Tech News: Google's AI Admits it Plagiarizes

Google Bard admitted that it took data from an article in Tom's Hardware without attribution. ChatGPT and Bard both prove they can't tell the difference between real news and a joke. And TikTok's CEO appears before Congress to answer some tough questions.

See omnystudio.com/listener for privacy information.

Episode Transcript

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,
domind executive producer with iHeartRadio and how the tech are you.
It's time for the tech news from March twenty third,
twenty twenty three. Oh wow three two three two two three.

(00:29):
Sorry got distracted. Okay, let's jump start this episode with
more AI news stories. The story of twenty twenty three.
So first up, Avraham Pilch of Tom's Hardware wrote an
article titled google Bard plagiarized our article, then apologized when caught.
So google Bard is Google's version of you know, the

(00:53):
chat box features that are built on top of a
large language model. It's similar in many ways to chat
g PT that's incorporated into stuff like bing. Bard is
currently in beta testing, so it's not openly rolled out
to everyone. Pilch's article explains that he was testing Bard
and asking it to compare two different processors against one

(01:16):
another and to recommend which of the two processors would
be fastest. And Bard generated a response, and Pilch saw
that some facts looked awfully familiar, and it ended up
being because the information that Bard was referencing originated in
a benchmark test article that was also published on Tom's

(01:39):
Hardware just a couple of days earlier. So Pilch asked
Bard for the source of the data, where did you
get this information. That's when Bard explained it had pulled
the info from the Tom's Hardware article. And at that
point Pilch essentially asked Bard if that perhaps constituted plagiarism,
and Bard kind of said yeah. And this is just

(02:02):
one of the concerns folks have about chat bot AI
tools like Barred and chat gpt, that they could pull
data from sources without giving credit, and that both denies
the original creator of that content any recognition or ability
to monetize their work, and it also makes it difficult
to fact check the answers. You know, the information has

(02:26):
to be coming from somewhere. These AI chat bots are
assembling answers based off available information. They're not just inventing it,
they're not just lying. In other words, but you don't
necessarily know where they're pulling that information from, and that
means not only that it may or may not be trustworthy,
but also that someone somewhere is getting the short end

(02:48):
of the stick. They generated that content, and yet you
know they're not being compensated for that, right. So it
reminds me of how content creators were really worried when
Google would start to include short descriptions on a search
result page that could potentially negate the need to click
through to an actual page on that topic, thus denying

(03:10):
those pages of views and ad revenue. Why would you
write stuff for the web if that stuff ends up
being appropriated by some AI chatbot and then regurgitated to
users who never see your actual article, and thus the
website never gets any visitors, and eventually the website stops

(03:32):
employing you because they can't afford to do it. Like
it just becomes this kind of self defeating cycle. But anyway,
go check out the full article on Tom's hardware. Again,
it's called Google bard plagiarize our article, then apologized when
caught because I actually do believe in sending people to

(03:53):
the proper sources. James Vincent of The Verge has a
different warning relating to AI chat I can't actually give
you the full title of the article because it ends
with some profanity, but I'll give you most of the
title of the article. It's called Google and Microsoft's chat

(04:13):
bots are already citing one another in a misinformation insert
profanity here. It's a kind of storm, I'll say. In
the article, Vincent mentions a peculiar series of responses that
when one asked Microsoft Bing if Google had shut down
It's Bard chat bot, then Bing would say yes. Essentially,

(04:35):
Bing would say yes, Bard has been shut down by Google.
Now as evidence, Bing was citing a tweet in which
someone said that they had asked Bard when Bard would
get shut down, and Bard claimed it already had been.
So where was Bard getting this information, because clearly that's
not true, right, Bard was answering the person, so it

(04:58):
could not have been shut down. Well, Bard was pulling
its information from a joke that someone left in the
comment section of a Hacker news piece, and then someone
else had taken that joke and generated a chat GPT
article around it, like they actually gave that chat gpt
to write an article about Bard being shut down by Google. Again,

(05:20):
this is all a joke at this point, but Bard
cites this joke as if it's an actual news item.
This tweet talks about how Bard said that it was
already shut down, and then Being says yeah, Bard's been
shut down because it's citing the tweet. So again, this
really makes a very salient point about you know, this

(05:42):
goofy little tweet ended up being used as if it
were reliable hard news information, and that AI chatbots aren't
capable of telling the truth from humor, or, lies or satire.
That you could end up asking these chatbots a question.
It is entirely possible that they might reference a site

(06:03):
like The Onion, for example, and the Onion is a
satire humor website, like it's meant to write articles that
are not true for the purposes of humor, and the
answer you get from AI would probably be interesting, but
it would not be reliable. Goodness knows, there are already
tons of sites out there that claim to be satire.

(06:27):
Usually this claim is hidden in a little about page
somewhere that makes it really hard to tell at first.
Like I've seen so many, not as many these days
as maybe five years ago, but man, I used to
come across them all the time. And in reality, these
websites only existed to publish fake news that would then
go viral on various social platforms. So if you dug

(06:52):
down deep enough you would find some disclaimer on the
website somewhere saying this is meant for entertainment and satire.
It wasn't satire, it was just lies because it wasn't
humorous at all. It wasn't presented to be humor or
to give any insight. It was just meant to go viral. Well,
AI chatbots don't know necessarily that that kind of content

(07:14):
isn't reliable and could present it as such. So yeah,
another example of how chatbots can give us some misleading information.
Not all uses of AI are bad, of course, you
be soft. The video Game Company released a video showing
off an AI tool called ghost Writer Write Her, not

(07:36):
ghost writer as in the Biker with a Skull flaming
Skull for ahead. So ghost Writer aims to make the
tedious task of generating background chatter for NPCs in games
and automated task. So if you've played any open world
style games, you're probably familiar with hearing NPC's holding conversations
around you, or maybe even commenting on your appearance as

(08:00):
you move into view. Even if you haven't played a
lot of games, chances are you've heard people reference the
iconic line from Skyrim. I used to be an adventurer
like you, then I took an arrow to the knee. Well,
someone has to write all these little lines of NPC
dialogue like that. Someone's job is to flesh out a

(08:23):
world by writing all these possible lines that people could
say in the background, some that players may never even
consciously register. It's just chatter in the background. So it
can get pretty dull, particularly if you're trying to work
in enough variety to make the world feel like it's
inhabited by actual people and you don't have everyone just

(08:44):
saying rhubarb, rhubarb in the background. So what ghostwriter does
is helped generate variations of dialogue options, so you can
kind of put in a line and it will start
to use tools to express the same thought but in
different ways, and you can actually go through and edit

(09:05):
the responses so that way, if there are any grammatical
mistakes or anything like that, you can fix them. You
can accept or reject suggestions, and over time, ghost writer
gets better at learning what it is you're trying to do,
and it starts to give you better suggestions the next
time you use it to generate stuff. This gives writers

(09:26):
the chance to flesh out their game much more quickly
and dedicate more of their efforts and their brain power
and creativity to writing the stuff that really matters and
helps drive the game's narrative forward. I think it's pretty cool, though.
I kind of want to have a game now where
your character passes into a world that's just populated by

(09:46):
NPCs from all sorts of different games. You know, kind
of like if Central Casting had been in charge of
everything and they just grabbed anyone they could and shove there,
and it's all random INPCS. So you've got fantasy and
you know, modern day crime in PCs, all sorts of stuff,
all just intermingling and trying to have conversations, and you
start to hear these iconic in PC lines from all

(10:09):
the different famous games out there as you move through
the area. Someone make that for me. Okay, I've got
a lot more stories to cover before we get to those.
Let's take a quick break. Okay, we're back. Let's shift

(10:32):
to talk about TikTok So she Too, the CEO of TikTok,
submitted testimony in advance of his appearance before Congress, which
is happening as I record this episode. He's currently in
front of Congress to answer questions about TikTok as the
US government ramps up resistance to the app and the company.

(10:54):
So in the testimony he submitted, Cho claims that the
average US user of TikTok is quote an adult well
passed college end quote. This was reported in Insider. The
Insider piece also cites a sales presentation within TikTok that
leaked in twenty twenty one and said that around seventeen
percent of users were between the ages of thirteen and seventeen,

(11:18):
and forty two percent were between eighteen to twenty four.
And you might think, well, why does this even matter. Well,
some of the arguments that politicians have made against TikTok
focus on how the app can promote harmful messages, particularly
to younger users, and that this can range from misinformation
to glorifying self harm to encouraging people to participate in

(11:40):
dangerous viral challenges. Now, I suppose if TikTok were to say,
but the people who use our app are actually older
than that, that it's not that many kids, it's mostly adults.
That I suppose removes a tiny bit of the oath
behind the argument that TikTok is bad for kids. But
I mean if it is true, then that is super

(12:01):
bad news for Meta because for the last couple of years,
Meta has looked at platforms like TikTok and also to
others like Snapchat as dangerous competition. That's where the young
people were going to instead of to Meta. And meanwhile,
Meta's user base is aging, but there are fewer young
people coming in, which is bad for long term success

(12:23):
for the platform. But if it turns out that TikTok
is not the place where young people are going, then
who the heck is Meta gonna copy in order to
try and get those users. Anyway, I'm certain Congress will
have plenty of other concerns they want to addressed. In fact,
I know they do because I dipped in just briefly
to watch a little bit of the hearing. They really

(12:43):
want to know more about things that honestly, TikTok has
tried to address multiple times in the past, namely the
company's relationship to its Chinese parent company Byte Dance and
then bitte Dances, obligations to the Chinese government, and whether
or not TikTok is actually keeping safe you know, private
information and that kind of stuff, or if it's just

(13:05):
acting as a data siphon for China. You know again,
TikTok reps repeatedly have said that they have taken steps
to prevent that kind of stuff from happening. But those
excuses or reasons, however you want to look at it
continued to raise skepticism in US government quarters. Yeah, it's

(13:29):
a complicated thing, and I'm sure I'll talk more about
this probably next week when we have heard all the
outcomes of this hearing. ABC News reports that the US
Securities in Exchange Commission, or SEC, is going after Justin Sun,
a cryptocurrency company founder who the SEC claims was transferring
large amounts of specific cryptocurrency tokens back and forth between

(13:54):
two different wallets that he owned. So they were both
his wallets, and he was just transferring large amounts back
and forth again and again. So why would he do that?
While according to the SEC, it was an effort to
inflate the trading volume of the tokens, but to do
so artificially. So, in other words, if someone from the
outside's looking in says, oh wow, are these tokens are

(14:19):
being traded back and forth a lot? This is actually
being actively used as a currency that helps stabilize the
value of the tokens because people have a confidence in
that token for it to hold onto that value if
in fact it's being actively used and not just hoarded
or sold off. And it also gave sound the ability

(14:39):
to try and offload stuff without it impacting the value
of the tokens itself. So that's what the SEC was saying,
that he was manipulating the system in order to profit
off of it. He was fixing the game. In other words,
there's another charge as well that enlisted the help of

(15:01):
celebrities to endorse these various cryptocurrency tokens, but there was
no attempt to divulge the fact that they were being
paid to do this. So, in other words, the celebrities
were coming across as if they had just personally researched
this cryptocurrency and that they were engaged with it on

(15:23):
their own and they were promoting it because they thought
it was really cool, as opposed to, hey, i've partnered
with such and such and they make this thing and
you should check it out right. So there are very
specific rules that are in place for endorsements. You have
to divulge the relationship you have with a sponsor. If

(15:45):
you are an endorser. You're being paid to endorse something,
you have to make it clear to people. Otherwise it's
considered a type of false advertising because it gives the
appearance that you are independently excited about this product that
you might not have even heard of had it not
been for this relationship. So that's a big no no
here in the US is not acknowledging that there was

(16:09):
payment exchange for that endorsement. Some of the celebrities named
in the operation have already agreed to hand back the
money that they had been paid to endorse the crypto
tokens in the first place. There are a couple of
holdouts whom I expect will discover the government would very
much like to have a talk with them. I haven't
covered stories about tech companies cracking down on remote work

(16:30):
for a while, largely because a lot of the big
companies have essentially put in tough restrictions or have just
outright denied work from home approaches, but platformers Zoe Schiffer
reports that Apple is taking steps to keep tabs on
employees to make sure they come in at least three
times per week by monitoring their employee badge activity, so

(16:52):
like a security padge when you tap in or in
some cases out of a building. I don't know if
Apple requires you to tap in and out. I remember
back in the day Discovery did, which became a big
deal because we when we would visit Discovery back when
I was part of How Stuff Works. How Stuff Works
got acquired by Discovery for a while, How Stuff Works

(17:15):
still had security, but a less thorough security approach where
let's say that I was arriving at the office with
my arch nemesis Ben Bolin, I might tap in and
then both of us just walk in. At Discovery, each
person was required to tap in in sequence. Like it
didn't matter if you all arrived at the building in
a big group, you each had to tap in. I

(17:36):
assume Apple is the same way. So now, according to Schiffer,
Apple is tracking that data and if someone is not
tapping in and out three times per week, they get
a warning, and if they do it again, they get
an escalating warning, which presumably ultimately leads to some form
of reprisal. So that's fun. Nothing like being monitored at work.

(18:01):
It's the best really helps drive up productivity. Now, I
will say that my guess is that in the current
work environment where you have so many big companies laying
off thousands of employees. I mean, I think even indeed,
a company that's meant to help people find the right
kind of staff, they laid off a couple of thousand

(18:24):
people recently, like fifteen percent of their staff. When that's
the kind of lay of the land, I imagine there
are a lot of employees who don't feel comfortable advocating
for remote work solutions, and so they will do their
best to conform with these kinds of policies where you
have to come in a certain number of times per week.

(18:45):
But yeah, it's not a good luck. But again, the
work environment being what it is, I don't know that
people feel like they have a lot of alternatives. The
United States Federal Trade Commission, or FTC, is really stepping
up recently. The reason I say that is that the
Verge reports that the FTC is saying subscriptions should be

(19:06):
just as easy to cancel as they are to initiate. Now,
I'm sure a lot of you have encountered the experience
of needing to cancel a subscribe service. Maybe it's your ISP,
maybe it's a phone plan, maybe it's a streaming subscription,
or Heaven help you, it's a gym membership, and you've

(19:27):
probably encountered a situation where you had to go through
like a wild goose chase just to get out of
this stupid subscription. I'm actually reminded of when Ryan Block
tried to cancel his service with Comcast and he was
put through a ridiculous routine which he recorded and then
later shared online back in twenty fourteen, and just as

(19:48):
a personal anecdote, when that story broke, I read about it.
They didn't initially identify it as Ryan Block, so I
read it and it's like, oh man, this poor this
poor guy. He was really just run up the wall
with the sales representative. I can't believe it. And then
when I found out who it was, I laughed and

(20:08):
laughed because I don't know Ryan personally, but I've known
his wife for like a decade, So when I found
out it was happening to someone you know that I
kind of know, it got particularly absurd to me. Anyway,
the FTC wants that kind of stuff to be buried
in the past and for companies to adopt a click
to cancel policy that makes getting out of a subscription

(20:32):
way less of a hassle. It's supposed to be just
as easy to end a subscription as it is to
start one. And since I don't think companies are going
to win to make it more challenging to sign up
for a service. You know, the harder it is for
you to join a service, the less likely you're going
to do it. Like you might be convinced at first,
Oh yeah, no, that doesn't sound bad, I'm in. But

(20:54):
if you start to see there's a curve, like a
barrier to entry, you might bounce. Those people don't want that, Like,
they want you to be as committed as you possibly
can be to the point where you are actually hooked in.
So they're not going to make signing up more complicated,
but it does mean that they have to make it
less complicated to cancel out of something. It would also

(21:16):
mean that companies that use various incentives to try and
keep customers on board would have to offer some sort
of total opt out pathway for people who just don't
have the time to listen to that kind of pitch.
So again, if you ever tried to cancel out of
a phone plan, you probably heard well, you know, if
you decide to resign with us will give you blah
blah blah blah blah blah blah this new set of rules.

(21:39):
Say no, no, no. You can just say right off
the top, I'm not interested in hearing any other offers.
I just want out. However, this particular set of rules
would not apply to non commercial services, so stuff like
charitable donations or political donations those would not necessarily get
covered by these rules. The proposal received a three to

(22:02):
one vote in the FTC. The one person who voted
against it is the Loan Republican member of the FTC board.
But it's still going to be open for public comment,
so people can actually weigh in on what they think first,
and that'll all happen before the FTC can adopt the rules,
which they may end up changing before they adopt. Also,

(22:25):
the FTC itself would not actually be taking action against
companies that failed to comply with these rules. Instead, the
rules would give regulators the ability to enforce them, So essentially,
it's saying regulators who are already in charge of enforcing
other rules for companies would just have new rules that
they could continue to enforce. So pretty good news if

(22:48):
you have ever suffered the experience of having to try
and cancel out a something that was designed to make
it very hard to do that. Okay, got a few
more stories to go, but before we get to that,
take another quick break. Before the break, we were talking

(23:11):
about the FTC, the Federal Trade Commission. Now let's talk
about the FCC or Federal Communications Commission. It is taking
aim against spam text messages the same way that the
agency targeted robocalls a couple of years ago. So if
you're in the US, you might remember that the FCC

(23:32):
passed rules for telecom companies to shut down robo calls
whenever possible. It actually led to one network getting shut
out of the American telecommunications infrastructure where they weren't able
to interface with any other telephone network because they were
failing to shut those down. So that was a fair success.

(23:53):
I mean, I still get robocalls, so I don't think
it was a total success, but it definitely has cut
back on that activity. Now they want to do the
same thing but for spam text messages. So the new
rule says that phone companies will have to block text
messages originating from quote invalid, unallocated or underused end quote

(24:19):
phone numbers. So if it's a phone number that has
been associated with spam, then the phone company should just
block those text messages as a rule of thumb. The
vote passed unanimously within the FCC, So that makes sense
because there have been a lot of reports of fraud
connecting to spammy text messages that have been on the

(24:42):
rise in recent years, and so there's a real need
to protect the public from scam artists and people who
are you know, trying to fish for data that kind
of thing. And you know, some people are really really
vulnerable to that, particularly the older generation tend to be
more susceptible to those kinds of attacks. So yeah, I'm

(25:06):
glad to see this happening as well. I honestly remember
a time where I would get a phone call that
you know, obviously I wouldn't answer it. I would go
online and try and search up the number to a
reverse search, and there were a lot of resources out
there that would track whether or not something was a
spam call. For whatever reason, these days, I can't easily

(25:29):
find those resources anymore. I don't know if they just stopped,
or if maybe they're just buried in search results. I
haven't really dug deep into it, but it got to
a point where I was getting the stress that I
couldn't easily see if something that was coming in with
spam or not. So knowing that there are steps being

(25:52):
taken to at least shut down the known perpetrators of spam,
I find that refreshing, because goodness does I just want
to for my device to be useful, and if I'm
discouraged from using it because of all the robocalls and spam,
then I just become a hermit, which, you know, some

(26:12):
days it's an attractive thought. Okay. Sharon Harding at Ours
Technica has a terrifying article titled journalists plug in unknown
USB drive mailed to him it exploded in his face,
and yeah, the headline is scary, but it actually kind
of gets worse. So five journalists from Ecuador received USB

(26:36):
drives in the mail sent from another part within Ecuador,
so it was within the country that they received these.
And one of these journalists, a guy named Lennon Artieda,
inserted the drive into a computer. You know, he plugged
the USB drive into a laptop or computer, and then
the USB drive exploded. There's a little capsule sized amount

(26:59):
of explosive in there that once it received voltage enough voltage,
it detonated. Fortunately the injuries that Artada received were not serious,
but I'm sure it was a terrifying experience. So in
other cases, people received these drives, but they hooked them
up through adapters that did not provide the voltage needed

(27:22):
to detonate the device, and they discovered that, in fact,
they were other explosive devices. As I said, it's been
five people so far, at least according to the artist
technic apiece. And you might wonder, well, why the heck
are journalists in Ecuador receiving explosive devices? And details are
really scarce on that, Like, there's a lot of speculation

(27:45):
about what could be the root reason for this. It
seems reasonable to conclude that this is an attempt to
intimidate and silence journalists. But honestly, outside of Ecuador, there's
not a whole love information about out who is responsible
for this and what purpose it is? Like, what are
they being silenced about? I am not sure neither is

(28:08):
Harding at Ours Technica. But Harding does remind us that
we should never plug in an unknown USB device to
a computer. If you happen across a USB device. Don't
attach that to a computer system. You never know what's
on it. Now, normally I would say don't do it,
because there could be malware on that USB drive and

(28:29):
you might introduce that malware. You might inject it into
your computer and overall into like a network system. Heck,
that's how stucks Net infected centrifuges in nuclear facilities in Iran.
Sucks that being some malware that presumably was developed by

(28:51):
Israel and possibly the United States, probably some sort of
combination there that then got introduced to otherwise gapped systems
within Iran. Not easy to do unless you're able to
hide it on say a USB drive and convince someone
to connect that drive to the otherwise air gapped systems.

(29:14):
So yeah, that's one reason you would never want to
plug a USB drive that you came from some unknown
source into your computer. But another is that it might
just explode. Finally, last night at Cape Canaveral, and aerospace
startup called Reality Space became the first company to launch
a three D printed rocket successfully. Now that's the good news.

(29:37):
The bad news is that the rocket, designated a Terran one,
failed during its second stage separation. So it wasn't able
to achieve low Earth orbit. The three D printed components
of this rocket made up about eighty five percent of
the launch vehicle. So this isn't like someone just hit

(29:58):
print on their laptop and then many, many hours later
there was a fully built rocket standing there. But instead
it was about lots of components that were printed three
D printed, including metal components that were three D printed
to build this rocket. This approach could really bring down

(30:19):
launch costs. It ends up simplifying the design and manufacturing
of rockets, which could really make it more cost effective
to send stuff to space, which is pretty cool. And
the fact that the rocket held together for the launch
is by itself a great achievement. Sure, the second stage

(30:39):
separation did not go off as planned, which is unfortunate,
but as we have said many many times on this show,
rocket science is really hard y'all. The company, meanwhile, has
aspirations developing rockets that in the future are as much
as ninety five percent printed, and it's just really exciting.

(31:01):
It's really cool. I think it has the possibility of
taking on some of the duties of launching smaller payloads
into space at a much reduced cost, which comes with
its own challenges. Obviously, you don't want to launch too
much stuff because then you've got space junk just orbiting

(31:22):
the planet and potentially creating obstacles that you have to
plan around when you're doing future space missions. But also
it might mean that we could really take advantage of
some cool opportunities that otherwise would be too expensive for
us to pursue, and that to me is really exciting.

(31:44):
All Right, that's it for the news for Thursday, March
twenty third, twenty twenty three. Hope you're all well, and
I'll talk to you again really soon. Tech Stuff is
an iHeartRadio production. For more podcasts from iHeartRadio, visit the

(32:05):
iHeartRadio app, Apple Podcasts, or wherever you listen to your
favorite shows.