Apple Fails to Overreact to the AI Revolution — With M.G. Siegler

0:00With Apple's AI plan now out in the open, we'll assess whether the company met the moment and what its announcements say about the future of AI for, as Apple put it, the rest of us. All that and more is coming up right after this.

0:16You're used to hearing my voice on the world bringing you interviews from around the globe. And you hear me reporting environment and climate news. I'm Carolyn Beeler. And I'm Marco Werman. We're now with you hosting The World Together. More global journalism with a fresh new sound. Listen to the world on your local public radio station and wherever you find your podcasts.

0:43Welcome to Big Technology Podcast, a show for cool-headed and nuanced conversation of the tech world and beyond where you're here. Recording in the aftermath of WWDC as Apple has made its big AI announcement. And we have a great show for you today because MG Siegler is back. He's a venture partner at GV, formerly Google Ventures, and the author of Spyglass, a great site covering everything in tech. Also, some terrific articles. In the run-up to WWDC and in the aftermath of WWDC and MG, we previewed Apple's big AI play in April. And here we are recording right after the announcement. So we're going to talk not only about Apple, but also about sort of what the set of announcements means for the trajectory of AI more broadly.

1:27So it's going to be a great in-depth show. Welcome back. Thanks for having me back, Alex. Yeah, I feel like when we talked last, I think we, I was going over that and looking, looking over that episode. And I think we hit a lot of sort of it pretty well on the head, but in the intervening time, I feel like I've gone down these wayward paths of like trying to figure out how Apple could integrate open AI and the backend and do all this crazy stuff and do a brain transplant for Siri. And instead, it ended up being pretty straightforward what they ended up doing, I believe. We're going to talk about this, the weird like buildup.

2:02And it wasn't, I mean, I think everybody thought, I thought, and the reports certainly seemed to indicate that Apple and OpenAI would have a major partnership. And it turned out to be pretty minor. So we'll talk about that in a moment. But first, I just want to, you know, I slept on it. I want to get your reaction to this because the aftermath of Apple's WWDC has sort of broken out into two camps, right? Very different camps. There's one group that says this event was Apple's biggest moment in a decade. And Embracing AI is going to set up a super cycle of upgrades for the iPhone. And it was a master stroke.

2:42And then there's the other side where, and this is sort of like reflected in this poll that I did. actually as the event was basically wrapping up, asking people what they thought, whether Apple's AI announcements were wow, cool, or meh. And yesterday, right after the event, the poll had like 49 % of people saying meh. Now 50.9 % say meh. And I listen, listeners, I know this is unscientific, but the reaction, it does capture what a lot of people saw. And maybe it was because German leaked all this, but basically that it was basic, obvious use cases of AI. I said, you know, bring Sam Altman up and I was looking for more fireworks.

3:23And so that's camp two. I think I guess I'm in camp two. I'm curious, you know, after having watched this, like which one is right? It's funny you frame it that way, because I almost feel like in trying to think about this today, the day after, 25 hours later after the keynote, I'd written something, you know, immediately afterwards, just trying to, to distill my thoughts really quickly on like the, the gut reaction to it. And I thought to your points, like overall, it felt pretty right down the middle. Um, in part, because as you mentioned, like Mark Gurman had broken so much of what was going to be announced and it went pretty much straight down that, that line.

4:01And it felt like, you know, perhaps a little underwhelming as a result of that, for those of us who like follow this so closely and again, had read all those reports and knew what was coming. Today, I'm a little bit, I'm almost in exactly in between your two camps. And I know that because I'm like trying to think about what my next sort of take to write about will be. And I sort of have these like two paths that I'm potentially want to go down. And I'm not sure which one I feel more passionately about. I'll probably end up writing both of them, to be honest, because I do feel like it's a very nuanced situation.

4:34But I do feel like, you know, in one camp, I have sort of the headline in my head of like, Apple failed to overreact to the AI revolution, which I think is, you know, a tongue in cheek way to frame it. But I do think that that's, that's sort of what we're looking at right now, right? Like everyone wants some giant, you know, new feature, some big partnership with open AI. And instead, they did what Apple often does, which is just sort of go down a very middle of the road, focus on sort of products and implementations that are going to hopefully resonate with their very big user base and not focus on the latest and greatest whiz bang features and instead do stuff that they think the masses will resonate with.

5:18At the same time, the other angle that I'm thinking a lot about is what Apple showed off looked good as it often does, right? Like visually, like appealing. And, you know, I think they made some good product decisions on what they showcase yesterday. But I'm also fairly skeptical about how well they'll actually be able to execute on that simply because they've been in market for years and years with Siri. And we've been told again and again that, you know, this is the year that Siri is better. This is the year that they fix all of the issues. Exactly. And the year of Linux. It was the year of mobile for like 15 years.

5:57Yeah. And I am pretty skeptical having watched the depth. So I watched the demo, you know, remotely. You're there, but I'm in London. So I was watching it just on the live stream. And then I went ahead and watched it again this morning just to see like, you know, second reaction, if I missed anything. And watching a lot of those demos, again, they look impressive, but I'm just watching them And I'm like, come on, it's not going to work as well as it is in this demo. And obviously, that's always true of a demo. But I am skeptical of Apple's ability, you know, maybe unfairly, but I don't think so, given the history with Siri, that they'll be able to execute on this.

6:32And I think in part, what you're seeing, what you saw yesterday, and especially the announcing that a lot of this stuff, it's not just that it's, you know, not going to be available on day one of iOS 18 and macOS and the other operating systems. it's going to roll out over, you know, the course of the next year. Basically, they gave these nebulous dates. And in some ways, that's probably right and good because that's not what Apple normally does, right? Like this iterative approach to software development. They instead focus on these key dates where they, you know, roll stuff out and make a big splash.

7:02But I feel like with AI, you have to sort of do it in a much more iterative way because, you know, models change and there's always sort of new states of the art. And so, again, in some ways, I think that that's good. But I do think it also speaks to they're rolling this out before they're ready to do it. And I think that, you know, the good news there, again, is that a lot of us are we're not going to see it for a while because the betas are out. I downloaded the new iPad OS beta because I have an iPad backup iPad that I'm not using for day to day work. And, you know, a lot of stuff's just not there right now.

7:37It's just not literally not in there. So can't even try. Right. I'm going to blow up my interview flow for this one because I was planning to leave this stuff for the end. But screw it. Let's just talk about this. Because one of the things I was thinking about all through the day yesterday is this kind of saying in Silicon Valley that it's AI until it works and then it just becomes technology. And on one hand, that's true. And on the other hand, that's not been the promise of AI. The promise of AI has been that it's not just working technology is that, you know, basically we're going to be in dialogue with these devices and AI is going to be like our own little agent and our bot with us.

8:18And, you know, eventually we're going to reach like maybe human level intelligence and we're going to really feel it in a way that like you can't really feel, you know, your phone. Your phone is a cool device, but it's not like this sort of almost sentient type of thing with you. And, you know, maybe my expectations are too high, but I also think that is it possible that the sort of basicness of Apple's announcements really show the limitations of artificial intelligence? There's this or at least today's generative AI. There's this feeling among people in Silicon Valley and in the tech world that we're really going to need basically 18 more months of model refinement before this stuff reaches like kind of the potential or reaches the point where it meets the hype.

9:01Like even though GPT-4 and GPT-40 models, which are the cutting edge, are cool, they're not, you know, exactly at the end state or they're not what people have been building around. And so I'm curious if just like what we saw from Apple is just the manifestation of the limits of the current stage of technology. And if so, what does it mean for AI right now? I definitely agree with that. I think, and I've written about this from a few different angles, both from sort of the investing in AI perspective, from Apple's perspective. We all hear all the hype. It's all hype all the time with AI and, you know, in some ways justifiably.

9:41So there's really cool demos that are out there like the GPT-40, you know, that OpenAI did a few weeks back. Really, really interesting stuff with vocal computing and sort of the promises there. And, you know, Scarlett Johansson and her stuff aside, it's a pretty cool demo. And it sort of gives you a sliver of vision into what the future of at least one aspect of computing could look like. But at the same time, to your points, I feel like everyone's a bit ahead over their skis right now. And I think that these models are both changing so fast and there's so many companies being built around this space on top of these models.

10:27There's so much funding going in because every investor basically has to believe that this is the next big thing. And there's so much capital out there flowing in where I feel like we're at some point relatively soon going to hit not necessarily a wall, but sort of a strong headwind of reality. And I feel like to your point that Apple's announcements in some ways are indicative of that, which is that there is a lot of cool bleeding edge stuff that's going on right now. I have no idea how you productize it in a way that's sustainable. There's a lot of cool demos that pop up that are popular for, you know, a day, some a week, some a month.

11:11You know, even chat GBT itself, you know, came out of the gates as the strongest adopted, you know, by some measures piece of technology ever. And then, you know, it's still an impressive technology. It's still being used in impressive ways, but it's not growing at the way that it was by most accounts, you know, in those early days. And so, again, I think that there's a lot of stuff that makes for good demos. There's a lot of stuff that leads to sort of excitement. But ultimately, how are these things going to help users save time, get things done? And I think Apple focused on that a lot. And that's what Apple does.

11:49And I think that that's a very smart thing to focus on for a longer term. Now, they could have, again, sort of going back to my notion of the headline that they failed to overreact to the AI revolution. I think, you know, if it was a different company, you can easily see a world in which they go all whiz bang and try to create all these new, you know, new features and functionalities based off of AI. But really what they're, it sounded like, you know, the messaging from the top on down was basically like, we're trying to figure out what we can do for our user base that they'll actually use and not just be, you know, a great demo.

12:22And in some ways that's underwhelming because again, a lot of this stuff was leaked, but it's also been stuff that has been table stakes. Google and others have been doing a lot of this stuff within Gmail and, you know, predictive text and things like that for a long time. And Apple sat back and waited. And I don't think it's ultimately going to be a mistake that they did that. Because again, exactly what we're talking about. I think that we're going to at some point face those headwinds where we realize there's, you know, cool demos you can do. But from a product perspective, there's, there's, it's going to take a long time to get to sort of that next echelon of computing.

13:01Yeah. And I didn't, I was kind of holding lots of different feelings at the same time today. I had like very, uh, quite a morning where I just kind of sat with all this stuff. And I had like these feelings of like, this is amazing. Like for instance, being in some of like the serious, serious stuff as well. Like you say, you know, when is my mom landing? And it will like search your emails and get her flight data and then check when the flight is coming in. Or, um, you say, uh, this was my favorite one. Find that podcast that my wife texted me or messaged me about the other day and it will find it for you.

13:35this stuff is amazing it makes devices super useful but then i also like wrote a note here that it sort of gives me a vision of how this fizzles and it's like super cool it's great ui but it is it really just like a better you know smart notification i know that's selling it way short but yeah but it is interesting where it's like it might be that the hype has been so great right we hear about like agi and this stuff's going to turn us to paper clips you know from anyone working at an AI research lab. But if the, if the height of the product, like the best vision that Apple can show us is that, you know, it will take a, and this is cool, right?

14:14It will look at like the fact that a meeting has been slapped on your calendar and then check the traffic until you're going to be late to pick up your daughter. Like that's amazing stuff. And I'm still trying to figure out whether that feels evolutionary or revolutionary to me. Yeah. I, I think that, um, Um, the, the stuff that we're going to end up using in the short term, you know, with, with regard to what Apple announced and, um, and AI, I feel like the things are almost, you know, the, the whimsical things like the generative emoji stuff is, is going to be like, you know, the stuff that's actually used.

14:49Whereas the stuff that you're talking about that they demoed, you know, or previewed because it's not going to come for a while, like the multi-step sort of, yeah, you know, my mom's on a plane type thing. That's the stuff, by the way, that I'm fairly skeptical that Siri will be able to execute upon. And it's like if it fails at that once, you're never going to do it again, right? Because it's more of a hassle to get the wrong information than it is to just do the laborious work of actually looking up the email manually. All that said, I do believe that there are some, you know, low hanging fruit stuff.

15:23But it's a lot of like just sort of big data problems. Like I think Aaron Levy talked about this. And I know you did an interview with him recently. And but he had a tweet recently about, you know, on the enterprise side, of course, like why AI is so exciting. And it's just like, we live in a world where there is just an increasing amount of data, both, of course, you know, within the cloud on enterprise side, but in our personal lives, like, I have 40 some thousand pictures on my iPhone and, and, you know, Apple showcase, like what you can do, um, you know, with, with those photos now, which is interesting, but it's also just like, I know, I remember, you know, in my head, a time with my family, like a few years ago, and right now, sort of the way to do that, you can try to do, you know, like a, a granular search of, of a date, but it's, it's a little hard to remember the right syntax to do that.

16:09Um, the easiest thing I do now is actually use Google photos and sort of go to a place because they're all geo tagged, right. And sort of zoom in on a map and find a photo that I was looking for, which is weird. Like, and you should be able to do that just by talking, um, or by typing something in semantically and it being able to find that. And I think that those are, those are the things that are, that are going to be low hanging fruit that we're going to be able to accomplish. But again, that sounds like it sounds boring when you sort of say it out loud like that, but it, I think it can be sort of profound from a technology perspective, but it's these things that are, um, you know, again, not, not the whiz bang features.

16:46So one way we're going to be able to measure how desirable these features are, right? This predictive assistant that we're going to get with Siri and the ability to generate emojis and the ability to create images of people in your context and have Apple help you write with its assistive tools. And there's the smart notification prioritization that they're going to do with AI. We're going to find out really soon, like whether this is desirable from a customer base because Apple's had effectively a slowdown in iPhone upgrades. I think you wrote about how they've basically saturated every market that is, that lends itself to buying iPhones and people aren't upgrading as fast.

17:24And they're going to make this available on just the iPhone 15 Pro. And of course, like some of like the, like older laptops with M1 chips, but I mean, it's the phone that matters. And we're going to see whether people, you know, in the fall, this is, this is going to come out in the fall and of course it's going to release more iteratively which is interesting but um we're going to see whether people upgrade to the 15 pro and especially the next iphone event is going to be the 16 and wall street's had like a weird really the kind of the strangest reaction to this so on actual wwc day the stock was down 1.91 and we're recording tuesday morning stocks up 5.43 on like no news other than the event and a major down day for the rest of the market right so do you think that this is going to drive that upgrade cycle i know it's hard to predict but i'm going to ask and then uh does this kind of i mean apple's now solidly in second place it's like really surpassed nvidia does this quote-unquote save apple's position in the market i think so first and foremost um you know what what i was talking about in what you were alluding to is the fact that Apple doesn't necessarily have as direct of a way to monetize off of all the AI capabilities, right?

18:42And some people were waiting yesterday to see, like, would they be selling AI as a part of a services narrative, right? Like, would they be upselling certain features? And it looks like not only is that not the case, from what they were saying, it sounds like, you know, you'll only have access to the free element of OpenAI. And you can bring your own premium version if you want, but they're not even necessarily going to be upselling you or allowing OpenAI to upsell you on that. And so the question for Wall Street would have to be how Apple plans to monetize this. And to your point, the most direct way that they can do that now, not planning to sell this as a service, at least not anytime soon, would be through that upgrade cycle.

19:28And yeah, you would hope that they would hope that these features, because they're restricted to newer phones, the new ones coming out and also last year's, you know, pro models will lead to the quote unquote super cycle upgrade. To me, I think that that only really happens if there are features that are live when those new phones come out that people want to use. And so again, I looked to like the Genmoji stuff, how they handle that. They didn't really talk about it. I don't know if any of the other, if you went to another panel or, um, you know, like the developer state of the union, I watched some of that, that video online, but like from what it sounded like, or what I would guess is that they sort of restrict that to iMessage.

20:14And it's not like, um, you know, you can't cross that to other, you know, WhatsApp or, or text to Android phones. Um, you know, the way that they would do that is obviously, you know, just basically send an image, I guess. But it sounds like technically speaking, it's not it's not an actual emoji, which is just text that, you know, the consortium puts together every year and rolls these out. They're sort of doing this clever like hack together version of it, which I think looks good because I think that they smartly made it look like what their emoji style is. But I think that's the type of feature that might get people interested in like seeing what their their friends with the new iPhones are doing.

20:53and then the the only new app at least that they talked about was image playground right and that's like the the thing to be able to create images which you can do it sounds like via some of the other apps themselves but they have an actual standalone app to be able to do that too so it's like those types of things which are forward-facing are the ones that are probably most likely to drive you know user interest in upgrading the stuff like predictive text and things like that i mean it's hard to see how that that would drive someone to spend upwards of$1 ,000 to buy a new phone when they don't need one.

21:28And so again, I have to think that it's going to be the much more forward-facing stuff, if anything. And it's a little bit hard to see that right now. I do think that that might be what Wall Street is getting excited about, that again, they're restricting these features to newer generations of phones and the new ones coming out. And so it will certainly lead to some sort of bump, but will it lead to a super cycle? I'm not so sure. All right. And why don't we talk a little bit about this open AI thing? I mean, that to me, the more I thought about it, the more puzzling it was to me that open AI was sort of, I mean, of course, in press reports played up as a major part of this announcement.

22:06I don't think they did much to put fire, you know, to try to douse that fire. And it turns out that all of the AI that you're going to see on the iPhone is going to be run with Apple models, a lot of small language models, right? Which is interesting. Not the big LLMs, but the small language models running on your devices, which is basically the same technology, but smaller, more efficient. Doesn't need to have the whole corpus of the world's information to run. And then there are going to be instances where Siri will suggest that you use ChatGPT. Like that was the extent of the open AI announcement.

22:38What's going on there? So, and it goes back to when you and I last spoke, because that was around the time when there were several papers were published about Apple's LLMs, right? And, you know, they got buzz at the time, obviously, because it's Apple and, you know, it's rare for them to put out sort of any sort of look, you know, behind the curtain of what they're doing. But that's sort of the state of AI and what they needed to do in order, you know, to attract talent and to keep their talent around the table. And so in a way that was interesting because that is, you know, more or less what they ended up sort of going with, right?

23:13They have their own several different sort of flavors of their own LLMs of varying sizes, certain smaller ones, obviously to run on device. And then the cloud ones that can be larger. And their secret sauce there, it sounds like is basically, you know, some sort of algorithm to determine whether it goes to the on-device LLM or whether it goes to the cloud or, you know, and I'm very curious to see how responsive speed wise this is to know, you know, like if they can basically instantaneously know if it goes to on device, to cloud or to open AI and how they sort of determine, you know, that that level of calculation will be fascinating to sort of see from a product perspective again.

23:59But yeah, I mean, again, it was what we talked about in that conversation. And, uh, I was sort of wondering and trying to talk myself into again, like the idea that maybe this open AI partnership was much more deep. And what if they even just used, uh, GPT four Oh, as the basis for Siri now, um, based off of that demo, you know, that was pretty compelling. Um, and again, what we know about the historical version of Siri, which is not good. And so, well, you know, was Apple willing to do that? And of course was always going to be a bridge too far, but it's still like, were they going to augment in some ways?

24:36And that's sort of what they're doing, but they're doing it in a very, um, you know, sort of bound way in that I'm not sure that you're going to need to opt in every single time, but it sounds like you might, right. In order to send anything to open AI, where you basically have to confirm that you're okay, sending it to them. Um, and you are acknowledging that the results you get back are from them and not, not from Apple. And so again, that's a, it's a very Apple way to do things, but it all comes down to what the product execution is both in speed, obviously in results. And then yeah, how they, how they do handle the UI, including like, you know, voice and, um, and everything else, uh, aspect of the UI.

25:14Yeah. And I actually kind of liked the fact, you know, in retrospect that it was weird for open AI that they were so small, but for Apple, I think it's good that they thought well beyond, um, sort of like a chat bot that gives you information and more just like how to use gen ai to like figure out what's going on on your phone go through all this all this text be more useful like this personalized intelligence like i mean we've talked i guess through most of this podcast about how it's like okay whatever but um the strategy is i think it's a good strategy for apple even if it's not gonna knock our socks off and it's like something that can get better as the technology gets better especially if we do see that leap in models in 18 months.