Microsoft is in its AI PC era | The Vergecast

  • 4 months ago
The Verge’s Tom Warren and David Pierce discuss the announcements from Microsoft’s Surface event, including the new Arm-powered Surface Laptop, and Copilot Plus PCs. Verge senior AI reporter Kyle Robison joins the show to chat about OpenAI’s GPT-4o demo and where we’re headed in the next few years of AI. Nilay Patel answers a question about iPads for this week’s Vergecast Hotline.

Category

🤖
Tech
Transcript
00:00:00 Welcome to the Verge cast, the flagship podcast of Arm-Based Chipsets.
00:00:03 I'm your friend David Pierce, and I am currently in the woods,
00:00:06 in the middle of nowhere in Virginia, at an Airbnb.
00:00:10 So, the last seven days have been my wife's birthday,
00:00:13 Mother's Day, and our wedding anniversary. All in the course of a week.
00:00:17 If you get to choose in your life to do all those things at once,
00:00:20 or spread them out through the year, I highly recommend spreading them out
00:00:24 through the year. But this is what we have. So, every year,
00:00:27 we try to combine all these celebrations and just go find a place
00:00:31 in the woods with bad internet and no cell connectivity,
00:00:35 and just relax for a couple of days. I cannot recommend enough,
00:00:39 every once in a while, just finding somewhere where you can't be online,
00:00:42 whether you want to or not, and just lean into it for a couple days.
00:00:46 It's the best. But, I'm in the woods, but the Verge cast goes on.
00:00:50 We have an awesome show for you today. We actually have a lot to get to.
00:00:54 It is like the newsiest time in recent memory,
00:00:58 and there's a lot of fun stuff to talk about. Today on the show, we're going to
00:01:01 talk about two things. We're going to talk about the Surface
00:01:03 event that just happened, where Microsoft kind of unveiled both its new
00:01:08 devices, but also what seems to be an entirely new generation
00:01:12 of laptops and Windows PCs. There's a lot going on. There's some big,
00:01:16 huge promises. We're going to talk about all of it. We're also going to talk about
00:01:19 OpenAI. I know we've been talking about the OpenAI
00:01:22 event from last week, and Google Gemini, and
00:01:24 this whole AI space that we're in right now a lot over the last couple of weeks,
00:01:29 but I can't stop thinking about it, especially what OpenAI believes
00:01:32 is the future, and how we're going to talk to computers going forward.
00:01:36 Lots to get to on that front. We're going to talk to Kylie Robison, our new AI
00:01:40 reporter, about all of it. We also have a question
00:01:42 about iPads on the Verge cast hotline, because we also can't stop talking about
00:01:45 iPads. Lots to do in this episode. All that is
00:01:48 coming up in just a second, but I just heard the hot tub motor go
00:01:52 off, which means it's hot enough, which means it's hot tub time.
00:01:55 This is the Verge cast. We'll see in a sec.
00:01:59 Welcome back. We've been hearing for the last couple of months that a whole new
00:02:03 generation of PCs was about to come out. Qualcomm had
00:02:07 this new lineup of chips called the Snapdragon X
00:02:10 that were supposedly drastically more powerful
00:02:14 and more efficient, and would actually bring the Windows PC world
00:02:17 up to par with the recent Macs that run on Apple's own silicon.
00:02:21 That's a big win on its own, obviously, but also the combination of power and
00:02:26 efficiency changes more than you might think. Not
00:02:29 only do things get better and longer lasting, which is a huge win,
00:02:33 but you can also do different kinds of device design
00:02:36 when you can count on all-day battery life. You can build them with 5G and
00:02:40 other connectivity because people will actually take them away from the charger.
00:02:44 Windows tablets would suddenly make a lot more sense for a lot of people.
00:02:47 So would more portable Windows gaming devices. It all sounds
00:02:51 awesome. It also, I should say, is a thing we have been promised
00:02:55 many, many times. Qualcomm has been claiming that the next one was the one
00:03:00 for years. Everything was going to be the year of Windows on ARM.
00:03:04 I'm skeptical of all of it. Well, yesterday, that new generation of PCs
00:03:08 showed up. Microsoft had a big event to launch new services. Lots of other
00:03:12 companies announced Snapdragon devices of their own.
00:03:15 Our own Tom Warren is at the event in Seattle to catch us up on all of it.
00:03:19 Hi, Tom. Hello there. How's life in Seattle? Yeah, it's good. It's sunny. It's not
00:03:23 raining, so I can't complain. I did appreciate that the weather came
00:03:27 up like three times at the beginning of the event this morning. They were very
00:03:30 happy that it was a sunny day in Seattle. Yeah, yeah. I mean, they put us
00:03:34 all in the tent today, so the last time they did that, it was like mudslides all
00:03:37 around the tent. So I'm glad that's not happened. Oh, boy. Yeah, so tell me a
00:03:42 little bit about this event. This was an event we've known was coming for a
00:03:45 while. It's the day before Microsoft Build, but it's not Microsoft Build.
00:03:49 Give me the vibes of the event a little bit. Yeah, so obviously, we kind of knew
00:03:54 that they'd been working on these sort of ARM-powered PCs. And kind of like the
00:03:58 event sort of talked us through this sort of reinvention of Windows, which
00:04:02 is Microsoft's promised a bunch on sort of Windows on ARM previously. But it
00:04:07 does feel like it's slightly different this time, mainly because these devices
00:04:10 have got more performance and battery life, all that sort of stuff. But also
00:04:14 because all the OEMs are actually doing their own variants of what they're
00:04:17 calling these CoPilot Plus PCs. And they're actually like the top tier
00:04:22 variants of their Windows laptops. So it's not like they're just shoving these
00:04:25 chips into some mid-range laptops and just calling it that. They're like, I
00:04:29 quote, "all in" on these CoPilot Plus PCs. So the event kind of ran that down.
00:04:34 Why that matters, the performance versus the MacBook Air, because Microsoft
00:04:38 thinks that this is kind of like their moment to sort of beat the MacBook Air,
00:04:41 which is kind of interesting. And then some of the AI sort of features that
00:04:44 these new laptops enable. Yeah, there's a lot of confidence coming out of
00:04:48 Microsoft in this event. I feel like normally, I'm used to sort of the
00:04:52 Panos Pane era, where he would come out and sort of tell you how lovely a
00:04:55 device is and try to convince you of how wonderful it is and how much the team
00:04:59 cared about it. And this felt like a very different Microsoft. They just came
00:05:03 out and were like, "We did it, folks. Here's the stuff. Check it out. Let's go."
00:05:06 Was that what it felt like in the room? Yeah, it definitely felt a little bit
00:05:10 different than some of it, because the previous Surface event that they had
00:05:12 last year, it was literally days after we knew Panos was leaving, and that
00:05:16 felt like... Oh, that's right. It was kind of like a funeral to Surface. It felt
00:05:19 like Surface was dying or something. It was a really odd event. This one, it
00:05:24 just feels like the Surface is like the platform, essentially the showpiece
00:05:28 for this platform, for these new advances on Windows and ARM. And that's kind
00:05:33 of... They didn't make Surface the most important thing, but it wasn't the
00:05:37 least important thing. It was kind of just there to sort of back up that they
00:05:41 are serious about this transition to ARM. Yeah, so talk to me about the
00:05:45 Copilot+ PCs thing, because you and I have been on this show for the last
00:05:49 couple of months, I would say, relatively consistently making fun of the idea
00:05:54 of an AI PC, which just generally means nothing. But this time, Microsoft came
00:06:01 out and much more clearly tried to define what a Copilot+ PC is, and what
00:06:06 makes it different, and what makes it special, and why this kind of deserves
00:06:09 its own category of PCs. What makes a Copilot+ PC a Copilot+ PC?
00:06:16 Yeah, exactly. They previously had these AI PCs that were kind of like...
00:06:22 They were trying to sell it on this AI vision of Windows, but they didn't
00:06:25 quite have the performance, essentially, inside for them to actually enable
00:06:28 these AI features that are part of these Copilot+ PCs. So essentially,
00:06:32 what they are is... I think the way Microsoft's defined them is that the
00:06:35 MPU, the Neural Processing Unit, is a new chip. I wouldn't say it's necessarily
00:06:39 new, but it's new to Windows. A new chip that will basically accelerate these
00:06:44 AI tasks. So they've got this new recall feature, which essentially... Say
00:06:48 you're working on your laptop, and you saw an elephant two weeks ago, and
00:06:52 you're like, "Where did I see that damn elephant? Was it in a tab? Was it
00:06:56 in an email? Sure, I saw it." And then you can just type in "elephant", and
00:07:00 it will be like, "Okay, I saw this in your email." And it doesn't need to
00:07:03 see the words that you've typed, like "elephant". It will literally
00:07:07 understand that that is an elephant. So it's quite good for recalling those
00:07:11 memories from your everyday use of Windows. Very similar to a Mac OS app
00:07:16 called Rewind, which does a similar thing. But obviously, this is built-in.
00:07:19 It's leveraging the actual MPUs. It's not hitting the CPU or GPU to do this.
00:07:23 You can just do these whilst you're watching videos or processing, I don't
00:07:27 know, like Premiere Pro or something like that, and it's not going to hit
00:07:30 that performance. There's also live captions, and they showed translation
00:07:34 there, which was pretty cool. But you can't save that stuff. So say you
00:07:38 wanted to watch, listen, or watch a podcast that was in Japanese, you could
00:07:42 literally just load it up, and it would just do that live translation for
00:07:45 you. So that was pretty cool. I have to say, I thought recall was really
00:07:50 cool. And I actually thought, in general, Microsoft did a really good job
00:07:53 of not only explaining why you want lots of power on your computer, but
00:07:59 also showing the kinds of features that you can do on your device when you
00:08:04 have this stuff. The translation was a good example, where you were able to
00:08:06 have multiple languages being translated to each other in real-time on a
00:08:11 video call, which is the kind of thing you genuinely only can do on device
00:08:15 because you're sending all that to the cloud. It just falls apart. And with
00:08:19 the live captions and some of the image editing stuff, they did a really good
00:08:22 job of saying, "This is actually what you need all of that power locally in
00:08:26 order to do." And I feel like, for me, recall, which is, I think if I
00:08:29 remember correctly, all private, all on device, not uploaded anywhere, not
00:08:33 even leaving your computer, really, really smart way to show, "This is what
00:08:38 you can do when you have this thing on your computer." I found that very
00:08:41 compelling.
00:08:41 Yeah, I think recall is their flagship feature for this AI stuff. But they did
00:08:47 also show, which was kind of more forward-looking, and it wasn't a
00:08:51 complete demo experience, but they showed some of the stuff where you could
00:08:55 have this co-pilot experience in a game, like an Xbox game on Windows. So
00:09:00 there was a really cool demo. And I don't know if this came through very
00:09:03 well on the live blog, so definitely check out their demo of it. But it was
00:09:06 essentially, someone was playing Minecraft, and then the co-pilot was
00:09:10 teaching the person how to craft in the game, because it could actually see,
00:09:14 obviously, what you were seeing. So it knew. And that was pretty cool. And it
00:09:19 was like a zombie in Minecraft appeared, and they were like, "Oh no, there's a
00:09:22 zombie. You better run away." It was literally guiding them around the game
00:09:25 as if it was their friend sitting beside them. So that was a really good hint
00:09:30 at where this could go, when you've got all this on-device AI models running
00:09:33 in the background, and it can actually see everything you're doing. And I
00:09:36 thought that one was particularly cool.
00:09:38 That is really cool. On the recall one, we talked when you were reporting on
00:09:43 this, I think it was called AI Explorer was the codename that you had found out
00:09:46 about, is that right?
00:09:47 Yes, that's right.
00:09:48 We were debating whether it was going to seem very cool or very creepy.
00:09:53 Having seen the thing in action, where do you land? Cool or creepy?
00:09:56 I think it's in the middle, right?
00:09:57 Okay.
00:09:58 Definitely, because it is kind of creepy that they are taking basically a
00:10:02 snapshot of everything you do on your PC. They say there's obviously a way to
00:10:05 manage that. You can disable it per app or on particular websites and stuff. But
00:10:10 there is still the fact that it's running at all times in the background
00:10:13 and capturing everything you do. But then again, so is your web browsing
00:10:17 history. That's always captured in your browser, not quite to the degree of
00:10:22 seeing it per se. So I think there is definitely a creepiness to it, but I
00:10:27 wonder if people will see past that for the actual functionality that it
00:10:32 enables. And it's all local. I am speaking to one of their security VPs
00:10:36 shortly to find out how they're going to protect against malicious activity.
00:10:40 Because you can imagine someone getting in through some sort of floor and then
00:10:45 using these AI models to cause chaos on your PC and search for stuff. There's
00:10:50 potential there. So I am curious how they've actually secured it beyond the
00:10:54 sort of vague promises that they had on stage. But yeah, I think it's
00:10:57 definitely in between. I think there's going to be a bunch of people that
00:10:59 will disable this on the hardware that they buy. But then I think the things
00:11:03 that it enables, particularly around that sort of Minecraft demo in the
00:11:06 future, I think it's going to be worth the trade-off, right? As long as they
00:11:10 can keep it secure.
00:11:12 Yeah. I mean, the demos like that where it can collaborate with you as you're
00:11:16 using your computer is like, that's the stuff that both Microsoft and others
00:11:20 show these bits and pieces of that I'm always like, "Oh, I get it now." Like
00:11:23 that's the kind of stuff I actually want AI for, not to write me business
00:11:29 emails, but like help me do the things I'm trying to do better. And in a sort
00:11:33 of automated, proactive way, like that's pretty cool.
00:11:36 Yeah. Being assistive and sort of like helping you along your way is
00:11:40 definitely the way that I prefer to see AI go than try and sort of replace
00:11:45 you, essentially.
00:11:46 Right. Totally. So let's talk about the devices. So the two big new ones from
00:11:51 Microsoft, the Surface Laptop and the Surface Pro, no numbers, which really
00:11:56 throws me off. It's just this, no. Are there numbers?
00:11:59 They're sneaky numbers. So the Surface Pro is called the Surface Pro 11th
00:12:03 Edition, which is very Microsoft, as you can imagine.
00:12:06 It is very Microsoft.
00:12:07 They're kind of just referring to it as the Surface Pro, right? Or the
00:12:10 Copilot Plus Surface Pro, if you want to get even more Microsoft. But yeah,
00:12:15 essentially there are no numbers.
00:12:16 Okay. You've seen them both. I would say my takeaway just from reading the
00:12:21 live blog and seeing the specs and stuff is basically similar-ish
00:12:26 footprints. I think the Surface Pro in particular is exactly the same size as
00:12:30 before. And the whole pitch is like gigantic, massive internal upgrade that
00:12:34 kind of changes everything. Is that reasonable?
00:12:37 Yeah. I mean, on the Surface Pro, they've got a new OLED display model,
00:12:42 which is nice. It's good to have OLED on there. It's not tandem like the iPad
00:12:46 Pro. I think that Dell's XPS actually is. So you've got a new display option
00:12:51 there. But yeah, overall, the actual device looks very similar, but they have
00:12:56 done one really cool thing that I like is the type cover. Now, you know from
00:13:00 reviewing these for years, the whole lappability argument with the Surface
00:13:04 Pro has raged for like a decade now. And what they've done is they've
00:13:08 reinforced it. So I couldn't really get it to bounce, honestly, when I was
00:13:13 using it on a desk.
00:13:14 Really?
00:13:15 Yeah. You know how it usually flexes and you get that weird bounce? It's
00:13:19 really annoying when you're trying to write a long document.
00:13:22 Yeah. As soon as you start typing fast, the whole thing sort of feels like
00:13:24 it's wiggling underneath you.
00:13:26 Yeah. It doesn't do that anymore. It's still a very slight flex, but it's not
00:13:32 really as bad. And I tried it on my lap as well. It was the same. And it's
00:13:36 because they basically reinforced the base of the Surface keyboard. They put
00:13:40 a battery in it now, so the actual trackpad is a little bit larger and it
00:13:44 has haptic feedback and it has a Bluetooth chip inside there. So you pull
00:13:48 it away from the display and you can actually use it wirelessly. So it's not
00:13:50 like just this dumb keyboard when you pull it off anymore. So some pretty
00:13:54 cool keyboard upgrades. And you can also use the old keyboards if you want
00:13:57 to. Yeah. And that and the OLED display, I'd say, are the biggest things that
00:14:01 you would notice on the outside. It's obviously the inside is they're
00:14:04 actually Qualcomm powered now, so it's Windows on ARM. The Snapdragon X Elite
00:14:08 and the Plus in the Surface Pro, depending on which model you go for. And
00:14:12 playing around with it for a good 10, 20 minutes, it just doesn't feel like a
00:14:18 regular Windows laptop, right? Especially when everything is running native on
00:14:22 there. We really need to review these to check out these big claims that
00:14:26 Microsoft's making. But yeah, it feels super slick and fast and speedy, as you
00:14:31 would kind of expect. I don't think you'd necessarily think that this was
00:14:34 an ARM laptop, put it that way. That's really exciting. I mean, and I think
00:14:37 that sounds, as you say it, almost like faint praise, but I just want to make
00:14:41 sure it's really not faint praise. Given what we've seen from Windows on ARM in
00:14:47 the past, you could tell on the last generation Surface Pro, whether you were
00:14:52 using the Intel model or the Qualcomm model, like immediately. Like
00:14:56 immediately, you could tell the difference. And so just the fact that you
00:14:59 pick it up and it just feels like a Windows laptop is already an enormous
00:15:03 win.
00:15:04 Yeah, exactly. Like it does. I couldn't get it to like really lag, you know?
00:15:09 And all right, I wasn't really pushing it. We will like, I say faint praise
00:15:12 because we have to check this out when we review it. But from everything I can
00:15:16 see, it does look like they might have done it this time. But it really is
00:15:19 going to come down to the app compatibility. And they've made some big
00:15:22 promises about new emulator. And I think just the pure fact that these chips are
00:15:26 a lot better. They're sort of brute forcing their way through this in many
00:15:30 ways, but they're not doing it in a way that's going to kill the battery life
00:15:32 either. Which is interesting because they're making big claims about that as
00:15:35 well. It's like 16 hours browsing the web, like 20 plus hours if you're
00:15:39 watching local video. So if that actually holds up or anywhere close to it, then
00:15:44 yeah, that's going to be very competitive compared to sort of the MacBook Air.
00:15:48 Whereas usually when Microsoft says they have like a 10 hour battery life PC,
00:15:52 it's usually five hours, you know?
00:15:55 Well, and you got to see, if I remember correctly, like you actually got to see
00:15:59 them run these demos next to a MacBook Air, right? I find Microsoft's whole
00:16:04 fascination right now with the MacBook Air and the M3 chip in particular,
00:16:07 really fascinating. But just the fact that they put you in a room and tested
00:16:12 the MacBook Air next to the new surfaces is just wild to me.
00:16:17 Yeah, I spent an hour watching a ton of demos. That's ranging from like
00:16:22 Cinebench and Geekbench, you know, like the sort of the basic PC benchmarks,
00:16:26 all the way up to like, they made scripts for like doing Photoshop filters,
00:16:30 which they showed some of those on stage. And then some AI tasks. And yeah,
00:16:34 I'd say out of that hour, probably at least every five minutes, the MacBook Air
00:16:37 was mentioned. I'm not joking, I'm not over exaggerating. It was literally
00:16:40 relentless.
00:16:41 Why is Microsoft so obsessed with the MacBook Air?
00:16:44 I think because they know that they've fallen behind that like premium set of
00:16:49 devices over the past few years since Apple transitioned to the M1. They've
00:16:54 really been behind it. And Intel hasn't like got them there on battery life or
00:16:58 performance to sort of match that kind of great offering from Apple. So I think
00:17:03 their confidence is like, yeah, we're finally here. You know, we can be
00:17:06 competitive or if not better, and we can do it with all these AI sort of
00:17:10 features on top. There you go, Apple. You know, we're here. Yeah, so it's
00:17:14 definitely a new level of confidence.
00:17:16 Yeah, that makes sense. And speaking of that, actually, I was very confused by
00:17:19 the fact that Microsoft at the beginning of the event said a bunch about how all
00:17:24 the OEMs, all the chip makers, everybody's in that this co-pilot plus PCs
00:17:28 thing is, is like an industry wide move, and then spent the whole rest of the
00:17:34 hour talking about Qualcomm powered devices. What is what do you make of
00:17:38 that?
00:17:39 Yeah, they kind of did this thing that was like, oh, AMD and Intel are going to
00:17:43 have these co-pilot plus PCs, you know, soon, but not right now. So we've been
00:17:48 partnering with Qualcomm to do this. I think they're kind of, I think they're
00:17:51 downplaying a little bit their ARM transition there, because obviously they
00:17:55 want this to be all partner focused. That's the whole idea of Windows. But
00:17:58 they've made such a big effort, even in all those demos, like there was no
00:18:02 mention of Intel and AMD, you know, it was all about these Qualcomm devices,
00:18:06 all about the transition to ARM, like all about how they've recompiled the
00:18:10 kernel and done all this work in the background, so that they're confident
00:18:14 that ARM is finally going to be, you know, competitive with the Intel based
00:18:18 models that we have of surface devices right now. So I think they, they kind of
00:18:23 had to throw a bone to Intel and AMD and had to have them included, obviously,
00:18:27 in this event. And I'm not, that's not terribly surprising, but their stage
00:18:31 presence was, was very little compared to all the ARM talk that they had on
00:18:34 stage, right? It was all, all of those AI experiences were lit up by a
00:18:39 combination of these new chips. So it's going to be interesting to see when
00:18:42 Intel and AMD can actually have an answer with the MPUs that are as
00:18:46 performant. Because they're already, like they're talking, these chips are
00:18:51 45 tops, whereas the Intel ones are like less than 10 right now. So by the
00:18:56 time Intel matches, will Qualcomm then jump ahead? Like, it's going to be
00:19:00 interesting to see where the roadmap sort of pans out here.
00:19:04 I did enjoy, this wasn't the spec heaviest event ever, which I thought was
00:19:08 really interesting. They spent a lot of time doing like app demos and a lot
00:19:11 less time talking about the details of how these chips work and all that
00:19:17 stuff. But they did keep talking about 45 trillion operations per second over
00:19:22 and over and over. That was like the one spec they said like 10 times. It was
00:19:27 like, they are, they're like, our chip is so fast. And we're, we're going to
00:19:30 remind you 11 times during this event, how fast our chips are, because they're
00:19:35 so proud of their fast chips.
00:19:36 Exactly. And depending on how they calculate it, and we'll have to see, I
00:19:40 think it will be faster for AI operations than the M4 perhaps. So it's
00:19:45 going to be interesting when that roadmap shakes out. Because obviously
00:19:48 Qualcomm have had ARM chips for years and they've never been quite this
00:19:51 performant and they acquired Nuvia. And I think that is the key to why these
00:19:55 chips are different. And I mean, that's, that's the key because those Nuvia
00:19:59 chips were built by ex-Apple engineers. So it's almost like, you know, they're
00:20:05 building the Windows Silicon. So I think, I think that is the key to why
00:20:08 this all feels a little bit different. It's that the chips are actually
00:20:12 performing like you would expect them to.
00:20:13 So does all that add up to this being the moment Microsoft seems to want it to
00:20:20 be like the, the sense I got reading your reporting and from talking to you
00:20:23 about this for the last couple of months is that this is kind of a day
00:20:26 Microsoft has had circled on the calendar for two years. It's like, this is
00:20:31 the beginning of the new era of Windows. We have the power, we have the AI
00:20:35 stuff. Copilot is starting to get there. Like the, there, along with this,
00:20:39 there were, I think every OEM on planet earth announced a bunch of Copilot
00:20:44 plus PCs. Like this, this was clearly designed to be sort of the beginning
00:20:49 of an era of Windows PCs. Do you think it got what it was trying to do?
00:20:53 Yeah, I think so. I think I was surprised by some of the AI features that
00:20:57 they showed on stage. I think those, those look really promising. So I
00:21:00 wasn't expecting as much of the AI features as they showed. And just
00:21:04 coming to campus this morning, it's like, they literally have like signage
00:21:08 everywhere. It was like the new AI of the era has is here sort of thing. So
00:21:13 that they're obviously like very confident in it. And what they showed
00:21:16 and just, just playing around with the devices, it does feel like something
00:21:20 is different here. Like this is definitely different from like the 2018
00:21:24 push and definitely different from the surface RT push. So yeah, I mean, I
00:21:28 want to try them out. I want to review them next month and give them a good
00:21:31 sort of spin just to see if this is actually, you know, this can replace my
00:21:35 own laptop and see if they can really nail that battery life then. Yeah. I
00:21:40 don't know. I don't know what Intel and AMD are going to do because I know
00:21:44 Qualcomm might've delivered it.
00:21:46 Yeah. I feel like you and I are in exactly the same headspace, which is
00:21:48 like, I desperately want this to be everything that they said it would be
00:21:52 because it would be so much fun if the whole windows ecosystem just kind of
00:21:56 leveled up. But also we've been burned on this specific promise several times
00:21:59 in the past.
00:22:00 It's like, if I sound cautious, it's because of those two burn moments in
00:22:04 the past.
00:22:04 Exactly. But it does. I think if it's, if it's ever going to happen, it feels
00:22:08 like it might happen right now.
00:22:09 Yeah. I think if it, if it doesn't happen right now, then I mean, they've
00:22:13 blown it, right? Everything that they've showed and all the, all these big
00:22:16 promises, it'll be pretty crazy if, if they, it doesn't live up to that. You
00:22:21 know, if the performance isn't there, these AI features don't work properly
00:22:24 or like the battery life just sucks, then Microsoft's blown it. Right. So
00:22:28 they kind of have to have done it this time. Otherwise there's a lot on the
00:22:32 line, given the way that they've been talking and the confidence that they're,
00:22:35 that they're putting into this.
00:22:36 Totally. Yeah. I realized going in that there were a bunch of people who bought
00:22:42 iPad pros last week who were looking at the surface pro this week being like,
00:22:45 Oh, maybe I want a surface pro. And it's like, Oh, Microsoft's doing that.
00:22:49 They've done it. They're back.
00:22:50 That's true. Yeah. It's not, it's not as thin and light as sexy hardware as
00:22:55 the iPad pro I'd say. But yeah, I mean,
00:22:58 But it runs all my apps, Tom.
00:22:59 It runs all your apps that you can actually use it. Yeah. So when you're
00:23:02 processing video and you tap a notification, it's not going to kill that
00:23:06 video in the background. Like it's an actual OS. So yeah.
00:23:10 A hundred percent. All right, Tom, you got meetings to get to you. Thank you
00:23:13 for, for jumping on and doing this. I appreciate it.
00:23:15 Yeah. No worries.
00:23:16 Good luck with the rest of your stuff in Redmond.
00:23:18 All right. Thank you. I'll speak to you soon.
00:23:20 All right. We got to take a break, but actually real quick before we do,
00:23:23 you can subscribe now to Tom's new newsletter called notepad about all
00:23:27 things, Microsoft. I'll put a link in the shout outs. It's brand new and it's
00:23:30 already fantastic. All right. Quick break. Then we're going to talk open AI.
00:23:34 We'll be right back.
00:23:34 All right. We're back. We talked a bit on last week's show about the open AI
00:23:44 event from last Monday in which the company announced GPT 4.0 and a super
00:23:49 personable, maybe even flirty voice assistant.
00:23:52 Hey, I'm Mark. How are you?
00:23:55 Oh, Mark. I'm doing great. Thanks for asking. How about you?
00:24:00 So I'm on stage right now.
00:24:01 If you hear that clip and you're like, Oh, weird. That sounds like Scarlett
00:24:05 Johansson in her. Well, so did everybody else. I mean, listen to this from her.
00:24:11 Okay. Let's start with your emails. You have several thousand emails
00:24:14 regarding LA weekly, but it looks like you haven't worked there in many years.
00:24:17 And now this clip from the open AI event.
00:24:19 I see it now. You wrote down three X plus one equals four.
00:24:26 Yep. I could do this forever because the similarities are deeply uncanny,
00:24:32 but here's just one more. This is from the GPT 4.0 demo. When one of the
00:24:36 researchers wrote, I heart chat GPT on a piece of paper and then held up his
00:24:40 phone camera so it could see it.
00:24:42 Of course. I'd love to see what you wrote. Show it to me whenever you're ready.
00:24:46 Okay. So this is what I wrote down. What do you see?
00:24:48 Oh, I see. I love chat GPT. That's so sweet of you.
00:24:55 And now here's a moment from her.
00:24:57 Thank you, Theodore. I'm kissing your hand.
00:25:05 Side note, by the way, her is such a good movie. If you haven't seen it or
00:25:11 just haven't seen it in a while, do it. It's even better than I remembered.
00:25:14 And with all the context of the last few years of AI, I watched it really
00:25:18 differently this time. Anyway, other than the fact that the actual product
00:25:22 doesn't work quite as well and sounds a little robotic. I mean, the vibes are
00:25:26 very similar, both in the product in general and how you're supposed to use
00:25:29 it and the voice specifically to the point where actually over the weekend,
00:25:34 open AI pulled the voice that you're hearing here, which was named sky from
00:25:38 the product altogether. Open AI says it wasn't trying to mimic Scarlett
00:25:41 Johansson in her, which I find hard to believe for reasons we'll get into.
00:25:46 There are lots of questions here about what we've learned from that GPT 4.0
00:25:50 demo, about what open AI is up to in general and about where we're headed in
00:25:54 the next few years of AI. And to talk through all of it, I grabbed Kylie
00:25:57 Robison, the Verge's new senior AI reporter. Kylie Robison, welcome to the
00:26:01 Verge cast. Thank you for having me.
00:26:03 This is what day nine for you at the Verge as we're recording this?
00:26:07 Precisely, yes.
00:26:08 And you've already been in like 465 news cycles.
00:26:12 That is 100% correct. Maybe one more today, but yes, exactly.
00:26:17 How's it been so far? How are your first nine days at the Verge?
00:26:21 It is so fantastic. It's pure chaos. And I think a lot of people looking are
00:26:25 like, "Why are you doing all this? Are they killing you?" No, it's because it's
00:26:28 so fun and I can't stop agreeing to write stuff. So it's been great.
00:26:32 Yeah, you fit in very well in that respect. Mostly I want to talk about
00:26:35 OpenAI. There's a lot going on, but I think to me, what's been interesting
00:26:39 is there's been a ton of news the last two weeks. And the only thing I
00:26:43 can think about is that OpenAI demo. Oh my god.
00:26:46 27 minutes of just super awkward OpenAI interactions. And I can't stop
00:26:52 thinking about it. I don't know. So I'm mostly just curious. You've been
00:26:55 covering OpenAI for a long time. You've been thinking about this company,
00:26:58 you've been thinking about AI for a while. You're our senior AI reporter.
00:27:01 What did you make of that event? There were all these rumors leading up to
00:27:04 this about what OpenAI was going to do. They did it ahead of Google I/O, I have
00:27:09 to assume deliberately. There were lots of rumors. None of those rumors turned
00:27:13 out to be true. This was like this other thing. Where was your head during
00:27:16 that event? What were you thinking about? Well, I was really excited for it to be
00:27:20 this huge, huge thing. They built it up so much. So the rumors were, "It's going
00:27:25 to be a search engine that rivals Perplexity and Google. It's going to be
00:27:29 GPT-5." So when we got GPT-4, oh, it's cool. It is. How natural this is, it's
00:27:35 very sci-fi. But I did my first TikTok for The Verge. And my takeaway was,
00:27:41 "This is so flirty. Do we need flirty AI? I don't think so." Just like, you
00:27:46 know, not to throw OpenAI completely under the bus. I said the same about
00:27:50 XAI's Grok. Like, do we need a sarcastic, snarky AI? I'm not sure. Let's get
00:27:57 some other advancements first before it starts being mean and flirty. Some
00:28:02 people in the TikTok comments did not agree. These Gen Z-ers were like, "Do
00:28:04 not take away my AI girlfriend, please." The announcement, it was interesting.
00:28:10 It's not a huge leap forward. I think they definitely wanted to get in front
00:28:13 of I/O. I did talk to the CTO, Meera Marathi, in a very short briefing after
00:28:18 the launch. And I asked, like, "Hey, why today?" And she's like, "You're
00:28:22 asking because of I/O?" And I'm like, "Well, of course." And she said, "You
00:28:26 know, we didn't even know I/O was happening." I'm like, "Okay."
00:28:28 - Yeah, cool. I mean, that's just a lie. That's, like, not even a good
00:28:31 answer.
00:28:32 - Yeah. I'm like, "Okay, we're not taking the bait. All right, moving on." But
00:28:36 yeah, so it was interesting. Obviously, I went to I/O, and I thought, you know,
00:28:42 this stuff is really cool. It was a two-hour-long keynote. I was there in
00:28:46 person. Pretty brutal. But yeah, GPT-4.0, it's cool. It has a long way to go.
00:28:51 It kind of feels like the rumors in the AI reporting world is that they had
00:28:56 something bigger that wasn't ready in time. So they're like, "Here's 4.0.
00:29:00 Microsoft Build is coming up at the end of this month, so I'm kind of expecting
00:29:05 something bigger for that." But yeah, it was an interesting demo. Flirty.
00:29:08 - Talk to me about the flirty thing, because I confess, this is the thing I
00:29:11 can't stop thinking about, right? And like you pointed out in the middle of
00:29:14 it, like, this is just her. And I think Sam Altman even said, like, referenced
00:29:18 her before Meera Marati was like, "No, we're not doing a her thing," which is
00:29:21 like, again, obviously not true. They're obviously doing a her thing. I'm so
00:29:26 torn on this, and I'm curious how you're thinking about it. On the one hand, I
00:29:29 totally understand why it makes sense for these things to, like, be fun to talk
00:29:33 to, right? And I don't know if flirty is the correct version of fun to talk to,
00:29:38 but again, I think this push towards, like, if you're going to spend all your
00:29:41 time interacting with this thing, it should be enjoyable to interact with and
00:29:45 not just some, like, cold metal robot that you talk to. I totally understand
00:29:49 that. I also think that is, like, a mess. - Yeah. - And just, for OpenAI and for
00:29:54 everybody else, will cause as many problems as it solves, maybe more, and is
00:29:59 going to make all of this really weird before the technology is nearly good
00:30:03 enough to actually have any of this work. Like, her is a fun example, because,
00:30:07 like, she doesn't get basic facts wrong all the time in her. Like, importantly,
00:30:11 it's very good technology. - Yes. - It's not just, like, flirty Scarlett
00:30:16 Johansson, but very dumb. Like, that's not what they're doing. And I don't
00:30:20 know, I just cannot wrap my head around whether this, like, 11-year-old movie
00:30:25 about a voice assistant is the correct place for us to be going, even though I
00:30:31 kind of understand why we're going that way. I don't know, and I feel like
00:30:34 everybody's had very visceral reactions to it, so I'm curious what yours has
00:30:37 been. - You know, a lot of tech reporters have said, like, "Please, for the
00:30:42 love of God, can these tech executives finish the movies? Like, see how it
00:30:46 ends?" - Right. - Yeah, no, not convinced we need flirty AI before we have
00:30:51 accurate AI. And something I noticed during Google I/O is that when they were
00:30:56 demoing their similar assistants, it was on video. It was not live. And
00:31:01 versus OpenAI's demo was live, so you got to see all the flubs, and if the
00:31:06 listeners have seen Nilai break the AI product... - We'll put it in the show
00:31:11 notes. It's very good if you haven't seen it. - It's very good. I was sitting
00:31:14 in the press lounge, and they came back. They're like, "Nilai broke it twice,"
00:31:17 which is what I expected. - That's the goal, yeah. - Yes, exactly. So, yeah,
00:31:22 flirty AI. As a tech reporter, it just makes me, like, "Okay, can we do
00:31:26 better than this?" I don't want to completely drag them, because it is cool
00:31:30 that we're finally getting, like, a working Siri. I was just thinking before
00:31:34 we started this episode, when I use Siri almost exclusively as a timer,
00:31:38 because it doesn't work for anything else, I do laundry, and it's 50 minutes
00:31:43 every time, and it always thinks 15 minutes, no matter how hard I try. That's
00:31:46 how mostly useless it is, so I have to say 55. - That sounds right. Yeah. - Yeah,
00:31:50 so flirty AI, not super useful, feels purposeful, obviously. Mira said,
00:31:56 you know, "Not the case." She said, "Someone else in the audience asked me
00:31:59 that. I don't know where they're getting that from. It just sounds so
00:32:02 natural." And I'm like... And then, meanwhile, Altman is like, "Her." - Yeah,
00:32:07 he just tweeted the word "her," didn't he? - Yes, he did. So, and he said,
00:32:11 "I've been at," I think it was a Dreamforce or something, at a conference,
00:32:14 and he said "Her" was his favorite movie, and all his tech reporters were,
00:32:17 like, noted. So, yeah, I think they're trying to inch towards that reality.
00:32:22 I don't know. I feel a bunch of different ways about it. What were you
00:32:25 thinking when you saw it? - I thought it was really impressive, and I just
00:32:30 keep coming back to, like, the immediate reaction everybody had to ChatGPT,
00:32:35 which is, like, "I can't believe it's this good," which is both a compliment
00:32:40 and an insult, right? Like, it's not very good, but it is genuinely impressive
00:32:44 that it is as good as it is. And I feel like I've spent a lot of time
00:32:48 reporting on AI in the last year and a half, and trying to constantly
00:32:52 calibrate that, that it's impressive, that it's as good as it is, and it's
00:32:55 also not very good. Like, both of those things are true, and it is so
00:33:00 hard to hold both of those ideas in your head at the same time. And this,
00:33:04 to me, was like that, too. Like, I actually went back and watched—it's
00:33:07 like a 26-minute video. Everybody should watch it. It's weird and awkward
00:33:11 and delightful, but it's actually, like, it screws up a lot. Like, a lot in
00:33:16 the demo. And again, that's, like, it's fine. This is new, early technology.
00:33:19 They're having fun. This is not—they didn't promise that it was perfect.
00:33:21 They just sort of showed how it worked. But it screwed up a lot, and nobody
00:33:25 talked about it, right? Everybody came to the end of that thing, and all
00:33:28 over the internet, the immediate reaction was basically, like, "OpenAI
00:33:31 just destroyed all of the AI competitors forever. It is leaps and
00:33:36 bounds ahead of everybody else. This is the only thing that there is
00:33:39 forever and ever." And I just don't think that's true. Like, it's cool,
00:33:44 and it's a good demo, and the fact that it feels at all like a person and
00:33:49 does convincing-sounding, like, ums and laughs, and—
00:33:52 - Yes.
00:33:53 - What was it? He showed the, like, "I, I heart GPT-4.0," and it was like,
00:33:58 "Oh, that's so sweet." Like, that is both impressive and nothing simultaneously.
00:34:03 - Exactly. Exactly. I think a really good example of that, too, it's like an
00:34:08 OpenAI effect, is with Sora. People are like, "This has just put all of these
00:34:12 text-to-video AI companies out of business." And people like to say they
00:34:16 launched it, but neither of us can use it. It's still in demo mode, so, you
00:34:20 know, people are like, "Wow, OpenAI destroyed everyone." It's, you know,
00:34:23 this technology is not even out yet, and clearly it has a long way to go when
00:34:29 you see it breaking in demos. So, yes, it's very hard to hold both of those at
00:34:33 the same time, and day nine at The Verge, I'm still figuring out how to balance
00:34:37 that.
00:34:37 - It's tough. I'm curious—I think the Sora thing is actually an interesting
00:34:40 example, because OpenAI was perceived to have this big lead, right? Like, it
00:34:44 kicked off a lot of this stuff with ChatGPT. It has been the company we have
00:34:47 talked about the most as we talk about AI. But now, like you said, we're in
00:34:51 developer conference season. Google just announced, I would say, in a
00:34:54 significantly more polished and less interesting way a lot of the same stuff
00:34:59 that OpenAI and ChatGPT have been doing for a while. Microsoft today, as you're
00:35:04 hearing this, is presumably going to launch some of the same stuff,
00:35:07 potentially with its own technology. Apple is likely to do some of the same
00:35:11 stuff, potentially with OpenAI. Anthropic is out here doing stuff. There's just a
00:35:14 million competitors now. So I feel like one thing I've been thinking about, and
00:35:17 I'm curious how you're thinking about this as you try to sort of map out this
00:35:22 space in your mind. Where is OpenAI in the AI world right now? Is it still
00:35:27 winning the race, does it feel like?
00:35:29 - In the hype, perhaps.
00:35:31 - Fair, yeah.
00:35:32 - They get a lot of hype. Rightfully so. They do deliver a lot of mind-bending
00:35:36 sci-fi kind of stuff. And that's part of why I joined this beat, because, wow,
00:35:40 I loved reading sci-fi books. This is all very weird and fun.
00:35:43 - Totally.
00:35:44 - I was speaking to someone at a major AI company who said that essentially,
00:35:48 this year you're just going to see them leapfrog over each other as they all go
00:35:53 towards the same thing. They're all sort of training models in the same way,
00:35:58 they're all trying to deliver the same products. So it's hard to say, like,
00:36:02 tomorrow things can change. This beat, this technology moves so fast. I felt
00:36:08 this way when Sora came out, because I had just interviewed the CEO of Runway
00:36:12 for a piece.
00:36:13 - What's Runway?
00:36:14 - Runway is a text-to-video startup, AI-generated startup. And they helped
00:36:21 create everything everywhere all at once. So some of the AI tech was used in that.
00:36:25 So they're a very cool company. And I thought when I saw their demo, this is
00:36:29 sci-fi and cool. But then Sora came out and I was like, "Oh my God, this is crazy."
00:36:34 And I sent it to the Runway CEO and he was like, "Welcome to our world." Like,
00:36:37 things change so quickly every day. So how is AI in the race? Like, where are
00:36:44 they in the race? They're still leaps and bounds ahead of maybe someone like
00:36:49 Amazon. But things can change at any point. And I think we're going to see
00:36:53 that a lot this year.
00:36:54 - Yeah, I do. I've come to think that that is why it's so chaotic, is because
00:36:59 there is this sense that this is like a land-grab moment, and it's all happening
00:37:03 so fast that you actually still kind of can catch up. Like, I remember a year
00:37:08 ago hearing stories from people that were like, "Oh, only a few companies are
00:37:12 going to be able to do this. Training the models is so expensive and so hard,
00:37:14 it's not going to work." And I think there's some truth to that, but also
00:37:17 there is so much open space in this market right now, and there's so much
00:37:21 money in it that it feels like we're probably at least a couple of years away
00:37:25 from any of this feeling at all settled. And so if you're a company that is
00:37:29 like, "We want to find a place to make a lot of money very quickly, and we kind
00:37:32 of think we can win because no one really has yet," this is probably the
00:37:36 place to do it, for better and for worse.
00:37:38 - Definitely. I agree.
00:37:39 - So you mentioned all of these companies sort of working towards the same
00:37:43 thing, like that the North Star is coming into view. What do you feel like
00:37:46 that thing is that everybody is pointed towards?
00:37:48 - You know, as an avid VergeCast fan and listener, I've been hearing you talk
00:37:53 about hardware in your ears, and I think it was Meta, a scoop came out from the
00:37:58 information about Meta doing headphones for AI. Yes, I thought... I was at a
00:38:04 South by party, and we were going around asking questions about technology, and
00:38:08 something I said that's way ahead of its time is hardware, like this AI
00:38:11 hardware, which turned out to be true. That was a couple months before we saw
00:38:15 the R1 and the HumanePin. Something that was really shocking to me as I joined
00:38:18 this beat is how much these AI companies are investing in robotics and bringing
00:38:23 this AI into a physical plane, because I thought that that's way off. We're not
00:38:28 going to be there anytime soon, but they are seemingly investing quite hard
00:38:32 into... Like, you have the Ray-Ban Meta, the Meta Ray-Bans. It looks like Sam
00:38:36 Altman is doing something with LoveFrom, which is Johnny Ives' startup. So,
00:38:42 where is this headed? I think we're going to see more of those conversational
00:38:45 AI assistants. We have what we saw at Google I/O. I would kind of love this
00:38:51 world where AI is doing everything for me, because I have too many emails and
00:38:54 too many notifications. But I'm also pretty torn on how they're using my data,
00:38:59 obviously. But yeah, so I think we're going to see more advances in hardware.
00:39:05 I'm so surprised to hear you say you think hardware is the thing. It is music
00:39:09 to my gadget-loving ears. But I've been wondering at this moment, after Humane
00:39:15 failed so badly and Rabbit failed so badly, and those companies aren't dead,
00:39:19 and this stuff moves fast, and my Rabbit R1 updates every other day at this
00:39:23 point, and it's ridiculous. But that one particular version of that kind of
00:39:28 thing is just clearly not going to work, right? And I think it was very telling
00:39:32 to me that that OpenAI demo, they're just sitting there talking into a phone.
00:39:36 - Yes.
00:39:36 - And I saw a bunch of people pointing that out, that this is actually the way
00:39:41 that we do this, is we sit here and I hold my phone in my hand and I talk to
00:39:44 it. And that is the interface for AI that we are doing right now. And I have a
00:39:48 lot of theories about why that's not the case forever, but I did kind of wonder,
00:39:52 like, are we going to put off this dream for five years in order to let some of
00:39:57 that stuff get better, and then we'll worry about the hardware, because phones
00:40:00 are great and they'll get us where we need to go. But I have to say, I find it's
00:40:03 very exciting to me that these companies are still buying into the idea that
00:40:07 hardware is part of the equation, if maybe not the whole thing, at least not
00:40:12 for now.
00:40:12 - See, this is where you know a lot about this, but I was surprised,
00:40:17 obviously, that they're doing anything in hardware that they still think that
00:40:20 this is the case, because the HumanePen and the R1, as nifty as they seem to be,
00:40:25 I kind of think they're just destined to be in drawers collecting dust and not
00:40:30 really the next big thing. But I'm wondering if you have any tinfoil hat
00:40:34 theories about it being an iPhone on stage, because I thought, you know, there
00:40:37 is, for the listener, there's reports that OpenAI and Apple are going to do
00:40:41 this deal where chat GPT is integrated into the iPhone, which everyone's like,
00:40:46 "Oh my God, once again, Siri might work. That could be great." I do see the
00:40:50 future just... I mean, who doesn't want to just carry their iPhone instead of
00:40:54 multiple devices, like you've mentioned? When I say hardware, just as you've said
00:40:58 on the Verge cast, headphones, glasses, I think you said you're not into the
00:41:02 glasses idea. I wear glasses, so it seems perfect to me.
00:41:05 - I think if you wear glasses, it is like a perfect, obvious place to go. For me,
00:41:10 like, I have the Meta Ray-Ban glasses, and my problem is, I don't know how long
00:41:15 it takes. This is just going to be David complains about wearing glasses for a
00:41:17 minute. But my problem is, how long does it take to wear glasses before you stop
00:41:22 noticing that you're wearing glasses? Because I have these, they have clear
00:41:25 lenses, they're transitions, they're actually really useful. I like having them
00:41:28 on all day, but I never stop noticing that I'm wearing glasses, and it drives
00:41:31 me kind of nuts. And part of that is because they're heavy, but part of that,
00:41:34 I think, is also just like, it's going to be hard to convince people who don't
00:41:37 wear glasses to wear glasses. But maybe I just need to wear them longer. Like, do
00:41:41 I just need to put these on for three days, and I'll get used to it?
00:41:44 - Well, it's taken me about 20 years of wearing glasses, to be fair. So I don't
00:41:50 even notice them on my face. But, well, the answer to that is, if Meta makes this
00:41:54 earbud push, I don't know if you read the article, but they're still working out
00:41:57 some of the kinks, reportedly.
00:41:58 - Well, because it was earbuds with cameras in them, right? Which is just the
00:42:02 wildest idea.
00:42:02 - But you feel like, they're like, found out that if you have long hair, the
00:42:06 cameras are going to be covered, because obviously. And one of my favorite parts
00:42:10 is that they're figuring out how to stop it from being too hot, which is, I think,
00:42:14 something that was experienced with the HumanePen, is that people were like,
00:42:17 "This is getting really hot." So, you know, catching fire in your ears doesn't
00:42:21 sound like a great product. I hope they work that out, especially for you, since
00:42:25 you can't wear glasses.
00:42:25 - Yeah. I'm bullish on the idea of headphones being the next thing. But again,
00:42:29 it's partly just as a conduit to your phone, which I think works fine.
00:42:32 - Exactly.
00:42:32 - But I want to go back to your point about the iPhone, because one of the
00:42:35 things that OpenAI, I think this was Sam Altman who said it, that there's a new
00:42:40 chat GPT app for the Mac. And there were a lot of people who I would say
00:42:44 reasonably asked, "Where's the Windows app? Microsoft owns an enormous
00:42:48 percentage of your company. You have been partners with them for forever. Like,
00:42:52 what the hell?" And I think it was Sam, right, who just said sort of offhandedly,
00:42:56 like, "Oh, Mac is just where our users are." And then there's an iPhone on
00:42:59 stage. And it's like, how many conspiracy theories should I read into those two
00:43:04 facts? You tell me.
00:43:04 - You know, okay, my tinfoil hat is that there's beef. I've been told that it's
00:43:09 not necessarily true, not by anyone, like just from coworkers. From Tom. Tom was
00:43:14 like, "I have opinions about this." But I just think that there's beef. I was
00:43:20 covering OpenAI during the ouster, the meltdown. And I'm just imagining Satya
00:43:25 during his Thanksgiving week, like, "Damn it. Why do I have to deal with this?
00:43:29 He's on Pivot. He's on, like, CNBC." So I think what we're seeing with inflection,
00:43:36 them bringing in inflection, it seems like there's sort of this potential Apple
00:43:40 deal. It feels like there's this decoupling because both are realizing that
00:43:44 they've relied so much on the other and that it's like a codependent
00:43:49 relationship. They're like, "Maybe we should take some time apart," is where I
00:43:52 see this going. Especially with the iPhone on stage, the Mac app, it feels
00:43:56 like there's a lot of signs, but that's just my tinfoil hat for now.
00:44:00 I do think it is generally true that companies do that kind of stuff on
00:44:04 purpose. I think it's a thing that I have learned over the years is that you
00:44:07 want to just think, "Oh, they just brought up their phone and did it." And
00:44:10 it's like, no, even the stuff that looks like it's not thought through and
00:44:13 rehearsed is usually pretty thought through and rehearsed and is not an
00:44:17 accident.
00:44:18 Exactly.
00:44:19 There was a conversation about what phone they would use for that demo. I'm
00:44:22 very confident about that fact. To your point, part of the reason I find this
00:44:26 space so fascinating right now is it feels like that uncoupling you're
00:44:29 talking about is literally happening right now in real time at insane speeds.
00:44:34 Yes.
00:44:35 Between Google I/O and WWDC, which is essentially what? Four weeks, we are
00:44:40 going to potentially completely upend the economics of the AI industry. It's
00:44:46 insane. You think about what the Google search deal has done for Apple and all
00:44:51 the regulatory stuff that that has caused and the amount of money that
00:44:53 changes hands. If you put something like that into the iPhone through OpenAI,
00:44:58 it blows up that whole industry in a crazy way. Or if they just randomly pick
00:45:01 Anthropic, which to me is the most chaotic answer, and I would love for that
00:45:05 to happen.
00:45:05 Delicious, yes.
00:45:06 Just go open source. Be like, "We went with meta. We did llama. Apple Siri
00:45:11 powered by llama and just blew up the world." It'd be amazing. But it just
00:45:14 feels like all of this was starting not to settle but to be understandable. You
00:45:20 kind of knew where everybody was and what they were working on. And now it
00:45:23 feels like between all these relationships, and I think especially with
00:45:26 Apple, all of the balls are up in the air again in a really interesting way.
00:45:31 Yes. And something I've been thinking about a lot is sometimes these AI
00:45:35 researchers are like, "We're building GPT-5 and it's hard to predict what it's
00:45:39 going to do." It's hard to predict what any of this is going to do. And that's
00:45:42 strange as a reporter and just a human in this world. We don't know how this is
00:45:47 going to change our lives at any point.
00:45:49 Do you think that's part of why they're leaning in on the flirty personality
00:45:54 side of it? Because if it's delightful, it doesn't matter?
00:45:58 Yes.
00:45:58 That having fun with it becomes the point and not the "Is it any good at
00:46:02 tasks?" which is what they want to get to eventually.
00:46:05 I've heard this point before. And starting as an AI reporter, people told
00:46:09 me not to anthropomorphize the technology because then it takes the blame off
00:46:16 the company and they can blame the AI. So I'm so cognizant of that as I
00:46:20 report. And then they come out with flirty AI and it's like, I'm confused.
00:46:26 Because I think OpenAI also said at some points, "It's tech. It's not
00:46:31 human." And we saw that with the Google guy who was like, "This is real and
00:46:35 it's human." And so to see all of that and then go, "Here's your flirty AI
00:46:43 girlfriend," like her, it feels strange. But yes, I do think it's an
00:46:48 interesting critique to have. Like, "Look over here. This is fun. Don't
00:46:52 pay attention to how it barely works," I think is part of the point.
00:46:56 Yeah, I agree. I will say, by the way, that the New York Times has
00:47:00 cornered the horny AI beat. So we're going to have to find something else
00:47:04 for you to do. I'm really sorry. You were too late to the game for the
00:47:07 horny AI beat. It's just, you can't have it.
00:47:09 I was wondering if I could say this on The Verge cast. This is a huge day
00:47:13 for Kevin Roos. I forgot to message him about it. He's going to have 10
00:47:16 more girlfriends by the time we finish this.
00:47:18 Kevin Roos, vindicated every day. It just gets better and better to be
00:47:21 Kevin Roos. Kevin, we love you. Come on The Verge cast.
00:47:24 One more thing before I let you go. You mentioned the brief Sam
00:47:28 ouster. And the one other bit of OpenAI news last week was that Ilya
00:47:33 Sutskever, who was a big part of that story, is leaving OpenAI for good.
00:47:39 What do you make of that story? What do you know about what's going on
00:47:41 there?
00:47:41 Oh my God. That was about 4pm while I was at Google I/O. And I was like,
00:47:45 "Oh my God, please, no more news, please." Yeah, that seemed to have
00:47:51 caused other people to leave OpenAI as well. The sort of Ilya cult inside
00:47:55 of OpenAI. I don't know. I think I'm really hoping we get some fun
00:48:00 discovery, maybe from the XAI lawsuit against them, to figure out what
00:48:05 went down. Not to be totally off course, but when the government is
00:48:09 deciding to ban TikTok and everyone's like, "Where's the receipts? What
00:48:12 happened?" Similarly here, what is happening at OpenAI that caused all
00:48:17 this drama and now Ilya is leaving? And it all feels very strange. My
00:48:22 thoughts are we need a lot more information. I think that is sort of
00:48:27 owed as this changes people's lives and how they interact with technology.
00:48:32 The meme is, "What does Ilya know? Where's Ilya?" I think we deserve some
00:48:37 answers there. I don't think we're going to get them from OpenAI. We're
00:48:39 going to get them from great AI reporters who dig into it. But it's
00:48:44 strange. It's really strange.
00:48:45 I would think the most simple way to read it is that divide that everyone
00:48:52 sort of ascribed the issues to, but then kind of waved off was the doing
00:48:57 AI for good versus doing AI for profit. The fact that not only did Ilya
00:49:02 leave, but then a bunch of people, like you said, in the Ilya cult also
00:49:06 left, suggests to me that that divide or whatever the divide was still
00:49:11 exists. And there was some reporting that said Ilya basically never came
00:49:14 back to work after all of this. And so it definitely suggests to me that
00:49:18 whatever happened last fall is not done happening. Like Sam won in whatever
00:49:23 that looks like. But whatever is going on inside of that company seems to
00:49:28 still be going on inside of that company.
00:49:30 I agree. I think that rift is still there. The reporting suggests that it
00:49:34 was like what EAC versus the D-cells, people who want this to move fast and
00:49:38 people who want this to be a research lab that's for good. I think that rift
00:49:43 is still there. And I think that's something you're going to continue to
00:49:46 see in this AI race. I think the people who want to make money are going to
00:49:50 continue coming out on top as we see here, though.
00:49:53 Yeah. I mean, as of right now, I think just thinking about Google I/O, it's
00:49:57 like, why did they make any of these choices? And the answer is like, oh,
00:49:59 because there's so much money in it right now, right this minute. There's so
00:50:03 much money in it. And will there be forever? Who knows? But there is a lot
00:50:07 right now and it is making everybody do truly wild stuff.
00:50:10 Yes, exactly that.
00:50:12 Fair enough. All right. Well, we're going to have to have you come back on
00:50:14 when we get more stuff from Apple and build and everything. So welcome to
00:50:19 hell. And by hell, I mean the Vergecast. Kylie, thank you for being here.
00:50:23 Thank you.
00:50:23 All right. We got to take one more break and then we're going to talk
00:50:26 iPads on the Vergecast hotline.
00:50:28 [Music]
00:50:37 All right, we're back. Let's get to the hotline. As always, the number is
00:50:40 866-VERGE-11. The email is vergecast@theverge.com. Send us all of your
00:50:45 questions. We've gotten a lot of iPad questions, and I think at some point
00:50:48 here, we're going to stop answering iPad questions. But we try to do at
00:50:51 least one on the show every week. And at some point, we'll stop talking about
00:50:54 iPads. But this week, we have a question about iPads, and it comes from
00:50:58 Clifton.
00:50:58 Hi, this is Clifton from Tacoma, Washington. I was just listening to the
00:51:04 case for the iPad Pro podcast. And in your comparison to the iPad and the
00:51:09 MacBook, I was wondering, do you think that the majority of that want for
00:51:13 macOS on the iPad would go away if Apple just made a touchscreen MacBook?
00:51:18 Also, on the same token, do you think that if they made a two-in-one, that
00:51:22 the work case for Pros on the iPad would go away, and it would probably
00:51:26 just go back as a media consumption device? I think it's similar to Apple
00:51:31 thinking they know what customers want instead of actually doing what
00:51:35 customers want. Thanks. I appreciate it.
00:51:37 So there's like a big thinky thing at the end there that I kind of want to
00:51:41 ignore for now. Neelai Patel is here. Hi, Neelai.
00:51:44 Hey.
00:51:44 We've talked around this for two weeks and also like 10 years. So let's just
00:51:49 talk about this. What do we want from Apple is I feel like basically the
00:51:53 question being asked here. There's the iPad, there's the MacBook, and
00:51:56 somewhere in the middle, there is this idea of like a perfect device that
00:51:59 would solve all of our problems. What is it?
00:52:02 I think it's an iPad that doesn't have a totally restricted application
00:52:06 model. That's the problem. The problem is not that it doesn't run macOS.
00:52:09 What macOS represents to people is freedom. I can just use this thing like
00:52:14 a computer and developers can show up and fill the gaps that Apple has left
00:52:18 open. You cannot do that on an iPad. So it's just whatever Apple wants you to
00:52:24 do. And if they would just let go, I think these things would evolve on
00:52:29 very different trajectories and they just won't let go, which I guess comes
00:52:33 back to what do we want Apple to do?
00:52:36 Well, I think I largely agree. Like I got a bunch of crap on threads because
00:52:40 I posted something to the effect of like, I agree that I don't want huge
00:52:44 sweeping changes to the iPad, but it would be nice to be able to put
00:52:47 Windows where I want them to. And a bunch of people are like, you can put
00:52:49 Windows wherever you want. And I'm looking at right now, I can see nine
00:52:52 different apps on my Mac right now. I can see them all. And that is not how
00:52:56 it works on the iPad. And there are like a million little things like that to
00:52:59 me. But the evolution thing, I'm curious how you think about it. Because I
00:53:03 think the case a lot of people would make is that if you just let both of
00:53:06 these things kind of be everything, that eventually they would just crash into
00:53:10 each other and the iPad and the Mac would essentially be the same thing.
00:53:13 But you don't think that's the case?
00:53:14 I don't. I think about a lot of iPad use cases that I see that I think are
00:53:18 really interesting. I'm married to a lawyer. Some lawyers show up at
00:53:22 meetings with iPads just to sign contracts and get through them. You don't
00:53:26 need a Mac laptop for that. You don't even need the keyboard case for that.
00:53:30 Right. And also the fact, by the way, that it lies flat on a desk or table
00:53:34 is actually really important. Like I was thinking about this just form
00:53:36 factor wise, right? Like if you just put a touch screen on the Mac, does it
00:53:39 suddenly solve all of your iPad problems? And like, no, because drawing
00:53:42 like that sucks. Signing things like that sucks. Like having a thing that
00:53:46 just lies down like a piece of paper on a table does change things. I think
00:53:50 you're right.
00:53:50 I will note that you have to put a case on some of these iPads to get them
00:53:54 to actually lie flat. But you know, I agree with you. There's just a bunch of
00:53:57 places where you want a computer in a tablet form factor that is where the
00:54:01 interfaces natively touch. Great. Again, we get notes from people who are
00:54:05 like interior designers and architects who are like, this thing is my
00:54:08 computer. I walk around taking pictures with it and showing people things.
00:54:12 And drawing on it all day long. And I killed a battery in my head every day,
00:54:16 which is admittedly not so hard to do, but there's just a world of use cases
00:54:21 that way. And then there's a world of use cases for the laptop form factor.
00:54:24 And I think the problem is Apple thinks the form factor is important and not
00:54:30 the use cases. So they're like, we'll just make it a laptop. And now it's a
00:54:34 laptop replacement. And I think most people are like, well, I use my laptop
00:54:37 for all these things and it can't do those things. Right. And then I, I,
00:54:41 this is just me imposing a belief on Apple. I think Apple wants computers to
00:54:47 be easier than the market wants them to be. You know, like people, you give a
00:54:51 kid a laptop, like crazy stuff starts happening. Yeah. Right. Like they, they
00:54:56 start to expand on the boundary of what you can even do with some of these
00:54:59 applications. That's really cool. I've seen entire graphic design businesses
00:55:05 that have existed by teenagers using the Instagram create mode in ways that no
00:55:10 Instagram product manager ever thought that thing would be used. That's the
00:55:14 beauty of computers. And to whatever extent that has not happened on the
00:55:18 iPad, I think that's the tragedy. And I think it's because of the restrictions
00:55:22 in the operating system. Yeah, I, I tend to agree, but I think something you
00:55:25 said at the, at the very beginning of this, I think is, is very striking,
00:55:28 which is the assumption that the best version of this thing is closer to an
00:55:32 iPad that does more stuff than it is to a Mac with a touchscreen. Right.
00:55:37 Cause I think like, if you just to boil this question all the way down, it's
00:55:39 like, does a Mac with a touchscreen solve all of our problems? And I think
00:55:43 the answer is clearly no. And I'm not even sure it solves many of our
00:55:47 problems. Like I actually, I look at my Mac and I'm like, I don't know how
00:55:50 often I want a touchscreen on this thing. And I think to me, when I look at
00:55:53 the iPad, the best version of that is so much more interesting than kind of
00:55:58 the next version of a Mac. If that makes sense. Well, I'm confident the
00:56:03 criticism we will get, because you said that is for a bunch of windows people
00:56:06 who use touchscreens on their windows PCs all day long. And they're like,
00:56:09 this is great. Sure. Actually being able to, I think Tom Warren has
00:56:12 published a version of this piece. We've had this conversation for so long.
00:56:15 I can't tell you if it was a decade ago or yesterday, Tom Warren published,
00:56:19 like touching the screen on a PC is pretty good. Cause we've been talking
00:56:23 about it for a long time and it is pretty good. The people who use computers
00:56:26 that way, they get used to it. They like it. To me, it's, it's really just
00:56:30 about the application model. And in particular on the Mac, because the web
00:56:34 browsers are so good and you can run different web browsing engines and
00:56:39 Chrome and Safari and whatever else, there's a richness to the experience
00:56:42 that you just get. You can just have a lot of the latest things on a Mac
00:56:46 because all of the latest things tend to run in the web browser. All of the
00:56:50 AI in the world is happening in web browsers, right? You cannot do that on
00:56:54 iPad. Like the web browser is still pretty much mobile Safari. They fake it
00:56:57 in different ways. And it's just that it's that right. It's Apple has just
00:57:01 restricted a bunch of stuff, maybe to protect Apple's business model, maybe
00:57:05 because they have an idealistic vision of what an iPad user should do or how
00:57:09 they should work or how easy it should be or to keep you safe or secure,
00:57:13 whatever you want to believe about why this has happened, it has happened. And
00:57:16 I think those are the limits that people keep bopping up against and saying,
00:57:20 Oh, just let this be a Mac. Cause the Mac represents those limits going away.
00:57:24 And I really do think you can see it, you know, on other kinds of tablets
00:57:29 that if you let that form factor evolve, it will head towards a different place.
00:57:34 Like, I think it'd be really interesting if the solution to the input problem,
00:57:38 especially with all the AI we have now, isn't just put a keyboard on it.
00:57:41 Like we, we, we just haven't like tried because we are like, maybe the, maybe
00:57:47 everything will just, you know how like everything evolves into be a crab. It's
00:57:51 like everything is evolving to be a laptop again. And like, I don't know,
00:57:53 maybe there's some other ways to do it. Well, yeah. And that's the thing. I
00:57:55 think like when people have asked me over the last few weeks, like, why are
00:57:59 you so hung up on the idea that Apple should or could or ought to make this
00:58:03 thing more like a laptop? I just find myself wanting to hold the thing up
00:58:06 with the magic keyboard and be like, because Apple made it a laptop. Like
00:58:10 the two accessories this company has invested in are a pencil, which makes
00:58:15 perfect sense and is extremely iPad-y. And I actually think it's like iPad
00:58:18 plus pencil is by far the most compelling case for an iPad in every way you can
00:58:24 imagine. Or it's a laptop. Yeah. Like they just, it just looks like a laptop.
00:58:28 It acts like a laptop, but the Gmail app is worse. Like that's just, that's the
00:58:32 difference. And I'm sort of with you in that sense, but like to the Windows
00:58:37 point, right? The Surface rhymes with an iPad. I would say almost more than it
00:58:42 rhymes with a Mac, right? Like it runs Windows and it can do all the stuff, but
00:58:45 in terms of like the spirit of that device, it's so much an iPad. Yeah. And I
00:58:50 think actually like Microsoft got that really right and just has not really
00:58:53 solved a lot of the software stuff around Windows. Like the Surface isn't
00:58:57 everything it could be because of Windows in the same way that iPad is like
00:59:01 hamstrung by iPadOS, but in completely radically separate ways. Well, it's two
00:59:05 companies reacting to the web in totally different ways. Microsoft just lost to
00:59:09 the web. No one's developing hot shit, Win32 applications for the Surface.
00:59:14 Outside of games. Right. I just have to like set those aside, but like the
00:59:18 hottest new applications in the world happen on the web. Like that is a thing
00:59:23 that's just true. The web has become the dominant place to deploy desktop
00:59:27 applications. Great. Microsoft reacted to it by just letting go by being like,
00:59:31 you know what? We built Azure. Put your web applications on our cloud service
00:59:35 and run open AI on our cloud service and we'll make the money there and whatever.
00:59:39 We'll just try to get you to use Edge. Yeah. And that is one reaction. A big
00:59:43 company, especially because they lost to mobile. They basically had no choice.
00:59:46 Apple's response to it is like, no, no, that won't happen at all, sir. Right. And
00:59:50 on our tablet, you will deploy applications for the iPad. You will use the app store.
00:59:56 And indeed, there are some cool iPad apps because of it. That's great. Like Procreate
01:00:01 is an entire company that exists to develop iPad apps and use the iPad in the way that
01:00:06 Apple wants people to use. That is great. That's a success story. But there's not a
01:00:09 lot of great iPad apps. Like they squeezed the balloon and it didn't get a lot bigger.
01:00:13 Like mostly the air went out. There's some balance in there, but you can see those are
01:00:18 two reactions to the same problem. All I keep thinking as you say that is like,
01:00:22 boy, it's ridiculous that Google never figured out how to make decent hardware out of this.
01:00:27 That like this circling Google did around Chromebooks and Android tablets and all this
01:00:32 stuff. Like nobody should be more incentivized to figure out how to do this well than Google.
01:00:36 And it just didn't. And that makes me angry. And I just look at my pixel book and I'm like,
01:00:39 God, you could have been something. It could have been great. I mean,
01:00:42 I famously bought my mother a Chromebook pixel, the thousand dollar Chromebook
01:00:46 several years ago. She still uses it because it was easier to use than a Windows PC or a Mac.
01:00:51 But you open a Mac today, it has 50 different interface paradigms all happening at you at once.
01:00:56 There's a launchpad and Mac apps and a web browser, and it's trying to get you to use Safari
01:01:00 and all this other stuff. And you can run iPad apps and it has an app store, but you can also,
01:01:04 and that's like, that's too much. Like actually just go to a bunch of web pages and use whatever
01:01:10 you need to use there. And here's a really nice piece of hardware to do it. It's pretty good.
01:01:14 And I think you look at kind of all the like the Steam decks of the world that are running games
01:01:18 and Linux now, and you're like, oh, you could build a kick ass Chromebook right now. And is
01:01:23 anyone going to do it? It's like, I don't know. We've come a long way from the iPad where I'm
01:01:27 like, you could build a kick ass Chromebook. But that feels like one version of computing where the
01:01:32 restraints are lifted and it does everything everyone wants. And the simplicity that Apple's
01:01:37 chasing with the iPad is still there. - Yeah. I really do agree with,
01:01:41 and you said this on last Friday show that a desktop web browser, like a true great desktop
01:01:46 web browsing environment would immediately solve a lot of people's iPad problems. I think that's
01:01:51 really real. - Yeah. Let us just run Chrome. Like
01:01:54 honestly, maybe it'll destroy the battery, it'll light on fire. Chrome has a lot of problems. But
01:01:59 if you just let me use an iPad, like a Chromebook, eh, you know, like I would still buy the hardware.
01:02:04 I would still pay for Apple's various services. I would still need iCloud storage. Like you'll
01:02:08 make your money, Tim, but like, let me use the computer the way I want to.
01:02:12 - So maybe part of the reason people want a touchscreen MacBook is that that seems like
01:02:18 the more likely way to get closer to that perfect middle. Cause I think the iPad thing you're
01:02:22 describing is like, what if we just blew up the whole way Apple thinks about itself? And that'd
01:02:26 be cool, but seems less likely than there might be a touchscreen on a MacBook. And Mark Gurman
01:02:33 has reported that Apple's considering that again. Like it has seemed anathema to Apple for a long
01:02:38 time, but I would track that as probably a more likely sort of nudge towards the perfect all
01:02:44 things for all people computer than let's upend the app model entirely.
01:02:47 - Well, I think there's two ways of thinking about it. One, at some point, you're just like,
01:02:51 it's an M4 and it's a keyboard and a screen. They should be the same.
01:02:56 - I mean, I would assume by the way that they will be, like, I suspect we are going to get a Mac
01:02:59 with basically these exact same specs before very long.
01:03:02 - Yeah. The only difference between a Mac and iPad is that Mac doesn't have a random
01:03:05 LiDAR sensor and a camera on the back. But I think the other thing that is happening is,
01:03:11 in Europe, there's a bunch of regulations that say you have to let other web browsers on these
01:03:16 devices with their own engines. And maybe someone just makes a great desktop web browser with the
01:03:22 desktop Chrome engine, maybe. And I think as we've already seen with emulators, it really
01:03:27 recontextualizes what you can do when you get out of Apple's restrictions or those restrictions fade
01:03:32 away and other people are allowed to use the hardware to their advantage. So I think that's,
01:03:36 the iPad is in all the regulations now. There's some back and forth on it, but the Europeans are
01:03:42 Europeans and they're like, no, we're taking that too. Great. So I just think there's something
01:03:45 else that might happen there that could really change how we feel about iPads.
01:03:48 - That's fair. I guess I just mean, I don't see a world in which Apple does that of its own
01:03:53 volition. Like, especially with the iPad, Apple seems so precious about the way that that thing
01:03:58 works that every teeny tiny thing it has done to open it up has felt extremely kicking and
01:04:04 screaming against its will. - Yeah, because it's the one that they
01:04:07 control and it's the one that isn't a monopoly, I guess. Like it's hard to make the case. I guess
01:04:12 the Europeans are regulating it for various reasons because they think it's a gatekeeper,
01:04:15 but it's Apple's. Apple has long said, this is our vision for the future of computing.
01:04:20 It has not won in the market. It just hasn't. Like they sell a lot of iPads. They sell more
01:04:25 tablets than anybody else, but people still use their laptops. Like I can look at our own usage
01:04:30 in Google analytics on our webpage and tell you no one's browsing our website on a tablet.
01:04:35 Like it's just not happening. And so I think they feel protective of it because it's not
01:04:40 under pressure. Like they are the market, but the market's really small.
01:04:43 - Yeah. And I think they have this sense of what it could be and it is like beautiful and perfect
01:04:49 and extremely Apple. - Yeah. If only everyone would do
01:04:51 what they want. - Then we wouldn't need
01:04:54 touchscreen Mac books. It would all be fine. All right. We should go. Neil, I thank you as always.
01:04:58 - See ya. - All right. That is it for the
01:05:02 Verge cast today. Thanks to everybody who came on the show and thank you especially as always
01:05:06 for listening. There's lots more on everything we talked about at theverge.com. Tons from the
01:05:11 Surface event and all the new Microsoft things, all of our open AI coverage, everything about iPads,
01:05:16 all of it. We'll put it in the show notes, but also read theverge.com. It is just a crazy newsy
01:05:21 week. Microsoft build is today. Developer conferences are everywhere. AI stuff is
01:05:25 everywhere. There's just a ton going on. As always, if you have thoughts, questions,
01:05:29 feelings, or PC recommendations for me, you can always email us at vergecast@theverge.com
01:05:34 or keep calling the hotline 866-VERGE11. Like I said, I love hearing from you. We have a Slack
01:05:40 room that just pumps in every voicemail and we get to listen to it. It's so much fun. It is my
01:05:44 absolute favorite thing, both about maybe my job and definitely about being in Slack. If you've got
01:05:49 ideas, thoughts, questions, keep it all coming. This show is produced by Andrew Marino, Liam
01:05:53 James, and Will Poore. The Verge Cast is a Verge production and part of the Vox Media Podcast
01:05:57 Network. Neal, I, Alex, and I will be back on Friday to talk more about build, more about AI,
01:06:02 and everything else going on in tech this week. We'll see you then. Rock and roll.

Recommended