemed
to happen after those customers had tried to call AT
and T customers. I should also add that Cricket Wireless
actually piggybacks off of AT and t's network, so maybe

(01:32):
it's not such a big surprise that they had some
issues with their customers. The Federal Communications Commission, or FCC,
has launched an investigation into the incident. The US Cybersecurity
and Infrastructure Security Agency also looked into it, plus the
Department of Homeland Security started to investigate it. But this morning,
AT and T says the problem that came down to, quote,

(01:53):
the application and execution of an incorrect process used as
we were expanding our network end quote. That's according to
an actual AT and T release. This lines up with
an earlier analysis from the Cybersecurity Agency. They said that
the most likely cause of the issue was a fault
in a cloud computing configuration error. While it took hours

(02:14):
to restore service to all customers, AT and T says
that process is now complete so hopefully if you have
AT and T you don't have any issues right now.
And now a quick update on Cruise. Not Tom Cruise,
but rather the autonomous taxi company that famously got in
trouble when one of its vehicles struck and then dragged
a pedestrian for twenty feet or so while the car

(02:35):
was pulling over, and then allegedly Cruz was attempting to
cover up the incident by not sharing their video of
the actual accident. Anyway, the state of California has revoked
Cruise's license from operating within that state, but the company
is looking at other cities and is reportedly preparing to
return to service with actual drivers in the seats as well.

(02:58):
Spokesman Pat Morrissey said, our goal is to relaunch in
one city with manually driven vehicles and supervised testing as
soon as possible, once we have taken steps to rebuild
trust with regulators and the public. End quote. Now. I
would like to think that our trust would require a
lot of work to rebuild, but I just don't know,

(03:18):
because some folks just love them sell some technology. But
then again, Cruz has to turn things around if it
is to remain a viable business. Since that accident, the
company has had several top leaders jump ship, and it
has also downsized more than twenty percent. So they definitely
need to make things start to happen, or else there's
not going to be a company anymore. Google is back

(03:40):
in the hot seat because it's newly renamed AI, which
is now known as Gemini, has a tendency to go
a little overboard for the sake of diversity. Now, I
think most of us understand that in most contexts, diversity
is important. It brings fresh perspectives and methods to bear.
Generally speaking, we all benefit when we concentrate on making

(04:02):
sure we are championing diversity. But if you wanted to
create a historically accurate portrayal of a group of people,
such as, oh, I don't know Nazi soldiers, you might
be shocked if the AI generated image included people of
color dressed in Nazi uniforms. The Nazis famously were not inclusive.

(04:23):
This sounds a bit like Godwin's law, right, I mean,
that's the law that says any conversation on the Internet
will inevitably gravitate toward Hitler as folks disagree with one another.
But the genesis of the complaints gets a little bit
more murky because they start in the anti diversity crowd,
or folks who are kind of like white rights people,

(04:43):
or they just are I guess they would position themselves
as anti woke. But these were people who are complaining
that if you asked Gemini to create images of like, say,
an American woman, that it would create some images that
included people of color, and maybe you'd end up with
more people of color than white people. Maybe you'd even
end up with a group of people and there's no
white person represented in that group. This has led to

(05:06):
some of these people arguing that Google is pushing anti
white propaganda on the public, which, you know, let's be serious.
I mean, the examples I've looked at at least have
had white folks in those examples. It's not like they were,
you know, completely left out. And goodness knows, America has
a very long history of producing media that studiously ignored

(05:26):
the fact that people who are not white also live
in America. See also the sitcom Friends. You would think
that New York was almost entirely white based on that show. Anyway,
the Nazi thing indicates that perhaps Google has been a
little heavy handed when trying to force diversity into image generation. Right, Like,
there are different ways to have diversity work its way

(05:49):
into your product, and brute force is not necessarily the
best way. You can almost imagine that the algorithm has
a little like switch at the end of the sequence,
as is the group of people that you've created diverse
if not insert people of color. Now, obviously that is
not appropriate in every context, and honestly it points out

(06:10):
that true diversity and fairness is really hard to achieve,
even if we're just talking virtually. That doesn't mean that
we shouldn't bust our butts to try and achieve it.
And I would recommend Adam Robertson's article on the Verge
titled Google apologizes for missing the mark after Gemini generated
racially diverse Nazis. If you'd like to read more, it's

(06:31):
a very good article. In other Google Gemini news, Google
incorporated an AI writing generator into Google Chrome this week.
Now to use this, you actually first must enable it.
You have to go into settings to do this, and
I did that this morning, but I haven't really had
a chance to take this out for a spin to
see what it's like. I'll be sure to report back

(06:51):
once I've had some time with it. I did try
the tab organizer option. There's an option that's supposed to
reorganize your tab so that they make sense, but I
had already kind of grouped my tabs into relevant research.
So I guess there just wasn't anything left for Gemini
to organize because nothing happened. But you know, it's either

(07:11):
that or it just didn't work anyway. You can also
do things like you can click within a text field online.
Let's say you found an online form, like maybe you're
leaving a review for a product, and you can choose
help me write by right clicking, you know, write write.
That will get Gemini to either write or rewrite your stuff.

(07:33):
So maybe you want to sound particularly educated while you're
writing your next review about a pair of sneakers or whatever,
you can use this option. I have not tried this yet,
so I don't know what it's like, but it's good
to note that if you do use this feature, Google
will collect some data about the whole process, like what
site you're on and what you wrote or had Gemini

(07:55):
write for you, and further this then could be reviewed
by human beings further along the chain, as a way
to make sure that Gemini is on track and is
improving over time, So there could be human review added
into this, So for the privacy minded, you might want
to think twice before you activate this feature, you know,
and otherwise the things you write are possibly going to

(08:18):
be reviewed by human beings, and if you're writing stuff
that you'd rather human beings not actually see, which I
don't know why you'd be writing on the internet if
that's the case, but maybe don't use this. The website
Reddit is about to go public, so it's about to
hold its initial public offering, as in, you know, become
a publicly traded company. This is a really big deal
for several reasons. Reddit the site that lets users create

(08:40):
and manage pages called subreddit. It's about pretty much any
topic you can imagine, and probably a few you can't.
It has served as a sort of gathering ground for
interesting links to all sorts of stuff online, and it's
been around for years, but it has never turned a profit.
In fact, it lost more than ninety million dollars last year. However,
it did post a twenty one percent growth in revenue

(09:02):
year over year, so it made more money, which meant
it lost less money overall. It also recently went through
a really rocky situation where the company changed its Application
Programmer Interface policy or API policy, and in the process
it alienated a whole bunch of its user base. You know,
Reddit saw users deliberately sabotaged several popular subreddits as a

(09:25):
result of that, and the experience left many Reddit users
sour on the site. Now that being said, Reddit has
offered something of an olive branch, at least to the
most prolific Reddit users. They will have the chance to
purchase shares and Reddit before it goes active on the
stock market. That means you can potentially buy shares at
a lower price than what it will be when it
is trading on the stock market, though that's not a guarantee.

(09:49):
They can be pretty exclusive, as you know insiders in
this case. So there might be some prolific redditors out
there who still hold a grudge against Reddit for those
changes the API. But I have seen people really fall
over themselves for the chance to be part of an
exclusive group, even if that group isn't really impressive. So
I'm guessing a lot of the folks who have the

(10:09):
opportunity to make those purchases will in the interest of
full disclosure, I'm definitively not one of those people. I
use Reddit for research, but I do so pretty much
only as a lurker. I think I may have posted
on Reddit perhaps half a dozen times total across my
entire history with the site, and it's almost always as
a comment to an existing thread. But in other interesting news,

(10:32):
Open Ai co founder and former CEO and then reinstated CEO,
Sam Altman is one of the top shareholders on Reddit.
He boasts an eight point seven percent share in the
company that's more than twice as much as CEO Steve
Huffman possesses of his own company. This means that Altman
can actually outvote the CEO on shareholder matters. Altman's shares

(10:55):
means he's the third largest shareholder overall. The company's Advance Publications,
which owns Conde Nast among other things, and the Chinese
company ten Cent, which owns pretty much everything, are ahead
of him. Also probably not related, but Reddit has started
to negotiate with AI companies to allow them to use
Reddit to help train AI in return for paying a

(11:16):
hefty amount of money. So Google recently forked over sixty
million dollars and have agreed to do that per year
for the privilege of using Reddit to train AI. Which
is interesting that Sam Altman has such a prominent place
in the shareholders and Reddit is taking money from Google,
a competitor to open Ai, for this training privilege. I
really wonder if open ai will have to cough up

(11:37):
money to do the same thing. I guess we'll find out. Okay,
we're going to take a quick break to thank our sponsors.
We'll be back with more news after this. Okay, we're back.
So Yahoo is shaking things up over at in gadget.

(11:58):
If you're not familiar with the inn gadget, that's a
site that reports on tech topics and offers up stuff
like product reviews. The company laid off ten people in
the editorial department, including top leadership positions like the editor
in chief, and it also split the editorial department into
two groups, so one group will focus exclusively on news
and features and the other one on reviews. Reportedly, this

(12:21):
is to realign the editorial staff in an effort to
drive more traffic to the site, because more traffic means
more advertising revenue. And Gadget's general manager issued a memo
that said the changes will quote increase our velocity in
the quote that raises huge red flags for me as
someone who used to write articles for a website. Increase
our velocity sounds a lot like making deadlines much tighter,

(12:44):
requiring writers to write far more pieces per given amount
of time, and giving writers less time to research and
understand a topic, which makes it harder to write something
that isn't just a regurgitated press release. And that doesn't
really help anybody. Now. I don't have any attachment to Engadget.
I don't know any people who work there or used
to work there, but I really hope everyone find it, finds,

(13:06):
you know, gainful employment. They end up in a good place,
because it's a rough time to be out of work
as a web based writer. Now I probably should have
put this next story after the cruise story, but here
we are. So. The State of California has put Alphabet's
application to expand Robotaxi services in the San Francisco and
Los Angeles areas until June nineteenth. The California Public Utilities

(13:27):
Commission states that the application is suspended for further staff review. Now,
this is a standard procedure if the agency does not
approve an application within thirty days of it having been submitted.
So it's not like this is totally unusual. It's not
like it. It's definitely an indication that California is pushing
back against alphabet and WEIMO. My guess is that the

(13:48):
state of California is reevaluating their stance on autonomous vehicle
operation in the wake of the cruise incident as well
as other accidents involving autonomous vehicles, and that it might
be a wild for it issues any approvals to expand
service until the state legislation decides upon suitable rules and regulations.
Go back just a couple of years, and a lot
of car manufacturers boasted of plans to switch over to

(14:11):
only making electric vehicles in the near future. Mercedes Benz,
for example, said the plan was to go all EV
by the year twenty thirty, but now the company is
backing out of that promise. So the problem you see
is that folks are not buying as many evs as
the companies thought they would, and so the companies are
scared that going all EV will just devastate their industry,

(14:32):
and so companies are starting to change their tune, saying
that they won't be quite as aggressive in pushing evs
to the market, and that they will continue to produce
gas and hybrid vehicles instead. That's a blow to environmentalists
who are really hoping to see consumer transportation make a
really big change. But the market dictates where these companies
are going to go, and it's hard to argue that
a company should produce itself out of business. That doesn't

(14:55):
do really anyone any good. Now, one could argue that
issues like vehicle costs and charging station availability are really
having an impact, but in fact it's a really complex
challenge and maybe it's just too complex to follow through
on promises that were made just a few years ago.
Blue Sky, one of the federated social platform alternatives to
x slash Twitter, made a really big move this week.

(15:16):
The company will now allow you to choose who hosts
your data. So by default, blue Sky hosts all your information,
but you can designate another party to host all of
your data, and that includes stuff like all the stuff
you post, all of your likes, all of your follows, etc.
And that even includes yourself. If you want to store
your data on your own server, you can do that

(15:37):
and it doesn't have to be forever. Blue Sky structure
allows you to port data from host to host, so
if you change your mind, you can move your stuff.
Now this is huge since most social platforms require you
to host all of your information on their own servers
because your data is valuable and the platform wants to
leverage it. Being able to choose where it all goes
gives more power to the user, and if the user

(15:59):
determines that they're the current host is doing something shady
or it's a poor steward of their information, they can
change where it's stored. If you want to read more,
there's a post on blue Sky. It is titled blue
Sky an Open Social Web and it has all the details.
Our next story is cool because it suggests we can
achieve incredible boosts and processing power without having to fight

(16:20):
against the laws of quantum mechanics to do it. A
UC Riverside Associate professor named Hungwai Seng has published a
paper titled Simultaneous and Heterogeneous multi Threading and it explains
this concept. So basically saying points out that modern computers
typically have multiple processors. You know, you might have a

(16:41):
CPU and a GPU, you might have hardware accelerators for
stuff like machine learning or AI, and all of these
handle information separately, kind of like as if they are
in a silo, and then information can travel from processor
to processor. But this means the overall process is inefficient.
So Saying In his team demonstrated that by multi threading

(17:02):
processes across all of these different types of units, he
could speed things up considerably. He nearly doubled processing speed
while at the same time almost cutting the amount of
energy needed to complete those processes by half. So not
only does it boost computational speeds, does so in a
really energy efficient way. It's pretty insane that he could

(17:23):
get nearly double the output just by using multi threading
across processors without having to do a single hardware upgrade.
And this has the potential to create enormous benefits. So
imagine that you were able to double the output and
half the energy requirements of a data server farm for example.
Like that would mean that you would need less water

(17:43):
to cool the data farm, you would be using less
energy to run it, so it'd be more cost efficient.
You'd be able to double the output. That's incredible. Of course,
we could probably just screw that up by doubling the
number of servers we're storing in that server farm, and
then we're right back to where we started. But you know,
you get the point. Satan himself is actually cautioned that
a lot more work has to be done to figure

(18:04):
out when this methodology will be practical and applicable, because
it might be that for some applications it's ideal, but
for others we might see far less impressive gains and
that it wouldn't make sense to implement it. So it
is good to keep that in mind. But still pretty
darn cool. Now, before I sign off, I do have
an article i'd like to recommend to y'all, and this

(18:24):
one is from Wired. It's by Lauren Good and it's
titled tech Job interviews are out of Control. So the
piece explores how the culture in the tech sector has
changed dramatically over the last couple of years. You know,
once upon a time, developers and engineers found themselves courted
by headhunters and they were constantly being lured to greener
pastures on a regular basis with bigger compensation packages and benefits.

(18:47):
But now an interview process might require proving your worth
by working on a project without you know, actually being
an employee. So there you are doing real work for
free in order to prove that you're worth high hiring,
and meanwhile you're taking up time and energy you could
be using to actually land a gig, which is pretty
crazy stuff. But the full article goes into way more

(19:08):
detailed it's well worth reading. As always, I have no
connection to Wired or to Lauren Good. It's just a
good article and I think you should check it out.
That's it for the tech News for February twenty third,
twenty twenty four. I hope you are all well. I'll
talk to you again really soon. Tech Stuff is an

(19:32):
iHeartRadio production. For more podcasts from iHeartRadio, visit the iHeartRadio app,
Apple Podcasts, or wherever you listen to your favorite shows.

TechStuff News

Advertise With Us

Follow Us On

Host

Jonathan Strickland

Jonathan Strickland

Show Links

AboutStoreRSS

Popular Podcasts

2. In The Village

2. In The Village

In The Village will take you into the most exclusive areas of the 2024 Paris Olympic Games to explore the daily life of athletes, complete with all the funny, mundane and unexpected things you learn off the field of play. Join Elizabeth Beisel as she sits down with Olympians each day in Paris.

3. iHeartOlympics: The Latest

3. iHeartOlympics: The Latest

Listen to the latest news from the 2024 Olympics.

Music, radio and podcasts, all free. Listen online or download the iHeart App.

Connect

© 2024 iHeartMedia, Inc.