What Apple's Failures Say About AI's Limitations —  With M.G. Siegler

0:00Are Apple's AI failures happening in a vacuum? Or are they a damning sign of the limitations of AI itself? Yes, we're going there. Plus plenty more observations from WWDC. That's coming up right after this.

0:17You'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 Buehler. 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.

0:45Welcome to Big Technology Podcast, a show for cool-headed and nuanced conversation of the tech world and beyond. Let's break down what happened at Apple's big developer event this week and look at whether the company's AI limitations say something worrying about AI itself. We're joined today once again by the great M.G. Siegler, author of Spyglass, which you can find at spyglass.org. M.G., welcome back. Thanks so much for having me on, Alex. All right, let's just get quickly to the event itself. I'm just going to go read the last line of your post to begin with, because I want to get your reaction as to what we saw from Apple, because they did have their big annual developer event this year, introduced a new user interface, said shockingly little about AI.

1:31I think you said that they said more about the phone app than they did about Apple intelligence. And your conclusion is overall, I'd probably give this WWDC two out of five stars, but it easily could have been a zero due to their mistakes of the past. So quick reaction to WWDC and why just two of five? Yeah. So, and obviously that was an allusion to their ending, which they did this cute little ditty song made up of their collating different app store reviews for different real apps in the app store with a six out of five, which is a very catchy song, nicely done. But that may have been unfortunately sort of one of the highlights of the event right there at the end, because, you know, this, of course, was the elephant in the room event where there was all the talk leading up to it that Apple just would be shying away largely from AI this year, given what happened last year at WWDC, where, of course, they may be overpromised and ended up under delivering on what was actually promised.

2:31And so, you know, when I say two out of five stars, I'm giving Apple, you know, a little bit of benefit of the doubt here, just because like, it easily, I really think could have been a zero five stars, given how little they had to talk about AI. And so, you know, they went back to sort of one of their historic strengths. And we can talk about this more in a bit, but going to UI instead of AI, right, going to the overall sort of look and feel of their operating systems, again, what has been considered a strength for them in the past. And I think they were able to sort of smartly navigate down that path, giving people things to talk about when they weren't going to be talking about, again, that elephant in the room.

3:09Now, they did talk about Apple intelligence. I don't know if they ever used the word AI as, you know, last year they sort of famously said or didn't say that, you know, AI, they went after Apple intelligence, their own branding of it. They did talk about Apple intelligence quite a bit. But sort of my interpretation and read of the situation is that they were just totally sort of downplaying it and skirting over, you know, what's actually obviously being talked a lot about in the industry and sort of just making it just like, you know, this little feature that we don't really need to talk about this year because there's just little things that it's doing for you and it's not that big of a deal, which of course I think everyone outside of Apple would disagree with.

3:49I think that's so perceptive. Like one of the things that we saw from Apple last year is it at least realized that was one of the companies that could build this sort of masterful AI product drawing from all different information sources. And then if you look at the AI perspective this year, it was much more, let's do a couple of features that aren't sort of tied together in a coherent message. This is from Twitter user Signal. Watching this WWDC felt like watching a dinosaur try to do ballet. Technically and aesthetically competent but spiritually lost they talked about apple intelligence that like like they just discovered the concept throughout the presentation there was zero ecosystem thinking and no real attempt to rethink a sitting a single interaction model i mean this was really the problem right it was just like that that um and obviously they've had problems executing and we can touch on that uh but that broader vision of what ai can be it's just gone yeah and i mean if we take a step back and they sort of kicked off with this.

4:52Tim Cook kicked off with it. This is a developer event, right? And developers are arguably, have never been more excited to talk about anything than they have been probably about AI, you know, certainly since, you know, the introduction of smartphones and, you know, the web before that. But like AI arguably is even a step above those right now because the developer ecosystem is so much larger than it was back in those days, thanks in part, obviously, to the smartphone and all the different technologies we have now. But so this was supposed to be a developer-focused event. And again, you hit on it, and I joked about it, but it was like Apple spent three minutes talking about a phone, the phone app.

5:28Like, who uses the phone app anymore? Like, yeah, there's some nice little cute upgrades, but we didn't need to talk about that for three minutes when we talked about Apple Intelligence for roughly three minutes to kick off with. And so, you know, that dichotomy, it seemed very telling of it. And they just didn't have a lot to talk about with regard to developers and Apple Intelligence. They did talk about, you know, being able to leverage some of their on-device models that had leaked, of course, beforehand. I think Mark Gurman, as with most things, was able to scoop that ahead of time. But that was sort of the one bone that they threw to developers.

6:04And it's unclear, you know, how useful that will be. I think, again, at like a high level, I think it's interesting to have on-device access to models, you know, for all sorts of reasons. But are they really going to be as powerful as some of the other state of the art models out there? Obviously not. But can they be used for sort of other things that developers will actually find useful? That's obviously very TBD at this point. I just want to put this all in context because Apple has failed so bad on rolling out Apple intelligence that I think there's this belief out there that they had a mind-blowing presentation last year.

6:41But they didn't. And last year, okay, a couple of things. First of all, your headline was Apple fails to overreact to the AI revolution, which was like they had a bunch of features. They did have this ecosystem thinking, but even that would have kind of caught them up to the modern day thinking of a company like Google. So it wasn't like they were promising the moon. You know, it doesn't seem like they even delivered a lamppost. And then, you know, even just like looking back to last year, I did this poll. We talked about it on the reaction show. I asked people, your reaction to WWDC AI stuff mid-event, wow, cool, or meh?

7:21Meh was the winner, 50.4%. This year, I asked again, Apple WWDC reaction not even talking about AI. And now maybe my audience is primed because they've heard me talk about Apple for a moment, but it's probably the same audience as last year. and MAG gained 30 points, a 30 point increase in MAG. This is a landslide. It came in at 81%. So talk a little bit about sort of where Apple's vision is this year compared to last year and why you think they haven't even met these sort of muted projections for what they want to do with AI. Yeah, so you sort of hit on it, you know, a bit ago for a second. The notion last year, it felt like was, look, maybe Apple has a window of opportunity here to actually step in and productize some of the work that's been done underneath with AI, right?

8:11There'd been all the model breakthroughs. And obviously, there was ChatGPT doing some real consumer-facing work with AI. But a lot of the other stuff was underwhelming and or just seemed very disparate and not sort of put together in a nice, cohesive package, which, of course, Apple is famous for doing, coming into a market a little bit later than everyone else and sort of putting together a nice package around, you know, stuff that's not exactly new, but that they could sort of put it together in a new compelling offer, all-in-one offering. And so I think that was some of the hope after last year's WWDC.

8:46Again, a lot of people sort of felt it was still underwhelming. And certainly people, I think, with a more perhaps technological bent felt like, you know, it wasn't sort of up to the snuff of some of the other giant companies, what they were doing. But had Apple been able to execute on some of that, you know, maybe we would have been looking upon it differently now. Of course, they didn't, as you're sort of talking about. And so this year, yeah, it was basically going into it, you know, with that baggage hanging over them. And again, you know, sort of the maybe hope that they would do something to surprise and, you know, something that possibly hadn't leaked ahead of time, but that ended up not being the case.

9:25Every single thing had leaked, which I do think is a problem, increasingly increasing problem for Apple. And in particular was a problem here because, you know, had everything not leaked, maybe they could have gotten away with a little bit more leeway with the audience. It's like, you know, oh, they hadn't unveiled the liquid glass UI elements ahead of time. And, and they hadn't known about the name changes ahead of time. You know, those are silly things compared to sort of what's going on with the overall AI revolution, but it still would have at least given them some air cover, I think, with talking points on the media circuits and whatnot that, look, Apple's doing things a little bit differently.

10:02And maybe it would plant a seed of doubt in people's minds that, look, maybe this AI stuff still is too early and maybe Apple shouldn't be so focused on it right now. And that was, again, some of the talk after last year, maybe Apple was just getting in at the right time, or maybe it was still even a little bit early and they were doing things a little bit pressured by Wall Street, which I think in hindsight ended up being a lot pressured by wall street right that's why they couldn't sort of uh uh you know execute on some of those promises but this year again they just had they had none of that benefit of the doubt and so you're going into this looking like they're just sort of not talking about it because they literally can't right now because they had to rejigger the whole team they have to sort of if not completely start over from scratch at least sort of look at everything from you know starting over from scratch standpoint and think, you know, much like Amazon, it seems like had to do perhaps with Alexa sort of figure out, do we have something here that we can salvage, you know, notably in Siri, or is there, you know, are we going to really have to rethink this from the ground up?

11:04I want to give you some credit here because there were two reactions that you had on our show last year. First of all, was Apple fails to overreact to the AI revolution, spot on, even though they haven't even been able to meet their lack of overreaction. The other thing that you said, and I'm going to quote, is there's a lot of cool stuff. I just don't know how you productize it. You were skeptical from the very beginning of Apple's ability to build the things that it showed in these demos last year. So let me just use that as like a jumping off point to ask you, and then I teased this at the beginning, and I think this is really important to discuss, to ask you about the limitations of AI itself, because we saw Apple make this promise.

11:45We saw Amazon make a similar promise in February of this year with Alexa Plus. And some people may have Alexa Plus. I haven't met them. And I'm waiting months after they promised it would roll out, not here. Google might be the most far ahead, but Gemini has severe limitations, if we're being honest. I mean, if you're using Gemini and Gmail, you're probably going to be disappointed. Sometimes it works, but most of the time it's like, I think this should be better. So here's the question for you. Have we gotten carried away with what AI can do? Because if you think about what AI has actually been useful for, the chatbots are great.

12:25I think everybody agrees that. Everyone who uses them, ChatGPT's O3 model, that thing's amazing. But where has AI extended beyond the chatbot? Maybe a tiny bit of agentic work in enterprise, But beyond that, we haven't seen the ability for these companies to build these promised smart assistants, despite, I guess, we see another company talk about it at a developer conference, you know, every few months. So what's going on, MG? Yeah, I mean, I think that some of this is exactly what we're talking about with Apple, Amazon and Google. So all three of them notably had, you know, assistants in the past.

13:03Right. So there was Siri. There was Alexa and Google had, you know, the Google Assistant and Google Home products, too, which sort of everyone sort of, you know, talks past now. But that was also, you know, a multi-million unit device that was in a lot of people's homes and still is in a lot of people's homes. And all of them, I sort of chalk up to having major problems sort of upgrading those both devices and mentalities because they had some level of success. Now, we could argue about, you know, series level of success, but it was the first mover in a way, right, in the market. And, you know, it got a lot of buzz and had some early promise.

13:42Alexa came along and sort of blew it out of the water. And Amazon just had a different strategy of ubiquity and getting these cheap Echo devices in everyone's homes and everywhere. And then Google sort of matched that general strategy with assistant stuff. But again, that in a way became an albatross around all of their necks, because when it came time to this new wave of AI, they basically had all these devices that they still had to service and hope to upgrade in some ways. And as it turns out, it looks like now in hindsight, obviously, it's just a lot harder to sort of go back and try to shove this new technology into that old technology, even though on a surface level, they seem like they're similar.

14:26It seems like, you know, behind the scenes, they're not similar at all. And in fact, there's a lot of stuff that's like so simple that Alexa was doing with like timers and stuff that it's actually harder in some ways to get some of the newer technology to do. Now, to your broader question about products, I just speak to that because I think a lot of those companies had the mentality of, look, we already productized some of, again, the high level of what these assistants and chatbots are doing with Echoes and HomePod and things like that. So why can't we just do that again? Well, as it turns out, first of all, some consumers were turned away from those after trying them for so long and just not getting a lot of utility out of them, right?

15:07Beyond the timers and playing music, it's like, again, there's sort of this weird problem of success, quote unquote success, in that the people were trained to use these devices and now can they really go down that same path? And so, you know, I think the big companies are all sort of stuck in this weird position where they tried that path and it sort of worked, but it didn't work well enough. And it's probably not going to work exactly correctly with the new technology. And so do they try to use, you know, new devices or do they try in the case of, of course, Apple and Google, do they try to shove, you know, this technology into smartphones, which they're doing Google to a greater extent, of course, than Apple is right now.

15:50But what ultimately do the products look like there? Like you're talking about, is it just like, you know, being able to autocorrect Gmail or write emails for you in Gmail? Or is there sort of something new that should be done? And that, of course, gets to what, you know, Johnny Ive is potentially working on with Sam Altman and with the OpenAI team and the whole acquisition of the IO company. and, you know, by all the accounts, they're trying to build a newfangled device, right? Something that's completely new from the ground up that's not trying to, again, cram this new technology in an old device.

16:22And, you know, who knows? Obviously, we can all guess as to what that will be and how that will work. No one will know until it ships, including them. You know, they're talking a big game about it, but they always talk big games about all this stuff, especially Sam Altman, of course, as he famously does. And so it's really going to be hard to know for sure. But I do think that that path is probably a little bit more clear than what the, you know, the quote unquote incumbents or the big tech companies have right now, because, you know, they just they have all of this legacy and institutional knowledge of like what they've done in the past to productize different technologies.

16:55and I'm just not sure it's going to work that time because in a way, we kicked off talking about like AI versus UI. In a way, AI is a new UI, right? Like, and it's weird to think about in that way, but it really sort of, you need to, I think, think about it in that way. And so if you're trying to cram AI into sort of the older paradigms of UI and thinking like, oh, we need, you know, this beautiful interface or we need, you know, like an app specifically to do that. You know, we'll see. That will work for some services, I think, but not for everything. Yeah, so we might talk a little bit more about this Johnny Ive device, but just to touch on it quickly, because I've talked about it a lot in the newsletter and on the podcast recently, we had a great YouTube comment.

17:41And honestly, it's a little coarse, but sometimes this says everything you need to know about the internet, where a guy said, Unless you can tell me dudes can watch porno without a high definition screen, the AI device is a hoax. Fair. I think that's a fair criticism. Porn has driven a lot of technological adoption, of course, famously over the years and what that looks like here. But, you know, again, like this is not a new and exactly new thought. But I do generally agree with like the notion of if we consider that sort of AI and UI blending together in some ways. I think that AI ends up working in some capacity on a product side because it's going to be all different sorts of different interfaces and devices that we use it on.

18:29It's not just a phone. It's not just a computer. And it's not necessarily just a new device of whatever these various other companies, including OpenAI, are working on. There's got to be a screen involved. It'll be a screen and maybe a device like Ive's device will not have a screen. That's what the reports are. No screen, not a wearable. There's no way that becomes the device. Is it a device? Maybe. Probably, actually. Or maybe that entire user interface just happens with your phone speaker and microphone. Yeah, I mean, look to what Amazon's doing right now, right? I think you've talked to Panos Panay about this.

19:08He demanded I get a screen device. Right, exactly. And that's how they're rolling it out, to your point, if they're actually rolling it out. I mean, I think that a few people have it, but it's unclear how many. But they're restricting it heavily because they're restricting it to the newer Echo devices with a screen in particular. So I want to ask you if you're going to account for any probability that the AI technology itself just can't handle this. And I'll give the argument that I think would sort of capture why LLMs might not be able to do this. It's a lot of information to take into account.

19:46Think about how many emails we have, the data we have in maps, how many text messages we have. Is it possible that they simply cannot put the attention? Now, we have a million or two million, I think two million, token context windows. But is it possible that there's just, you know, these assistants that are trying to capture all your data cannot work reliably trying to handle such a huge amount of information at once? I think it's a good question. It sort of reminds me of something I wrote a little several weeks ago at this point. But when it was around some news that basically when all of the AI companies were starting to roll out memory features, right, OpenAI did it and Google's done it and a few others, Google like doubled down and tripled down on that at IO, right?

20:35Talking about how now it'll have the context of like pulling in Drive, pulling in Gmail and everything. And like, to me, when I hear that, so at a service level, that sounds like, yes, of course, like I want, you know, my AI to know everything and I want it to be able to have the historical context. My thought a few weeks ago, and I still agree with this, is that I wouldn't be shocked if that sort of backfires against these companies in some way. Like, I'm not really sure that people want to use this new technology to sort of dredge up the past and have all of that historical context. Now, obviously, there's the famous examples like dating back to Facebook of like, you know, servicing, you know, old girlfriends or servicing people that you don't want to see anymore and that kind of stuff.

21:22But I think even beyond that, just like everyone evolves as a person a lot. And if you have the context from your Gmails that you sent in 2004 or 2005 or whenever Gmail actually launched 20 years ago, is that really going to inform your interactions with AI in 2025 and 2026 and beyond? And what does that mean? Like you would hope that the AI is smart enough to know, like maybe discount something that you said in 2005 because you no longer believe that for whatever reason. But maybe it doesn't know that, right? Maybe it weights it just as heavily. You know, certainly there's there's been plenty of of fits and starts with different AI products where they sort of don't quite get the context that they need to with regard to sort of the way that you're using AI, you know, dating to the the Bing chats hallucination stuff and trying to get Kevin Ruse to leave his wife or whatnot.

22:23And so there's all sorts of historical precedent for why you might not trust these chatbots to sort of take the proper context into account when utilizing these things. But again, to your exact question, I, you know, I do think that there's, there's a concern that they're going down a path right now, thinking that it's the right thing to do when instead, like, maybe even something like OpenAI, which is obviously a much newer product, shouldn't be so heavily thinking about and worrying about all of this memory stuff and historical, you know, stuff to pull in there. Let's go to one more YouTube comment, because I think this sort of gets exactly to your point.

23:03We had somebody in my recap from the event on Monday says, they said, I don't give one flying flip for AI on my phone. Apple should just leave it out altogether. I don't want AI reading my messages. I don't want AI looking at my photos. I want AI to get the F out of my life. I wonder what percentage of the population that comment represents. Yeah, I mean, there's a lot in there, right? Like, I've seen that too. I've seen that general comment. People have left me similar comments, right? Like, especially about Apple. It's interesting, especially about Apple products, because I feel like, I don't know if it's the sort of historical privacy messaging or even the anti sort of meta and Google messaging, right?

23:46where it's like people who are, you know, diehard in the Apple ecosystem almost, at least historically, have viewed Google and Meta, you know, as sort of the villains, you know, who are just all about data hoovering and, you know, bringing this all in to feed their business models. Obviously, there's some level of truth to that because that is their business model. But there's, of course, a flip side to that, too, which is Apple's not the full-on white knight in this regard who's doing everything for purely above-board reasons. Everyone has their own agendas, right? And everyone has their own business models, and everyone has reasons.

24:30All companies have reasons to do what they're doing. But to that comment, yeah, I do think that there's a subset of people who look at this stuff with AI and think like, is this Apple sort of delving into a world really that's more of the Google and meta mindset with data and everything else? And like, I don't want any of this stuff. Now, I do think that there's a bit of a Luddite concern there, right? Like that this is just people not fully understanding or not liking the notion of having to learn a new thing or worried that, you know, this is just not going to be useful at all. And again, to our earlier points, like it's up to these companies to prove the usefulness of these products.

25:10I do think, I think that ChatGBT has done that. And I think that there's a reason why, you know, where there's all this talk about Google search. And I think Google's reactions to this and shoving AI front and center into search are more indicative of anything that they're saying out loud. It's just follow what they're doing. Like, it's obvious that this has worked to some extent, but will anything work beyond that? You hit on some of the other things, generative stuff, you know, some of the video stuff is cool, but it's so early that it's like, it's just not there right now. And so the only thing that is there right now is really chat gpt and um you know what what does that look like productized i mean open ai already did it and so either apple just you know fully more fully bakes it in which i still think by the way they should probably do with siri while they rebuild siri right just totally outsource it to chat gpt we did see that yesterday announced like that was sort of one of the more subtle interesting things like they did a bunch of more smaller partnerships uh with chat with OpenAI to integrate ChatGPT and notably with Xcode, where all the rumors had suggested that it might be Anthropic and Claude that were sort of being baked in.

26:21And it's probable that they're still sort of testing that behind the scenes, but they rolled it out, it sounds like, with ChatGPT. And so it's interesting that they're going further down that path with that company. But again, that sort of makes sense. Again, that's the product, that's the brand that's resonating with people right now from a product perspective. Right. I had this conversation, like along those lines, I had this conversation with someone on the broadcast riser yesterday when I was out at WWDC or on Monday, where they were like, well, I don't need an AI phone. I like my phone as it is.

26:53I was like, yeah, well, you don't have it. And that's sort of been the history of technology is that we, the best companies will give us things we didn't know that we wanted. And all of a sudden they become essential to our life. And that's - Absolutely. I will just give you one quick anecdote on that exact point, which is my own personal usage of, believe it or not, back in the day before the iPhone launched, I was super skeptical. I watched the keynote like everyone else did. I wasn't there, but I was watching it from afar. And I just thought, like, yeah, it looks cool. There's no way I'm buying a$700 phone.

27:27Like, you know, I had a Motorola Razr flip phone. I was like, this is great. It's super small. It fits in my pocket. That's all I need. Like, what am I going to do with a$700 phone? sort of like Steve Ballmer, I guess I sounded like back in the day. And then day one, I happened to be near an Apple store. I was actually back home where I grew up in Ohio. And I just, on a whim, I saw a long line of people. So I was just like, yeah, I have a few minutes to kill. I'll go wait in line and just see what the deal is with this iPhone. I picked it up in the store and I literally walked out of the store with it.