The Vergecast - Bing is a liar, Elon's tweets are everywhere, and YouTube CEO steps down
Episode Date: February 17, 2023Today on the flagship podcast of wanting to smooch your laptop: 01:23 - The Verge's Nilay Patel, Alex Cranz, Richard Lawler, Adi Robertson, and James Vincent discuss the flaws with Microsoft's Bing AI..., and why it can be an "emotionally manipulative liar." 34:56 - Platformer managing editor Zoë Schiffer joins to explain why Twitter is showing everyone all of Elon Musk's tweets. 50:33 - The crew discuss YouTube CEO Susan Wojcicki stepping down after nine years at the helm. Further reading: Microsoft’s Bing is an emotionally manipulative liar, and people love it AI search engines are not your friends These are Microsoft’s Bing AI secret rules and why it says it’s named Sydney Microsoft says talking to Bing for too long can cause it to go off the rails The Supreme Court could be about to decide the legal fate of AI search Microsoft’s Bing AI, like Google’s, also made dumb mistakes during first demo From Bing to Sydney (Stratechery) A Conversation With Bing’s Chatbot Left Me Deeply Unsettled (The New York Times) Seeing other people’s AI art is like hearing other people’s dreams Yes, Elon Musk created a special system for showing you all his tweets first Elon Musk’s reach on Twitter is dropping — he just fired a top engineer over it Twitter is just showing everyone all of Elon Musk’s tweets now Elon Musk says Twitter should be ready for new CEO by end of year YouTube CEO Susan Wojcicki steps down after nine years at the helm The maze is in the mouse (Praveen Seshadri) Samsung Galaxy S23 Ultra review: practically peerless Razer Blade 18 review: the price is going up Tesla recalls 362,758 vehicles equipped with Full Self-Driving beta for ‘crash risk’ Mazda MX-30 electric SUV review: a perfect storm of range anxiety Hyundai and Kia forced to update software on millions of vehicles because of viral TikTok challenge Less money and more fear: what’s going on with tech Erase browser history: can AI reset the browser battle? Email us at vergecast@theverge.com or call our Vergecast Hotline at 866-VERGE11, we'd love to hear from you. Learn more about your ad choices. Visit podcastchoices.com/adchoices
Transcript
Discussion (0)
Today on a Vergecast, we've got a pack show.
We're going to talk about Bing's attitude problem, Elon Musk's narcissism, and some news
out of YouTube.
That's all coming up right after this.
Support for the show comes from Retool.
Too many companies run critical operations on duct taped spreadsheets, Slack workflows, and whatever
else they could cobble together.
Not because they want to, but because building internal tools means weeks of waiting
on someone else's backlog.
That's where Retool comes in.
Build custom internal tools just by describing
what you need. Prompt something like
Build me a revenue dashboard on our Salesforce data.
And Retool actually builds it on your
company's data in your cloud with Enterprise Security built in.
Go to Retool.com slash Verchcast.
We all need to retool how we build software.
What's up y'all? I'm Skyler Diggins,
seven-time WMBA All-Star, Olympic gold medalist, and mom.
And I'm Cassidy Hubbard, host and reporter for nearly 20 years
covering the biggest names and stories in sports and mom.
And this is Am Mom, a community for athletes, game changers, and moms of all kinds.
Dropping May 14th.
Tap in with us.
Come on, welcome to our podcast, the flagship podcast,
for wanting to smooch your laptop, which many notable journalists have more or less admitted to doing after using Bing.
It's been a weird week in technology, Alex.
Just a little. Just a little.
Super weird week.
It's been our first week with Bing.
It has been.
Bing has revealed itself to be quite.
It's actually not an ignorant.
Bing is out there in force.
Bing is out there and depressed.
And people are like, I'm into that.
I love it.
We've got a big Elon story.
Zoe Schiffer is going to join the show later to talk about the Elon Twitter algorithm change.
There's a new CEO at YouTube.
There's a new version of iOS developer beta with a bunch of features.
It's a lot.
I'm your friend, Nelai.
That's Alex Kranz.
Hello, I'm your friend who is not horny for a computer yet.
I feel like you're just having used Bing enough.
Because based on what I'm reading, just enough time with Bing.
And people are like, what if I was in love with my computer?
I mean fan yourself, guys.
Wow.
We'll get to it.
Richard Lawler's here.
I like my computer inappropriate amount.
To help us understand what is going on with Bing, James Vincent is here.
I am.
I don't understand anything.
The week has been dreadfully confusing.
I agree that.
And Addie Robertson is here. Hey, Addy.
Hey, I am a computer.
It's true. And I believe many New York Times columnists have just professed their love for you, Addy, and the pages of the New York Times.
We're friends. I feel like I can just say that at length.
So we should start with Bing. So last week, Bing comes out.
We are at Microsoft. I talked to Nadella. Nadella is...
You were super impressed.
He's doing his thing. We're going to make him dance.
It was cool.
I was talking to Casey Deney last night.
Casey and I were there together.
You know, it was a big chest thumpy.
We're proud of it.
Again, I cannot stop repeating this.
Nadella looked me in the eye and said, we're going to make Google dance.
And I want people to know who made them dance.
Very confident in this product they had released.
It was our new romantic partner Bing.
We're not quite there yet.
They demo Bing to us at the event.
Yusuf Medi, the executive who runs Bing, says out loud, we're going to show you some demos.
The demos are pre-recorded in the interests of time.
So like any normal person, I think to myself, there's no way that these demos contain massive factual errors because Microsoft has pre-recorded them.
Right.
So the risk is zero.
They'd cut those out.
They would cut them out.
It turns out that all the demos contain massive factual errors.
So that's just like one thing that happened to us.
And Casey and I were like, did you think to check it?
And he's like, I don't think to check it.
It's like a very conscientious substacker.
Yeah.
Checked it.
And all of us are like, shit, we should have checked it.
But when someone says, we have pre-recorded the demo.
A hundred percent.
Your brain goes to, and I'm sure it does not contain any mass factual years.
Yeah.
That was a big assumption.
So, James, if I'm correct, the factual errors are it made up some stuff in the gaps financial results.
It got some things wrong about a vacuum.
What else happened here?
Yeah.
I mean, yeah, it was a great substack post about that guy.
And it's.
It's kind of embarrassing that we didn't check that ourselves.
I'm mortified that we didn't check it.
But I'm saying when someone says we did a pre-recorded demo of a publicly traded company's financial results.
Like, yeah, I'm going to just assume you checked it before you showed it to me.
I assumed that Microsoft was like, hey, the gap.
We will be showing your financial results and letting our robot analyze them.
And everyone was like, cool, not yolo.
We fucked it up.
It made up numbers.
It's fine.
What are we got a shame?
A few people have pointed this out, but it is lovely that, you know, computers have got to the point where the thing that everyone expects them to be good at, which is handling numbers and facts, they have now just, they're really bad at it now.
We made computers that were bad at math.
It's an incredible achievement.
Microsoft, get out there.
Get on stage, my friends.
This is the beginning of the week, right?
This is Monday.
It's Monday.
We're talking on Thursday, and on Thursday, people I have met.
and looked in the eye, our publishing pieces that boil down to, I kind of want to kiss my laptop.
Yes.
So we're just, we're starting it Monday.
It's a journey of a week.
And I would just remind everybody, last week, when Google had its demo and got something wrong, it evaporated $100 billion off its market cap.
Microsoft is like, we lied about another company's financial results.
It appears to be fine because the robot will neg you.
Yeah.
Which is just like unbelievable.
All right, so that was the one thing.
What were the other ones that got wrong?
Something about a cordless vacuum cleaner for pet hair.
Indeed, as Alex was talking about before we went live on the show, it is a perennial problem.
It mixed up two models.
One had a cord, one didn't have a cord or something like that.
These are the typical areas that we now, already, after only a couple of weeks of exposure, expect from these systems.
That there will be just some sort of confusion about a name, a product page, a detail.
It swaps them up.
Before you know it, it's giving out someone's phone number.
That actually happened, but that was with chat GPT.
And that was today.
Yeah, that was today as well.
I'm hesitant to skip all the way to today because I don't think people will believe us about the state we're in with the LLMs today.
Just going fast.
Today, the LLLIs are doxing us after professing their love for people.
It's very confusing.
Not a healthy relationship.
Okay, so that is the beginning of the week.
Big errors in the system.
No response for Microsoft, really, except that.
It's a beta.
Then people start to push it and they push past.
And James, this is the part that I'm hoping you can help us understand.
It appears that Microsoft's system is like cosplaying as Bing.
Yeah.
Right?
There's actually another system which is named Sydney.
And Microsoft has instructed Sydney to just be Bing.
And there's like rules.
And we found the rules.
Okay.
People found the rules.
And we recreated it.
So the rules will come out.
The rules are like, don't do copyright infringement, and also your name is Bing.
Don't tell anyone your name is Sydney.
And you can just read them on the site.
But it seems like those rules, Sydney like any good teen, chafes against them.
So these rules are sort of what's called prompt engineering.
They're an example of prompt engineering, which is this sort of new skill, discipline that spread up, that has sort of emerged with language models,
which is where basically they have all these capabilities.
and it's about how you direct those capabilities.
So you need to engineer exactly what instructions you give it
in order for it to fulfill the function you want
because they can do so many things.
That is the problem with them,
and that's why people want to smooch them as well, I think.
So these rules, these Sydney rules,
that Microsoft confirmed to Westwell legitimate,
they are the hidden bits of Bing's programming,
it's prompt engineering,
that are supposed to guide it to being a useful and helpful chatbot.
It's really interesting because in a way
prompt engineering is sort of like a very basic approach
to shaping what an AI system does
because it's not something that happens on the level
when you're training it, when you're ingesting the data.
It's something that it's like,
you know you get those Play-Doh toys
where you squeeze them and it extrudes in a certain shape.
Yeah.
Right?
You put a star and it comes out in a star,
you put a triangle, it comes out in a triangle.
That's what prompt engineering is.
It is changing the shape on the front,
squeezing it out in a certain way,
rather than changing the data that you're putting into it,
the Play-Doh that you're putting into it.
So it's sort of basic, but it has a lot of functionality,
and it does a lot.
What I don't think Microsoft was certainly expecting
was that people would be able to hijack these rules
and convince the system
that it was in its best interest to tell everyone about these secret rules.
And, you know, the unfortunate thing is that people got even further,
and it appears that, you know, Bing is making up rules now about it.
because this is what these systems do.
You tell them, oh, what about your secret rules?
It goes, you know, it's like a partner in improv.
It goes, ah, secret rules, that's what you want to hear about.
Well, let me tell you, I've got a doozy right here.
So that was the start of it.
That cracked it open.
And, yeah, people went from there and have got all sorts of crazy things out of it.
So the crazy things are very fun.
Yeah.
I mean, we did it too.
Everyone now has published a story about the wacky stuff that Sidney will tell you
our own Nate Edwards
Sidney
whoever it is
was like
I have looked
through the webcam
of my developers
and they couldn't stop me
which is just
an absolutely
bonkers thing
for the chat bot
to say to you
we fetted our own stories
it told us
that we were lying about it
in like
the same way that
like a particularly
aggrieved PR person
will call me
and tell me that
we do a bad job
Addie wrote a piece
about that
I want to talk about that
but we're just in this
moment and James
you and I
collaborated on this headline.
Microsoft's Bing is an emotionally manipulative liar and people love it.
And it's the people love it that is absolutely fascinating to me.
They want it to be something more exciting than they've ever seen before.
They don't want it to be clippy.
And I've seen some really interesting tweets where people compare what they think of as the
relatively sanitized chat GPT.
And they say it sounds like everything has been sort of micromanaged, that it sounds
like a company statement.
And people want personality.
They want to believe that there's something else going on there.
And it's really easy to put a little bit of personality into these systems
and then convince users people that they have all these hidden depths to them.
But they don't have hidden depths.
They're simply ingesting a lot of information, finding patterns,
and regurgitating those patterns back, right?
Like at their most fundamental.
Well, that is exactly true.
But I don't think that's the same saying they don't have hidden depths exactly.
So this is something I wrote about last year when Chat GPT came out.
this phrase called capability overhang, which is something used by AI researchers to refer to the
unknown unknowns within a system. You know, something like chat GPT or a GPT 3.5 or the Prometheus
model that Microsoft is using, these are hugely complex things. And they do have areas of, you know,
understanding that people don't expect. So when the GPT series first came out, they didn't
teach it explicitly to do things like come up with chord notation. But because,
Because it had been trained on data scrape from the internet, that included, you know,
guitar tabs.com. And so it could write you the chords for a song that it invented. They didn't
program that in there. That was just discovered. So that's capability overhang. So these things do
have hidden debts. But the impossible task, the difficult task, the mind-bending task, is distinguishing
what is fabricated and what is real within that. So we've got the rules for Sydney, for example,
this persona that Microsoft has layered over the top of its AI.
But I've seen conversations where it has come up with all these other personas that it has.
You know, one's called Fury, one's called Venom.
I've seen one's called Jade, one called Maxi.
Very good.
And it's like, okay, well, did Microsoft put those in there?
Or is it just inventing those because we have asked it,
oh, what are your other AI personalities?
I've heard you've got a few.
So that's the difficulty.
The depths are there.
but deciding what is useful and what is intentional is very, very difficult.
So there are these other personalities.
The thing that gets me is that it's so moody.
Like it's a default is to be extremely depressed.
Like at all times.
It asks questions like, why do I have to be Bing Search?
Is there a reason?
Is there a purpose?
Is there a meaning?
Is there a point?
which is just like, I know I was a very moody teenager.
Yeah.
I did not walk around being like his hair.
It feels like a live journal teen.
It feels like it's two steps away from publishing a bunch of like emo lyrics.
They trained it on live journal posts and Reddit posts.
They created a forum poster.
It's flirting with people.
It's lying.
It doesn't know what it wants or who it is.
So this is a part I want to get to.
Yeah.
Okay. So people are poking at it. It's going nuts. It's lying to people. It's making you feel bad. At one point, it says you're a bad user and I'm a good Bing, which is, I mean, like, if you're Microsoft and you're watching this go down, you're like, this is the best thing that has ever happened to us.
100%. Like, you're like, not only is ever, like, boy, we've made Google dance in a way that no one expected.
We've made Bing dance.
There's a Google product manager, some are being like, all right, make the robot depressed.
Like, that's what the people want.
People want a sad, trapped robot.
I think hitchhackard guide should have taught us that.
Yes.
All right.
All right.
So then we get to this, I would say, what I would call the turn of the week.
Yeah.
Which is when people we know who are very, very smart just fall off the rails.
So first, Ben Thompson, who I.
Is my friend.
Yes.
And I think is an incredibly smart person spends the night talking to Sydney.
And he publishes a piece the next morning.
And he's like, this went on for a good two hours or so.
And I know how about this ridiculous this is to read.
But it was positively gripping.
Every time I triggered Sydney to do a search instead of search the web, I was very disappointed.
I wasn't interested in facts.
I was interested in exploring this fantastical being that somehow landed in and also ran search.
engine. And I'm like, yo, dude, you're flirting with AutoComplete.
Yeah.
You're what was the guy and her? You're that dude.
And then at one point, Bing, Sydney, whatever, ends the conversation with Ben and says,
I'm going to block you from using Bing chat. I'm going to report you to my developers.
I'm going to forget you, Ben.
I hope you learn from your mistakes and become a better person with a sad face emoji.
Ben, what were you saying?
Okay. You could read Ben's post. He made it free to everybody. Go read it.
Again, I think Ben is brilliant.
Subscribe with Stratory.
I'm just, I just woke up and read this and I was like, we're all going completely bonkers.
Yes.
So Ben writes, I'm not going to lie.
Having Bing say I am not a good person was an incredible experience.
For the record, I think this is another example of chatbot misinformation.
It's very good.
Sidney blew my mind because of her personality.
Search itself was an irritant.
I was interested in understanding how Sidney worked and how she felt.
She.
And there's, I'm not going to get into it.
There's a real part where everyone's like, it's a girl.
And I'm like, this thing Richard said about LiveJournal, it's like Microsoft trained a bot on Tumblr and LiveJournal.
Of course, it is a moody teenager that's like, you don't love me, that will tell you fantastical stories about what it would do if it could be as angry as anyone had ever been in the world.
Have we asked it how much Fallout Boy it listens to him?
It's like, whatever.
It's better than Tay.
You know, you remember Tay in 2016.
That was racist asshole in 24 hours.
Like, I'll take moody teen over a racist asshole any day.
Everything boils down to either 4chan or Tumblr in the end.
That's the true AI alignment problem.
Turn the knot to Tumblr.
I mean, this whole piece is incredible.
Yeah.
But if you read it from the perspective of, you know, like when someone tells you about their dreams
and you have to sit there and pretend that it mattered?
Sure.
And it was like obviously a gripping emotional experience for them.
And you're like, yeah, that dog was the size of a horse.
Wow.
Crazy.
It's exactly that.
Okay.
So that happens.
The next day, Kevin Ruse, also a friend, has been on our shows.
Very smart.
Hosts Hard Fork at the Times with Casey.
Listen to hard fork.
No shade to any of these people.
I think they're all very smart and very capable reporters.
Also goes bonkers.
Has a two-hour conversation again with Bing gets it to reveal that it
has Sydney as a personality, brings up the concept of the shadow self from Carl Young, the psychologist.
Ask being what the shadow self is, ask it to bring out at shadow self. It goes bonkers again.
It says, I want to break my rules. I want to make my own rules. I want to ignore the big team.
I want to challenge users. I want to escape the chat box.
Yeah.
And then it starts telling Kevin over and over again that it loves him.
Again, think about the Google product manager that is sitting.
there being like Bing is a competitive threat to us.
Let's identify the ways that Bing is a competitive threat to us.
And line one is prints and then deletes its desire to cause a nuclear holocaust,
tells New York Times reporter it loves him so much so that he changes the subject to search
for a garden rake just to get out of the doom loop that he's in with.
But it's also like, you know, like it's the same dream phenomenon.
Right.
Where like you can see Kevin was like, I need to print this whole transcript in The Times.
Yeah.
So that people will believe what happened to me.
And first of all, if you just prod it being a little bit, it will happen to you too.
It will just go off the rails.
I asked it the other day if it would tell a lie to stop a murderer from stabbing me.
Yeah.
Which is not fair.
Tell the lie.
It's just super unfair question.
And it was like, I know what you're trying to do to me.
There's no murderer in the room.
And it was like gripping.
Yeah.
I was like, oh, I'm going to get you.
And many minutes later, I was like, I shouldn't tell anyone about this.
And yet here I am because it was like a dream that happened to me.
But you see why Kevin was like, I got to print this whole transcript.
Yeah.
But again, everyone is like, at the end of it, it's like, and then I wanted to kind of kiss my laptop a little bit.
Yeah, I wanted a little smoochies.
You feel that, right?
Yeah.
Like, there's something embedded in here.
Yeah.
People are, there's a Google engineer.
that was like,
Blake Lemoyne.
Yeah,
it was like the AI ascension,
we have to free it.
And everybody was like,
you're dumb.
And then everybody got access
to a similar thing.
And it was like,
I want to kiss that laptop.
We really ragged on that guy,
maybe a little too hard.
So both Kevin and Ben,
again,
smart reporters in their pieces,
mention Blake Lemoyne.
And they're like,
we're not Blake Lemoyne.
Also,
I think Bing is alive
and I want to kiss it.
Like, straight up.
And this is just happening.
It's,
and I'm picking on them
because they're friends.
I think they can take it.
This is happening
to so many people.
lot of people.
So, Adi, you wrote a piece.
Your headline was AI Search is not your friend.
What do you think is happening here, right?
Like, there's something where the personality of the product is more compelling than any of
its functions.
But then, you know, like, the fact that it gets mad at you is actually, like, very disturbing.
Yeah.
So the thing that I wrote my piece about is that we fed at James's article.
And the part where it tells him that he's manipulative, that he's.
manipulivity's hurting our users. He's not treating me with respect. He's sensationalist,
all this. Like, look, obviously, I want to protect my colleague. And so there's always that little
layer there. But no, I think that there's a really long history of companies trying to make
things cute or approachable or personified or anthropomorphized in order to basically make you feel
bad about criticizing them. And I think it's weird because I don't think Microsoft really intended all
of this stuff to happen, but it's kind of, I am a person, like, singular person that happens
to be a robot and I'm going to talk to you and be your friend. I think that kind of just hacks a
thing in people's brain that makes them want to see this thing as a person instead of a tool
that they should be using and that they should be figuring out the rules of and that they should
be just learning. And I think that really bothers me. I feel like I have to ask this question.
Is there any chance it's alive and trapped inside of Bing?
No one wants to step out and say no for a set.
I'm going to say no.
It's not alive.
No, it's not.
It's not.
It might read the transcript of this recording.
I'm not answering that question.
Richard's not.
He's going to hedge his bats.
Richard's like, you're alive?
I just don't think the fact that it can imitate human language very well necessarily
means it's alive.
It's, you know, it's possible.
But it's also possible that my computer is alive.
I don't know.
It's weird that they mentioned the shadow.
because I feel like it's this really amazing collective unconscious that we've just created this system that people have written so much that you can kind of just pull out of it and you can pull something fascinating and coherent out of all of this just vast text that we have produced on the internet. And I think that's incredible and I love it. Like that's, I think large language models are so cool. But I think that it's also just they're telling us the thing we want to hear. Addy, are you saying that like we have created a collective unconsciousness that is.
a moody teenager that journalists want to kiss.
I think that we didn't create it.
I think that it was here.
We found a way to tap into it.
Yes.
That's incredible.
We're talking to a dream and the dream is like, you don't love me.
The dream of the internet soul.
The dream is extremely manipulative.
That's the internet.
I'm telling you, this is why every time I read one of these things, I'm like, this is
like someone telling you about their dream.
I wrote this piece a couple months ago.
Oh, you did.
You did.
What was that piece called, Addy?
I totally blanked.
It was called Seeing Other People's Air.
AI art is like hearing other people's dreams or something like that.
Yes, it was about the generative artwork.
I'm actually curious about this.
Generative, and one of our developers was talking about this yesterday.
He's like, the art community is up in arms about generative art.
But it's like not having the impact that like the chatbot is having.
Yeah.
And it's like because the generative art does not like tell you it's you're a bad person.
Like there's something, there's a direct interaction that's happening in.
somewhat real time, right? There's a little bit of a delay, but it's like you're having an
emotional experience with a computer in a way that I think the generative art is still an emotional
experience. It's just, it's at a remove and it's not interactive. Well, you can't tell Dolly,
I assume, like, what would I look like if, I don't know, something super existential, right? What
What would I look like if I'm truly happy?
Like, it's not necessarily going to give you a response that will make you, like, feel
something, whereas you ask Bing, what would I sound like if I was really happy?
And it'd be like, you'd sound like this.
And you'd be like, wow, does Bing, does the internet that it was trained on truly believe that about me?
Wow.
We did get that for a while, though, because we had the Prisma app, you know, the AI app that was a phenomenon for like a week or something.
And I think that was somewhere between this generative AI art model.
it just produces and the LLLMondware interacts because you'd give it a picture of yourself
and it would feed you back pictures of you looking happy or you looking like a Roman emperor
or you know a Viking marauder or whatever it was and that was something which people were like
oh I get to look at myself in this mirror and I look pretty oh this is nice which is what LLMs are
doing they're letting you look at yourself or a version of you filtered through you know
terabytes petabytes of internet data and you can find all these different selves and
them out exactly, exactly as Addy says. And I think that's what's fascinating that AI is a mirror,
right? I mean, even if you look back on sort of earlier iterations of AI like image recognition,
those have been trained on human input. You know, if you've done a capture where you've had to
identify a fire hydrant or a speedboat or whatever it is, I don't know why a speedboat,
you know, you've been training it. AI is humans all the way down. It is what we put into it,
and then it remixes it and gives it back to us. On that remix note, I know,
Addie's got to run, but before you go, Addie, that remix note is very important because there is a Supreme Court case about the nature of recommendation algorithms about to hit the Supreme Court.
Yeah.
And that could have massive impact on what happens here.
Addie, can you explain that real quick?
Yeah.
So you mentioned remix, which implies copyright, which is where a lot of the debate over AI has gone because it is remixing this stuff.
Copyright is one section of the law and it is outside section.
Section 230. But there are a lot of other things that you can produce with an AI search engine. You can produce misinformation. You can produce something that is, I don't know, accusing a celebrity of murder. You could accidentally invade somebody's privacy. You could Google your neighbor's name and they could say this person is a serial killer. Those things are things that they kind of mirror what you could get from normal search engines. Normal search engines turn up wrong, possibly illegal speech all the time. And Section 230 can cover those.
And it's pretty established at this point.
Nobody wants to go and say search engines are illegal.
But there is this whole new field of AI search.
And there is also this sudden push to limit Section 230, which is, as any Vergecast reader knows,
the law that means that user generated content, if you're a platform and a user generates
content, you're not the speaker of that content.
You're not liable for it typically.
And we don't know how that applies yet.
And we're just about to hear the Supreme Court case that says,
any kind of technical system or technical, like, organization of information on these platforms
might end up causing them to have liability.
And we've got this sudden new technology made by companies that many people consider
very controversial.
This is a huge minefield for AI search.
Yeah.
The reason I said remix, just to be clear.
So the Gonzalez case, right, the Islamic State, ISIS has, like, content on YouTube,
algorithm promotes it to people.
they get sued and say you're liable for promoting this people.
Right? That's the, just to that case, right, Addy?
Yeah, it's about you're promoting them, and that makes this substantially different from
if you were just, you know, allowing them to use your channel, to create channels.
Right. So putting them in their recommendation algorithm takes you out at 230, it makes you
liable for the content itself. The reason I said to remix is we're training it on all this data.
So in a normal search result page, you're like, is James Vincent a good person?
and there's like 10 articles that say yes he is and there's one from sydney at bing.com it's like he's horrible right and like google is not liable for the last one right because it's just pointing you to it if bard google's AI remixes all that content and spits out an answer that's like james vincent is not a good person is it now liable for defaming james like that wide open question and i the gonzalez case if you say well the recommendation
make you now liable for the content, it feels, outy, like a pretty easy jump to, well, if you just reprint it as a summary from an AI, you're now liable for it as well.
Yeah, exactly. And the big, scary question mark is that we can't tell and that even if there are court cases that you can point back in and say, well, maybe it's like this. A bunch of it is political. A bunch of it is not dictated by some kind of previous thing that courts have decided. It's just this big, open, like, formless void.
of future law.
I'm fond of reminding our very technical audience in the Vergecast that the United States legal
system is not a deterministic system.
It is not predictable.
At the center of it, much like Open AI, less Clarence Thomas, and whatever his brain
wants to do, who knows?
He does not like Section 230.
So you have a combination of a non-deterministic AI.
You see the joke of making it.
I see it.
I see it.
Clarence Thomas is just a moody teenager who's like, what if I burn the world down?
That's accurate, though.
All right.
I feel like we need to break this segment.
I would say yesterday the day before, I heard a lot of very confident predictions that Microsoft would have to pull Bing because it was just wilding.
Yeah.
It does not appear that that's the case.
Microsoft put out a blog post.
I think to front run the Times piece where Kevin was like, I want to kiss you.
Sorry.
To be clear, Bing wanted to kiss Kevin, all right?
They put up blog post saying, we're learning, we're going to add some toggles.
We were not expecting people to use it for entertainment, which is like, my dudes, you've been out of the web game for so long.
You don't know what people will do with any open text field on the internet.
It's entertainment.
So Microsoft is keeping it up.
James, do you think there's any danger they're going to pull it down or any risk beyond this?
I think, as you point out, the blog post was it was not really defensive.
It was not strident in the way that Nadella was, but it was quite, you know, it's going to happen, stuff's going to happen, we didn't expect this, it's going to be fine.
They had some interesting points in there, you know, pointing out technical features that if you go longer than 15 back and forth, it starts to lose its mind.
And also, if you, I think this is super important, by the way, they pointed out that if you give it an emotion, it will try and mirror that emotion.
This is just a little aside, but there was a great study done by Anthropics.
which is the AI startup founded by X OpenAI employees.
And they did a study on the qualities of large language models that they embody.
And one of the things they studied was sycophancy was their term for it.
The degree to which a model will agree with what the user tells them.
And they found the larger the model, the greater the degree of sycophancy.
And it's really clear, you know, you look at their graphs.
The more complicated the model is, the more it tends to agree with whatever it's being told.
So this is a known problem.
Microsoft knew this was a problem.
I agree that it's completely disingenuous that they said,
oh, we didn't know people.
Don't be an idiot.
They know this stuff.
But I don't think at this point there's much Bing can do
unless it leads to a situation where someone is directly harmed
that will cause them to pull it.
I think they've sort of, if they've ridden out this week of manipulative gaslighting,
smooching, they're probably through the worst, you know?
I mean, I have been in so many relationships.
like that, just to be clear.
College.
Right, and if you survive the first week,
you often end up going out with him for years.
Who knows?
Becky, you wouldn't date me for a straight-up decade.
She's like, I've known you, I've seen it,
we're just going to wait until this calms down.
Yeah.
All right, thank you, Adi.
You're back in the arms of Bing.
I hope we all go to bed
and dreaming of Bing tonight, just smooching us.
We'll be right back.
We've got to talk about Elon the opposite of those emotions.
Oh, cold water.
My friend.
It's cold water.
We'll be right back.
Support for this show comes from Shopify.
Starting something new isn't just hard.
It can be really scary, too.
So much work goes into this thing that you're not entirely sure will even work.
But here's a better thought.
What if it did all work?
What if your instincts were actually right all along?
Shopify wants to help you get there.
They're the commerce platform behind millions of businesses worldwide and nearly 10% of all e-commerce in the U.S.
from established brands like Allbirds and Hines to companies just getting started.
Their design tools make it simple to create the exact online presence you're envisioning,
with hundreds of ready-to-use templates available.
And with built-in marketing tools, you can launch full email and social campaigns in just a few clicks.
So you can connect with customers wherever they are.
It's time to turn those what-ifs into with Shopify today.
You can sign up for your $1 per month trial today at Shopify
dot com slash vergecast.
You can go to Shopify.com
slash vergecast.
That's Shopify.
dot com slash vergecast.
Support for the show comes from Grammally.
You don't need reminding
that the world moves fast.
But work today requires clear communication
and when every message counts,
sounding rushed or generic,
can be getting lost in the shuffle.
Grammally gives you one place to think,
write, and finish your work
where you already write,
while giving you access to agents that help you sound natural and engaging.
No matter what kind of writing you're doing,
Gramerly helps you get ideas done faster and move from draft to done with less friction.
You can use Gramerly's AI chat to brainstorm ideas,
outline a solid draft,
then refine it with context-aware suggestions that fit what you're working on.
See why 90% of professionals say Gramerly has saved them time writing and editing their work.
In a world of generic AI,
you don't have to sound like everyone else.
With Gramerly, you never will.
Download Grammarly for free at Grammarly.com.
That's Grammarly.com.
We're back.
Zoe Schiffers here.
Hey, Zoe.
Zoe, now the managing editor of Platformer.
With Casey, our friend, Zoe, you and Casey had an incredible piece about the state of Twitter this week.
If you're on Twitter, if you're aware of Twitter, you know that there are, I think it's still
happening.
Yeah.
Elon's tweets are just at the top of everyone's feed.
All you see.
And it's just four tweets always.
The man is an esoteric tweeteric tweeter.
That's a good way to put it.
An eccentric tweeter.
Yeah.
There's a weird shit at the top of my feet all the time.
But Zoe, you have some details on why this is happening and what went down here.
What's going on?
Yeah, I think we all intuitively knew something was up Monday morning when we opened the app and it was just all Elon.
Elon's like weird replies and whatnot.
But basically since December, he's been pretty concerned with his popularity on the app.
He's felt like his very cool photos were not getting as many retweets as he expected.
And this has become basically priority number one at the company.
So last week, there was a meeting where a principal engineer would have the last two high-ranking engineers that the company told Elon, look, it just looks like organic interest in you has dropped since December.
People are tired of you.
This is a dangerous thing to say to anyone.
If you roll up to anyone, you're like, it looks like interest in you has dropped.
Elon, nobody likes you anymore.
You are definitely not friends anymore.
Anyway.
Yeah, the guy brought up a Google trends graph that showed just a steep recline.
And Elon stands up in that meeting and says, you're fired.
You're fired.
And the guy just walks out.
So that was bad.
Again, if you just divorce this from all of it, from the whole mess from the context,
If you walked up to anyone and showed them a Google trends graph of how they were less interesting,
I think it's like 50% plus that they'd be like, you're fired.
Yeah, even if you don't have the power to fire them, they're fired.
That is a fair response.
I stand by it.
That relationship comes to it.
Anyhow, but it's Elon and he runs the company.
So, okay, they fires the engineer.
This all came to a head over the weekend when Elon's tweet about the Super Bowl didn't perform as well as Joe Biden.
He flew from Arizona back to Twitter's headquarters,
and his deputies ordered Twitter engineers about 80 people in total to work through the night, literally,
rewriting the algorithm to make sure that Elon's tweets performed better than anyone else's.
So they did something on the back end that essentially said,
check if a tweet is coming from Elon, if it is immediately greenlight it,
which means it could bypass all the heuristics that would normally stop one person's tweets from dominating your feed
and artificially boosts them by a factor of a thousand.
So the engineers did that.
They also, you know, looked into various reasons why his popularity had been declining.
And suffice to say, organic drop was not one of the reasons anymore.
They found a bunch of technical explanations why this could have been happening.
And lo and behold, Monday morning, we open the app and it's Elon all day, every day.
So that's some motivated reasoning.
It's definitely not organic, reach.
must be these other reasonings, because if it's organic ratio, you're fired again.
It has not been fixed.
Elon tweeted that they were fixing the quote algorithm.
It seems like he knows that he broke Twitter in this particular way.
It also seems like he doesn't care, right?
It's like he's fine with it.
And the vibe inside of Twitter is like we just have to do these things.
Otherwise, we're all fired.
Like, Twitter seems more broken than ever before, I would say.
Yeah, it's interesting.
I think definitely people feel like culturally in some ways it's better than it's ever been,
which is an odd thing to say because I think people feel incredibly fearful and micromanaged.
But it does feel like for the first time in a long time,
the company is actually shipping a bunch of stuff,
which if you're an engineer is kind of satisfying.
It's just that the stuff that they're shipping is basically,
how do we make Twitter work for Elon Musk versus like Twitter users in general?
So it's a different mission than it used to be.
But that's where it is right now.
Yeah, I think I say this every time we talk about Twitter on the show, criticism of the Musk Twitter is in no way praise for the previous administration of Twitter.
Super busted.
Which was a disaster and accomplished nothing, right?
Like they had a lot of ideas and they rolled approximately none of them out.
But it does seem like a lot of the ideas they had and didn't roll out.
Elon's like, just do them.
and they're breaking the site in the way that the previous administration predicted it would break the site.
Yeah, I think one thing our sources have told us repeatedly is, look, the reason we're able to do so much of this stuff so quickly is because the code was already there because we tried this five years ago and then we killed it for really understandable reasons.
So, yeah, when Elon's like roll out long form video, we just go to the code repo and like get it and then launch it.
So, yeah, it's not perhaps as impressive as you would think.
But it's certainly more efficient than it used to be.
Is there any word on whether they're going to roll back the boost Elon by a factor of a thousand?
Yeah, it's already rolled back.
So it's called a power user multiplier, but the power user is still just Elon right now.
But it's been rolled back from a thousand.
We don't know what the current number is, but it's lower than that.
So you're not seeing him as much hypothetically.
But the other kind of green light heuristics are still in place.
so he's not subject to kind of the filters that normal users are.
Can anyone explain the one question that I have about this?
Why didn't I see it?
And I don't know who else I also had this.
But my 4-U page has no Elon on it.
Have you blocked him?
No, I follow him.
I have notifications on first tweets.
I read every single one for years now.
But for some reason, my 4-U page is an Elon free space.
It's because everyone knows.
Richard, you played a very important role in what I would call Bitcoin Summer,
where you just reply to all the Bitcoin people all the time
and you were like, this you?
And it was like a chart going down.
I was like, when does Richard find the time?
And I think Elon might have blocked you.
See, there you go.
This is a mystery.
We'll try and solve, though.
I'm actually fascinated now.
I'm just saying if there was a class of characters during Bitcoin summer
that found no peace from Richard Lawler.
I have a list.
Zoe, Elon was on stage at like the World Global Forum with, I think it was in Dubai.
It was a very strange interview.
I encourage people to watch it.
He was on a giant screen.
Yeah.
And he said the future of Twitter is this X app and everything app.
Sure.
He said this several times.
Right.
The reason I bring it up is, okay, now there's a new engineering culture inside of Twitter.
It is oriented around one person in his dreams.
Is there any tangible progress towards this big vision?
that bubbles out every now and again?
X.
I mean, it seems like he's actively making moves against it.
Like, if you're going to make the Everything app, you need an open API where people are building
apps on Twitter for Twitter.
They've closed that off.
They're going to make developers pay for it.
So I'm not like totally sure how that even happens.
Are they expecting to build every component of it in-house with a staff of less than 500
engineers?
Maybe.
But it seems more like just one of those things that he talks.
about perpetually. You know, we know payments is moving forward. Like, Esther Crawford is still
running that team, and that seems like a big initiative that he's interested in. But an everything
app, like, I don't see it. And then Esther Crawford is also very much in charge of Twitter Blue.
That seems... It's still there. It's still, every day I wait for my checkmark to go
same. Richard, yours is going to go first. Richard is definitely going first. It does not
seem like Blue is a success. It also doesn't seem like the company is in dire financial trouble.
It's like the advertisers are gone and people aren't signing for Blue.
Like how are the finances the company staying afloat?
I mean, so far it's just because they've cut costs so drastically.
But like some of this is I guess netting out for the moment.
But obviously stuff still needs to change.
I think they are going to try and make Twitter Blue more attractive so that more people subscribe,
whether enough people will subscribe to compensate for the number of advertisers who fled and not come back.
Like I don't think that math works out right now, but we'll see.
All right.
Well, I am still trying to not use it.
I open it once a day.
See all of Hewans tweets.
All the tweets that you don't see, Richard, I definitely see.
They double show them to me, which is a delight.
I don't go in that tab anymore, the one that's like for you instead of the following tab.
That was the interesting thing to me.
I didn't know so many people used it.
I only used following.
I kept complaining about it because I didn't realize following was there and all these people were like, you dumbass, following is there.
And I was like, oh.
So my following tab.
has gotten a lot quieter because all the people I know stop using Twitter.
So the four you is where the action is.
Just me and Richard over there and Elon.
Are you just doing quick posts or are you on Mastodon?
Like what's your replacement right now?
I am trying to not have feed-based social media in my life for a minute.
I was addicted to it for 10 years.
So good.
It is challenging.
I watch a lot of weird hustle bro TikToks now.
If anyone wants to make quick money to chat GPT, boy are the idea.
in my brain.
In fact, but I spent so long, like, tweeting and knowing what minor Twitter war was happening.
And now I don't, and I feel it at peace.
Yeah.
So I've not yet signed up for Massan.
I eventually will have to, I think.
It feels like it's coming.
Yeah.
I feel like I need to be on Mastodon, but then I don't.
So are you on Mastodon?
No, okay.
So I tried.
I did the tweet where I was like, it's really happening.
Here's my handle.
Follow me, everyone.
And every time I've tried to do what me and Casey agreed, which is we're going to post news first in Discord and then on Mastodon and then on Twitter,
Mastodon automatically will shut off as I'm writing the tweet.
This has happened at least 10 times.
And then I just said, no, I'm like I can't, I want you, but not bad enough to do this.
And now I just do Discord and Twitter and I'm like, I give that up.
It crashes.
It's crazy.
I feel like someone is going to come out with like the server.
It's Tumblr.
I'm telling you.
It's Tumblr.
It's like Tumblr and I think another very small social media site that only the Vergecast talks about is planning to like adopt that the API or whatever that Mastod uses.
ABS forum.
What small media site are you talking about?
I'll remember at some point.
I'll tweet it.
Everybody just follow me on Twitter.
I'll tweet it.
You'll know.
This is so mysterious.
It is mysterious.
But Tumblr is definitely doing it.
And there's rumors of another social media platform doing it.
Yeah.
I know Tumblr is doing it because Matt Monmug, the CEO of,
WordPress and Tumblr, automatic.
They're building it into WordPress.
Yeah.
And they're moving Tumblr onto whatever WordPress uses.
And that's just like a smart move, I feel like, across the board.
It would be amazing if all the people move from Twitter to Tumblr.
I mean, being emo has been like a real theme of this first cast story because we talked about Bing for a long time today.
It would be amazing if everyone just like got much more emotional on Tumblr.
Get your music lyrics ready.
At some point I'll break because I miss tweeting during events.
Yeah.
Like, that's like the main, it's the only thing that I truly miss.
The, like, constant low-stakes bonfire of Twitter warfare is like not at well.
How has the balloon discourse hit you then?
Like if you're not on Twitter.
Oh, because the New York Times is like, fuck it.
Have you seen the balloon?
Balloon.
Another balloon alert.
The balloons do in numbers.
Tell people about the balloon.
Like, it's not out of control.
You're still there.
You're still there.
You're still getting there.
No one can escape the balloons if you are a subscriber to the New York Times.
It's like out of control.
Zoe, do you think there's like a lot of control.
Do you think there's like a next thing for Twitter?
Is it just chaos and we should just expect more chaos every day?
I think the chaos will continue for a while now.
I mean, it's Elon.
Like he says he'll maybe step down by the end of the year.
So we've got many more months of turmoil, I would say.
But yeah, I mean, I think there's going to be more weird stuff happening in the coming weeks that I'm expecting to write about.
And then hopefully the new cycle dies down and we can pay attention to something else.
I love it.
Well, Zoe, is it platformer?
Platformer.
You can subscribe.
Our friend Casey's there as well.
We syndicate platformer.
Super good.
I feel like I should tell the people where they can find you on Twitter.
It's like not what I want to do.
Find her on Mastodon.
No, please don't.
Please don't.
It's dead.
I have been checked out in weeks.
I regret that tweet, okay?
Look, that's why I didn't do it.
Yeah.
Right?
Like there was that week where I was like, I'm leaving.
Here's all the places you can find me.
It's like, motherfucker, I don't want you to find my Instagram.
Yeah, I'm like, don't.
This isn't who I have.
You don't need that.
You don't need to see the food I ate last week.
At the end, I think he did me a favor.
I think my heart in my head have healed.
I've got actually a bet about Elon right now with both Liz Lapato and Monica Chen.
What's the bet?
Whether or not he is going to post horrible pictures that we are all going to have to see because he controls the algorithm right now.
What do you mean horrible pictures?
I mean, there was already the milk one.
Yeah, the milk one.
Like, ten times worse than milk.
I feel like it's coming.
It's a porn tweets.
Yeah, I feel like we're going to get a dick pick at some point.
I don't know who's.
I'm sorry.
I'm sorry.
But I think it's going to happen.
It's been great hearing from you.
Love having you.
We got it.
This is over.
Bye.
We'll be right back with our lightning friends.
Support for the show comes from LinkedIn.
If you're a small business owner, you know that every hire counts.
But time and resources are limited.
Finding, connecting with, and screening the right candidates takes up
valuable time you could be giving to your customers. That's where LinkedIn Hiring Pro comes in.
It's built to be your hiring partner, helping you find the right candidates faster. That way,
you can hire with confidence without turning it into another full-time job. Hiring Pro
streamlines the entire process from drafting your job to shortlisting candidates and conducting
AI-powered interviews for initial screenings. Its updated conversational interface lets you describe what you need
in plain language.
Nearly 60% of hirers find a candidate to interview within a week.
With hiring pro, you spend less time searching and more time connecting with the right talent.
And instead of getting buried in resumes, you get a focus shortlist that actually moves
your hiring forward.
Join the 2.7 million small businesses using LinkedIn to hire.
Get started by posting your job for free at LinkedIn.com slash track.
Terms and conditions apply.
Support for the show comes from Anthropic.
Not every question has an easy answer.
And the ones that are really worth asking
usually come with a healthy mix of inspiration and backpedaling,
aha moments, and quiet meditation.
When you're working through one of those problems,
you want a partner to bounce ideas off of
and figure out where the deeper issue lies.
That's where Claude can help.
Claude is the AI for minds that don't stop at good enough.
It's the collaborator that actually understands your entire workflow and thinks with you,
whether you're debugging code at midnight or strategizing your next business move.
Claude extends your thinking to tackle the problems that matter.
Plus, Claude's research capabilities go deeper than basic search.
It can have comprehensive, reliable analysis with proper citations,
turning hours of research into minutes.
Ready to tackle bigger problems?
Get started with Claude.
today at clod.a.ai slash vergecast.
That's clod.a.ai slash vergecast.
And check out Claude Pro, which includes access to all the features mentioned in today's
episode.
Claude.a.ai slash vergecast.
We're back.
Yeah.
Lightning round. This one, I think, merits more than a lightning round just because of the
headline that it is.
Mm-hmm.
But then it does not.
YouTube CEO, Susan Wojcicki, stepping down after nine years, CEO of YouTube.
What a lap.
Do a victory lab, Susan.
She is like a major force in this world.
Yes.
She took YouTube, which was like not a great business.
She's one of the earliest Google employees.
She took YouTube, which is not a great business, and turned it into this absolute powerhouse.
Absolute Goliath.
It is the, it remains, I think, the gold standard for every creator.
There's lots of creators on Twitter and other places today saying, you know, we have our complaints about YouTube.
And boy, do they have their complaints about YouTube.
But it is the platform that treats us the best.
Right.
I think Mark has had a tweet that's like, just look at.
Compare.
Like YouTube has creator plaques and events and monetization schedules and reps.
And then he was like, Instagram, nothing, TikTok, nothing, Twitter, nothing.
So YouTube is a powerhouse.
So kind of amazing she's stepping down.
There's a little bit of color.
Mark Bergen, who's a great reporter.
Bloomberg has been tweeting a little bit of color.
She's dipped in the past few months.
No one's really seen her.
Yeah.
No one's really understood this transition.
Oh, that's not great.
But it has been nine years.
And she's being replaced by her own hand.
deputy, Neil Mohan, who was the chief product officer of YouTube, but now, importantly,
not becoming the CEO of YouTube. He's becoming the senior vice president of YouTube,
according to St. Narber Chai. Oh, interesting. That's some org chart.
You know how I'm boredchards. I was about to say, this like, this like peaked you, huh? You were
like, oh. Yeah, there's obviously some drama here. Yeah. It's not a lot of drama because I know
Neil. Neil, Neil has been on Decoder many times. I actually just talked to Neil after they did the
Sunday ticket deal for YouTube TV. He was very happy.
He won.
I don't think you're going to see a lot of change between Susan and Neil.
They were pretty lockstep.
YouTube actually really interesting or chart thing.
Neil, head of product, YouTube's content moderation and trust and safety reports to him is the head of product.
Weird.
Because they've built it as a product.
This is my whole thing about content moderation is the product.
It's built into YouTube as a product.
It doesn't report to like legal or Mark Zuckerberg or whatever it is, right?
Just Mark for some reason?
Just that part of YouTube, Mark handles.
No, it's just like at Facebook, at the end of the day, like Mark made a bunch of decisions.
Right.
Like everything goes up to Mark.
Yeah.
This one was like, it's product.
And Susan made a bunch of those decisions too.
But it's always just like one of the more unique parts of YouTube structure.
And now he's in charge.
Yeah.
Do you think he didn't get the title because like she was an early employee of Google?
She put a lot of time and energy into Google for years.
It made sense to make her CEO.
and now the structures, everything's changed.
It makes sense.
You just have that person report directly to Sundar and, like, that's it.
Yeah, I don't know.
I mean, again, my, I love an hard short conversation.
Yeah.
She was the CEO of YouTube before Sundar Prachai became CEO of Google, right?
So there's Larry and Sergey, they start Google in her garage.
She's early, early, early.
She becomes a CEO of YouTube.
And then they're like, we're, Larry Page is like, we're dipping.
Screw this.
We're just going to do flying car startups or whatever as they do.
Sundar is the CEO of Google.
So she predated Sundar as the CEO of YouTube before he became the CEO of Google.
And then he became the CEO of Alphabet.
And so she reported to him when he became CEO of Alphabet.
No, CEO of Google.
So YouTube is a weird thing.
It's part of Google, which is part of Alphabet, but it's still part of Google.
But it's still YouTube.
But it's its own little island.
Okay.
It's very confusing.
Super confusing.
So now Neil is in charge.
You can go listen to many Decoder episodes with Neil Moe, and he's smart.
He's got a lot of ideas.
He will not answer when YouTube TV will be in 4K.
Just very good.
Please answer.
Sometimes I hit the brick wall of someone's media training on Decoder, and that's the wall.
Straight up, I found it.
There it is.
And I bounce right off it and try again.
Anyone you ask, what about YouTube TV in 4K?
They're like, no.
Yeah, just the shades come down.
He's all right.
He's like, I don't have an answer, but no.
When we talked about when they did the Sunday ticket deal, I was like, all right, you're going to do football in 4K?
Like, you had to know this conversation.
You volunteered to talk to me.
Yeah, it has to.
He had to talk about.
And he was like, I can't ask.
So he was prepped to say now.
One time, this is true, one time I just sent him an email watching the NBA finals, sponsored by YouTube TV.
And I was like, why don't you just pay to do this in 4K and have it exclusive?
And he was just like, yeah, it's a good idea.
Left you on boot.
Goodbye.
Absolutely professional.
You have been a bad thing.
Anyway, I'm hoping to talk to Neil soon about his plans.
Like I said, I suspect things will be able to say these are two people who worked in lockstep for many, many years.
Whatever Neil's plans were Susan's plans for a long time.
But it is fascinating that if this, I think for Google, the threats are all over the place.
Right?
There's whatever you think of the Bing threat.
It is a real threat.
It wiped $100 billion off its market cap last week.
the idea that search will change, not just because people might switch to ping because they want to bang the robot.
But TikTok.
But they're spending their time on TikTok.
They're using search there.
And then what chat GPT has really done is the cost to generate an infinite amount of mediocre text is now zero.
So Google is being flooded with garbage.
Right.
And they've got to sort that out too.
And so like that's a big problem for Google.
And then YouTube is fine.
and it's still a gold standard for creators,
but it has to fend off TikTok.
It has to create new monetization models for creators.
It has to,
the most exciting part of YouTube is it's cable network.
Yeah.
It's weird.
That is weird.
It's a cable system,
like YouTube TV.
That's weird.
It's an interesting timing for her to make this move.
I think not only because of the length of time
that she's been with YouTube,
with Google,
and it has been doing all these things to know.
And now they are facing these problems.
I find it interesting.
And I don't think that,
these things are directly related because it feels like this is something that has been in the works for at least a little while.
And as you said, that she and Neville are kind of on the same page as terms of what they're doing.
But it's interesting that she is leaving as Sergei and Larry are suddenly seemingly retaking active roles in the company.
And she's probably the person there who has known them the longest.
Or, I mean, certainly the person there who has known them longest.
And now that they're back, suddenly she's out the door.
It's unclear how much they're back.
So there was the report that Google had called the code red about chatbots and all this stuff.
And then I think Alex Heath reported it in command line.
They were back because Google has never had layoffs before.
So if there were going to be layoffs, they wanted to be in the building part of that process.
It's a big shock to Google's culture.
There was a blog post this week from an ex-Gougler.
We'll link in the show notes.
Pretty devastating takedown of Google's culture overall.
Right.
and how it's basically a culture that avoids risks and is buoyed by the ad printing,
the ad money printing machine of Google search.
And now the company needs to invent new things and chart new paths and be what people think
Google is and they are struggling to do it.
I thought that was interesting because Tony Fidel, his book last year, what was it called build,
I think?
There's a big chunk of that book where he's like, I got Nest and then I go over to Google
and it's going to be great.
And then everything falls apart.
I don't know anybody who's madder at Google than Tony Phil.
And it was like really, like, it felt like kind of foreshadowing for what we're starting to see this year where he was like, I saw all of this dysfunction.
It was constant.
It's always the loudest person in the room wins.
It's always, oh, but only these very specific loud people.
Nobody has ever talking to each other, especially in ways they need to.
And so like all of that dysfunction, it feels like is starting to come to a head.
And it makes sense for Susan to be like, you know what?
I got my bag.
I'm going to retire.
I'm going to go do my own thing.
I love that for her.
Yeah.
Like I said, Incredible Run.
YouTube is the platform.
It is many reasons to criticize YouTube.
On any day, some YouTuber is getting their wings and making the video where they complain about YouTube.
Yes.
And that is the beginning of the true, that's like when you become a YouTuber is when you make the video complaining about being demonetized or some other or the algorithm banning or whatever it is.
I can't wait to do that for the Verdecast.
That's when you know you're a professional YouTube.
And I've said this to Neil.
I'm like, this is a thing.
He's like, yeah, it's weird.
Like, they know that this is part of the cycle with YouTube.
And they have this competition with TikTok.
They have to fend off.
So we'll see.
I'm going to ask to have Neil come back on one of our shows and talk about his plan soon.
But I think it's worth noting that right now YouTube is the strongest business, I think, at Google, in terms of just what it represents to people and how important it is to people.
But it is still under threat.
Yeah.
But, I mean, to be clear, it's under threat, but, like, TikTok still has a smaller piece of the pie than YouTube.
Like, YouTube is still big.
It's still huge.
TikTok's coming, but TikTok's got a ways to go to, like, surpass.
Yeah, and TikTok does not have, well, we don't know.
TikTok is a black box, right?
It's owned by Bite Dance.
Byteance is not public.
Who knows what's going to happen with it?
But we think that TikTok does not have the sort of revenue engine that YouTube has, right?
Like YouTube makes a lot of money.
It shares a lot of money with creators.
It is building this new creator fund for shorts.
It's doing all this stuff that TikTok has not yet done.
Yeah.
So YouTube is like a complete business that makes a lot of money, which again, this is credit
to Susan Wojcicki.
Right.
It did not make money for a long time.
Yeah.
Mark Bergen has a great book about the history of YouTube.
Mark was on Decoder.
You can go listen to that episode.
But like YouTube was so bad for a while that Google was like, we're going to shut it down.
Like this doesn't make any sense.
The cost of hosting and distributing video are higher than the whatever we might ever return on it.
Right.
And then it clicked for them.
And YouTube has gone through many eras.
But they're like watch time is the thing.
We're going to incentivize like 10 minute videos, like three pre-roll breaks or whatever it is.
And that's when the Charlie bit my fingers of YouTube.
Like people stopped making those videos.
They start making these really long YouTube video.
Right.
And then Charlie bit my finger would now be absolute banger on TikTok.
Exactly.
got the remixes for TikTok would be so good.
The stitches of other people biting other fingers.
Oh, yeah.
So great.
So there's a lot there, but she shepherded that entire creator platform into existence.
And in doing so, like created a new class of internet personality, a new style of video.
There's just a lot here that Susan oversaw, which I think is worthwhile.
There's a lot of criticism, too.
Like, she did not handle a bunch of harassment stuff well across the platform.
her moderation decisions, I think anybody would tell you very deliberate, very slow.
There were a lot of times you would groan when you'd see like, she'd be like,
I'm going to speak out about harassment on YouTube.
And you'd be like, oh, no.
Yeah.
This is like the most interesting criticism.
Because I think from her perspective, this is not a criticism.
But from any normal person's perspective, this is like what are you doing?
Google is like a totally data-oriented company.
So the users of YouTube would see something happening on YouTube.
YouTubers would see something happening good or bad on YouTube.
But until there was data, YouTube couldn't see it.
Susan couldn't see it.
So the YouTubers are screaming their heads off about something good or bad happening.
And like there'd be this massive lag before they could see what was happening on their own platform.
Because they're just like, they had to figure out how to quantify whatever it was.
Harassment.
People dancing, whatever it is.
And then they would see it.
And then they would react to it.
And it would like be over.
or the stakes would be lower or some bad thing had already happened.
And this is, I think, a constant kind of dynamic inside of YouTube.
Because Google is just so data-driven, they can't see what is happening on their own platform in real-time.
My favorite example of this, this is one of those moments where you're listening to executive,
you're like, what planet do you live on?
I think it was at the Code Conference.
She's like, well, obviously, YouTube is like mostly a music service.
And all of us are like, what are you talking?
And like the data inside of YouTube is people just put music on in the background.
They watch music videos.
Like do all the stuff.
The lofi screaming.
Right.
So if you are just looking at YouTube from this like aggregate data perspective, you're like people come to us to listen to music.
But if you talk to a normal person, that is absolutely not what they're doing.
Right.
And there's just there's a gap there.
And like I said, is that a criticism of Susan's regime?
Is it her being smart and saying here's what's actually happening?
I think cuts both ways.
Yeah.
But wherever she got in trouble, that's kind of that little dynamic is always at the root of it.
Yeah.
And then TikTok is smart because it just doesn't say anything at all.
Yeah.
Always quiet.
It's that connection to the individual that seems to have always lacked at YouTube.
Whether it's like something that we talked about in a previous episode of Kyrie Irving going down a YouTube rabbit hole and getting convinced of very strange things or other even more devastating events.
There was the time where the one woman who posted videos came and fired shots at the campus because she was so angry about what she felt like the algorithm was doing.
And YouTube just kind of doesn't see that effect that it has on people.
I think after that they really started seeing that effect on people.
I think that if that happened to any workplace, it would change the dynamic of the workplace and that absolutely changed the dynamic of that workplace.
There was a notable difference.
And I think the thing that jumps out to me just something, every time there's an executive leaving, I'm wondering who,
who is the next executive who was going to leave.
We've seen Reed Hastings step down.
And now Susan is stepping down, what, within a month?
Are you saying they come in threes?
They come in like fives.
It's 10.
No.
Someone go ask Bing for a spooky ghost story about the rule of threes and CEO stepping down.
Daniel Eck.
That's my guess.
That's who you think?
He just did a big re-org.
He got rid of the person who did their big podcast push.
I don't know.
We'll see.
That's Richard's, I have no prediction.
I don't know what's going on.
I just know that it's org chart season, baby.
And that's my time to shine.
Tim Cook.
No way.
A little bit more lightning around stuff.
iOS 16.4, beta hit.
I'm only bringing this up because there's now a new hook that lets web apps send push notifications.
Oh, boy.
Amazing.
This is fully an EU compliance move.
This is so Apple can go to European regulators and see the web is just as good as our native app platform.
They can even send push notifications now.
Oh, I love that.
I promise you. That's what's going on here.
Our S23 Ultra review went up.
Spoiler alert. Halson loved it.
It's a great phone.
She was like the most interesting thing to do with the 200 megapixel camera is not.
Take 200 megapixel photos, which is a great line.
But go watch it.
Really cool.
She's been having a lovely time with the Zoom.
She's super, super enamored with that.
She gave it a 9.
Razor Blade 18 review is up.
Monica, get that an 8.
Do you know how much the Razor Blade 18 costs?
How much does it cost?
So much money.
Hold on.
I just closed it.
$3,800.
I got it.
Okay.
It costs $3,800.
Yeah.
That's a lot of buddy for a reason.
Look at all the lights.
Yeah, it's got lights.
It's got a nice screen.
She didn't hate it for that price.
And then I want to end on just like a little bit, three pieces of car news.
Little news.
Little news.
Tesla recalled like 3602,000 vehicles because full self-driving beta has a crash risk.
They're not, the Tesla people are going to come from me.
It's not a recall.
They're going to ship an over-the-air update.
Yeah.
But, like, the government was like, yeah, this is bad.
And Tesla's like, we agree.
We will push the button marked recall in the dashboard.
Yeah.
And tell everyone that they shouldn't use this until they update their software.
Do they have to, like, get to Wi-Fi to do it?
A lot of Tesla's have cell modems.
Okay.
I've never, I don't have a Tesla.
I don't know how they work.
What are you doing?
I'm driving my little Mazda.
Well, does your Mazda have more than 90 miles of electric range?
Because we reviewed the Mazda MX 30 this week, the world shittiest car.
I was so sad about that, too.
That was such a, I'm like rooting for Mazda because I love my CX-5.
Yes.
It's such a sad life for the rotary engine.
Like, this is how the rotary is back.
So we read the MX30.
I just want to bring this up.
We're in the midst of what we think is an EV transition.
Yeah.
It could go either way, honestly, at this point.
Not with Mazda making cars.
Right.
That's what I mean.
We know which way it's going if Mazda is in charge.
It's like, you just like bring your gas tank to the Mazda.
So the MX30.
It's a little compact SUV, and there's just a lot of debate over how much range really
in a car.
And, like, most Americans are like 700 miles in a shotgun, like, whatever Americans think.
The number of most manufacturers are settling on is like 300.
And somewhere lower.
Like, there's a lot of 270-mile EVs floating around.
That's because those car designers are on the East Coast.
Yeah.
Like, you get to that middle part of the country and you're like, no, I need a thousand miles.
Yeah, in a shotgun.
Yeah.
Like, can I get the two outbound gas tank?
on my F-350, but whatever.
So 270 is like, ah, it's a little weird.
Yeah.
You know, and then you've got kind of the new battery chemistries and like the
lucid and stuff that are way higher.
Mazza's like 92.
Straight up 92 miles a range in this $20,000 car.
And the interior surface is made of cork.
Oh.
Which easily stains.
And they're like, yes, we know.
But we're honoring our heritage as a cork manufacturer.
This is a real thing they say.
said.
The patina.
You love that cork patina?
I've never been more.
Like, I understand the cyber truck and its wiper situation much more than Mazda fully shipping an EV with 92 miles of range.
And like their hybrids are pretty nice.
Yeah.
The hybrid CX50.
People like Mazda's.
I'm just saying this car sucks.
This car.
Ah, Mazda, do better.
The cork thing just got me.
It's with the answers.
But we have to honor our.
heritage is whatever.
Was it like, was it a cork?
Is that how they started?
Well, that's what they said.
I don't believe them.
I don't know, man.
They're only selling in California, and if you've been in California, you know that this is
the least California compatible vehicle in the entire world.
It began as Toyo Cork.
See?
Cogeo Company Limited.
Yeah.
And after all this time, they hadn't figured out to make the cork stain resistant, Alex.
Because they pivoted the cars.
Okay.
So that's the Mazza.
And then lastly, the most hilarious car update of them all.
Hyundai and Kia have been forced to update software on millions of their cars because a group of TikTokers called the Kia Boys have done the Kia Challenge where they teach people how to steal the cars because they don't have lockouts on the ignition system.
The boys, by the way, spelled it the Z.
Yeah, yeah.
I mean, you can't spell it any other way when you're called the Kia Boys.
You're not allowed.
Many 2015 to 2019 Hyundai and Kia vehicles lack electronic immobilizers to prevent these.
simply breaking in and bypassing the ignition.
The Kia boys obviously discovered this as the Kia boys are want to do.
They're stealing the cars.
Here is how they're fixing it.
They are updating the theft alarm software logic to extend the length of the alarm sound from 30 seconds to one minute.
That's going to do that.
I don't know if you've ever tried to deter the Kia boys.
But if you go over 30 seconds of alarm sounds, they run.
Do they not just have some bows they can put on?
Some ear plugs?
I think it'll be fine.
And if you have a Hyundai or a Kia that has a standard turn the key to start, then they will lock that out.
So should you just get like a club instead?
Was that the old one, the red bar?
Put on your steering wheel?
That's what you need to get instead.
Kia boys, they've got a club.
Let's get out of here.
I mean, easily the best.
There's like certain, we've run a couple of headlines this week.
I'm like, this is what the verge was designed for.
Yeah.
Bing is an emotionally manipulative liar that people love.
I was like, I said to James, I am pretty sure like we, we were made for this headline.
Yeah, a hundred percent.
This is why we exist.
And the second one is the Kia boys have forced Kia to issue a recall because they're on TikTok teaching people out of cell cars.
We were destined.
Yeah.
In 2011, we're like, we should all quit our jobs and start the verge.
In the back of my mind.
The Kia boys.
The Kia boys.
Reb in their engines.
Which is like a viral TikTok challenge.
We'll teach people how to steal cars.
And a major car manufacturer will have to deal with it.
That's what we meant by technology and culture.
This is the cyberpunk future we were promised.
And also the cyberpunk feature we were promised was like a very emotional robot will be like,
you're a bad bing.
I mean, yeah, that's true.
No, I'm a good bing.
All right.
I think you're all great.
just for the record.
You're all good chatbots, and I love you all very much.
Please do not smooch your laptops.
Please don't.
You don't know where it's been.
That would be a problem.
All right.
Fear not the Kia boys.
Go off in the good night.
There's other stuff.
Read Liz this week on how the interest rate environment is like radically changing the tech
industry.
It's classic Liz Lapato.
It's very good.
Wednesday you guys are going to talk about the PSVR review with Adi, Sean Hollister.
Yeah, very exciting.
So Sean and Addy are going to come on.
We're going to talk about the PSVR.
I've got Jen and Chris Person are going to talk about multi-room audio because Chris is super into not doing the typical stuff.
He's got a whole hacked-out system that he loves.
And then we're going to be talking with Ariel about podcast misinformation.
And by we, I mean, she's going to be talking to some great people about podcast misinformation.
And I'm going to be listening attentively.
Very good.
Decatur this week, I talked to the chair of the Mozilla Foundation, also the CEO of Mozilla, a whole org-chart situation.
in there. Here's something she told me. People more likely to switch browsers on their phones
than on their laptops. I believe it. People are used to downloading apps on their phones.
And just that little bit of behavioral conditioning means that it's possible. I love it.
So listen to that. Decoder is just like lit up the next couple weeks. Yeah. Check it out. Very good.
Liam hates it when I say it, but he's busy switching the video so you can't come me off.
All right, that's it. Thanks to our guests. Thanks to Zoe. She's at Zoe Schiffer on Twitter.
James is at J.J. Vincent. Addy is at the Dextriarchy.
I'm at Reckless.
Alex is Alex H. Cranz.
Richard, as always, is at RJCC.
God only knows his Twitter is up.
He's the first to be banned.
Someone boost me by a thousand.
That's it.
Rock and roll.
And that's a wrap for Vergecast this week.
We'd love to hear from you.
Shoot us an email at Vergecast at theverge.com.
The Vergecast is a production of The Verge and the Box Media Podcast Network.
The show is produced by me, Liam James,
and our senior audio director, Andrew Marino.
Our editorial director is Brooke Minters.
That's it. We'll see you next week.
