576: Quiet Little Leech

30:07You're asking the wrong person. I can't. I don't see it either, so I'm not sure. I mean, this makes sense to me because the cameras on the Vision Pro, I mean, they're fine, but they're not going to be you know the the fidelity of things that are being drawn in vision os is obviously going to be as perfect as it can be because they're not passing through a camera and a sensor or whatever and yeah cameras with sensors that small especially if the room is dark they're going to have to keep the shutters open for a longer period of time to get enough light and that's definitely going to cause blur but what i had heard is for some people complaining that there was a blur on things when they moved their head including things like windows of vision OS applications.

30:43Yeah, in my experience, I think any kind of motion blurring is true of the Vision OS Windows as well. I noticed it in particular when trying to scroll text and read it in Safari, just within the Vision Pro Safari app. Again, it's not unusable, but if you notice the motion blur in other areas, you will notice it there too. We are brought to you this week exclusively by ATP members. Please consider joining and becoming a member today. Here's why you might want to do that. So for me personally, I listen to a lot of podcasts and my favorite shows, if they offer a membership program, I will almost always buy it because what they usually offer is an ad-free version of the show and usually also a little bit of bonus content.

31:27And to me, those are the two reasons to buy a membership. Our membership here at ATP offers both of those plus a few other things as well. So the ad-free version of the show, nice and easy. You get the same show everyone else gets, but with the ads removed. So it's very convenient. I always love that. I kind of feel like it's a premium experience when I'm listening to a show and there's no ads in it that I have to like, you know, skip over or whatever. Then we also do bonus episodes. These are episodes exclusively from members that come out about once a month on a pretty wide variety of topics.

31:54We do like tier rankings of different tech products. We do movie reviews. We do a couple of other like, you know, food things and a couple other fun things. So it's always a lot of fun. And if you like us and you like our content, you'll probably like a little bit more of it. And especially with the freedom to go off in a couple of different directions that we wouldn't normally cover on the main show. That's what the bonus episodes do. And we offer a bootleg feed. This gives you the unedited live stream recording. So this is everything that we record from beginning to end on the live episodes.

32:23It's basically the unedited raw version of the show. So you have a bit of bonus stuff on the beginning and end of that. Any topic or feedback that we end up cutting in the edit will remain in the bootleg. You get to hear Casey swearing. You get to hear any of my jokes that don't land that I edit out of the publisher version to make myself sound funnier. All that's available in the bootleg. And you get the bootleg right after we finish recording. So it comes out usually about 12 to 16 hours before the main show. You get a little bit early release in a more raw version of it. And a lot of our members actually prefer it listened that way.

32:52I think it's kind of cool, actually. So anyway, that's the main thing you're getting with our membership. Add free version of the show, access to the bootleg if you want it, and the bonus episodes for exclusive member content. Check it out today, atp.fm slash join. You can get it for just eight bucks a month or, you know, equivalents in a couple other currencies. We also have annual plans, gift memberships, all sorts of fun stuff. So once again, ATP.fm slash join. Become a member today. It is by far the best way to support the show. Thank you so much. And now back to the show.

33:28Gottfried Chen writes, I recently saw an interview using Meta's version of Personas called Kodak Avatars. They are basically photorealistic, but the scanning process is highly complex and not feasible with consumer hardware. I think it shows that the current technical challenge is the high-quality scan. The interview was done using Quest Pro, which I assume is inferior to Vision Pro regarding device sensors. So I wonder if Apple should offer high-quality Persona scanning stations in Apple stores so you can do an optional HD scan if you want. I mean, I don't know one way or the other, but I will tell you that I saw a couple of stills from this video, and my goodness, the quality is night and day better than the Personas.

34:03and i say this as a persona apologist well so here's the thing about the the the avatars these codec avatars obviously you know setting aside the fact that i don't know what they use to scan them maybe they use like the fancy laser scanner or whatever but it's not you know as we said the last time we discussed this the miracle of the personas is you take the headset that you bought and you just point at your face for two seconds and it gives you one this is something else but um the uncanny valley which people just throw out that phrase without re-explaining it just assume everyone knows what it is but the uncanny valley is the idea which i don't think has ever really even proven out but people just accept it because it feels right to them but anyway the idea is that as you make something more and more realistic looking it gets more and more appealing to you until you get really really close to it being like perfect like wow this is totally convincing as a human and you get really really close to it and then it takes a huge dip and that is the uncanny value it's like boy i was liking these avatars more and more as you made them better and better but then all of a sudden right when you got close to them being perfect now i don't like it at all because it's a scary death mask, right?

35:02There's a lot of questions with this. How close do you have to get to being perfect for the Uncanny Valley to take place? If you look at the original paper or whatever, it's just kind of like a wavy line craft and it's like they don't really make any strong stances about how close you have to get. It's just a general trend. It's like things look better right up until they don't, but then after you're perfect again, you know, people like them. And this is all about how does it feel to a person? Does it appeal to a person? Because you want your avatar to be appealing, not repulsive to a person. I say all this because these much more finely detailed, much more photorealistic avatars, the Kodak avatars from Facebook, are, I think, further into the Uncanny Valley than Personas because they are closer to being perfect.

35:44they're better like in that uncanny valley graph i think they are like over the edge of the roller coaster thing and screaming down in the middle of it like so you know the the uh the personas ones what they have going for them is that their blurriness and sort of slight cartoonishness helps keep them in my opinion anyway in my personal uncanny valley graph for these farther up the hill you know farther up the big dip that is the uncanny valley in that graph you know you can take a look we'll link to the the paper about this and facebook's video about it from what is it from a couple years ago a couple months ago um you decide whether you think that that is farther into your uncanny valley or not but part of the problem with doing like actual photorealistic humans where there is not a bunch of like fuzzy gaussian blur around everything and sort of you know a little bit of caricature is that as you know people who make 3d movies and games have known for many many years it's really difficult to do convincing humans because we know what they look like and things like subsurface scattering underneath translucent skin a technology that was sort of pioneered maybe 15 years ago and is now pretty commonplace is necessary for you to not look like you're not made of plastic and i think this is doing that but doing that well gets expensive real fast and then if you see these things in motion they're not you'll see the video of it but you can see the actual human wearing the headset and And you see their avatar.

37:08And I feel like if you look at the motion of the human, the avatar seems stiffer. Like maybe they're not moving all the parts of it. Like maybe the mouth doesn't move quite as much. And when the mouth moves, like the earlobe doesn't move. And you might not think those are connected. But on a person, like when you talk, more than just your lips moves. And that's very difficult to do unless you carefully model or observe that. Which is why I think Apple's personas, when you look at this, Apple's personas start to look a lot more like Memojis. because they, Apple has chosen the three or four or five most important parts that need to move.

37:42And they've probably put in some simple heuristics of saying, look, we can't tell when the earlobe twitches, but basically every time they make this noise at this volume with this mouth sound, move the earlobe a little bit or whatever. And I think they're doing a little bit more of like what animators do, which is just give me the gist of it and use what you know about how people talk to make it convincing. These ones look a little bit more death mask-y to me, but Hey, uh, everyone has their own opinion on these so you should take a look at it as for the question though should apple have high quality persona scanning stations well i think not if they end up looking like this but either way it's not it's not feasible like if they ever end up selling a whole bunch of vision pros you can't have people lining up at the the apple store to get their scans the whole point is the device should do it for you so that they should continue to work the solution the way they have you buy the device you take it home you do the scan the privacy of your own home after you've done your hair and makeup the way you like it tweak it to your heart's content and they should just keep making that better rather than saying you need to come in for a three-hour session in our laser scanning booth i disagree with you with regard to these uh codec avatars i think these look incredible i really really do and and to me the personas which again i did i am a defender of the personas i think that they are a pretty good idea and they work pretty well uh i think the codec avatars look way way better do you think they're on the other side of the uncanny valley or they haven't dipped into it yet do you know the uncanny valley chart i'm talking about i gotta Yeah, yeah, yeah.

39:00I think that the Kodak, or I'm sorry, the Personas are dipping into the Uncanny Valley. I certainly don't think they're at the bottom of the Uncanny Valley by any stretch, but I think they're dipping in. But if you think the Personas look better, then you, like, the Personas aren't even in the Uncanny Valley for you yet, right? Correct. Yep, I would agree with that. I would agree with that retelling of my opinion. I don't want to imply that you feel the same way. This is kind of like a Bezos chart. I just put a link from the Wikipedia page to the Uncanny Valley thing. And I mean, I guess, where is the dip?

39:30Is it like at 90-ish percent? It's very fudgy. Yeah, yeah, yeah. No, to me, I feel like the Personas are starting down the cliff, but they haven't fallen yet. And I feel like this is comfortably on the upswing, if not the top, on the other side of the valley for me. You're going to think, other side? All right, well, maybe you need to play more video games. Yeah, actually that's probably true. That's actually a really, really good point. Cause these look about as good as like, you know, the cut scenes and last of us part two, for example. No, that's a really good point. Actually. And I think you were saying it kind of tongue in cheek, but I, I think you're right because I very rarely play any video games.

40:05I don't know what the state of the art is on this sort of stuff yet. But kind of tangentially related to this recently, I did a podcast with Wojciech. I'm not even going to try to pronounce his last name. I'm so, so sorry, But it was a Polish podcast, which obviously I don't speak Polish, but he speaks phenomenal English. And we did this in the same style as Gruber in Sandwich with Adam Lissagor on the talk show. And we were doing this just with our personas. And I've never met him in person. And so I don't know what he looks like. And to me, his persona looked completely good, like completely great even.

40:43But I think that's in no small part, in fact, entirely because I don't know what he actually looks like. Like I've seen photographs of him, you know, from time to time. I bet his hair is not blurry in real life. Fair. But, you know, but I think you get the point. I'm driving that is when you don't have a relationship, a physical relationship with this person insofar as you've never been in the same space before. I thought the personas were great, actually. It didn't feel unnatural to me at all. Now, that's probably that would be very different if I was talking to you or Marco or Mike or Jason or what have you.

41:13But for an instance where this person who is not a stranger but is sort of kind of a stranger, like I'm underselling our relationship, you get my point, it was very passable to me. And there's a YouTube video, which we'll link in the show notes if you want to see this for yourself. But I thought it was really good. Teach your own. All right. Moving right along, there's a gate now for the Vision Pro, and it's Crack Gate. Some Apple Vision Pro units develop an identical crack in the cover glass, reports Mac rumors. A small number of Apple Vision Pro owners have claimed that their headsets developed a hairline crack right down the middle of the front cover glass, despite never having been dropped or mishandled.

41:51And there's a link in this article that makes it pretty clear what they're talking about. Yeah, I mean, this kind of thing, like, it looks like, you know, kind of a standard, you know, manufacturing defect or weakness with, like, a certain stress point in the glass that probably is, you know, no fault to the people who are using it. again it's unfortunate but it's a version one product of a very very complicated design so this kind of stuff happens they'll fix it it'll be fine in the long run all right there is breaking news this was just today wasn't it as we record uh apple has officially canceled project titan ending a decade long and allegedly 10 billion billion dollar efforts uh this is reported by Bloomberg and also the New York Times.

42:37We'll put links in the show notes. From Bloomberg, Apple Inc. is canceling a decade-long effort to build an electric car, according to people with knowledge of the matter, abandoning one of the most ambitious projects in the history of the company. Apple made the disclosure internally Tuesday, surprising the nearly 2 ,000 employees working on the project. Wasn't really that much of a surprise. Yeah. Fair, fair. The decision was shared by Chief Operating Officer Jeff Williams and Kevin Lynch, a vice president in charge of the effort. The two executives told staffers that the project will begin winding down and that many employees on the team working on the car known as the special projects group or spg will be shifted to the artificial intelligence division under executive john john gianandria something like that i'm sorry uh those employees will focus on generative ai projects an increasingly key priority for the company this is an interesting confluence of events here because this you know the fact that they're taking some employees i mean first some employees are going to be leaving the people who knew how to build cars i'm not sure apple needs those people but shifting the other employees to ai projects made me think of exactly you know i was even thinking of blogging this but didn't have time because it just came out today how i would sum up uh this you know 10-year foray into the car project as we always called it project titan um for apple uh and it it's related to ai um when i was a kid i read a lot about computers and stuff and i used to read a lot about ai because who wouldn't it's cool you watch 2001 you know you're learning about computers you want to learn about computers that are like hell 9000 but don't kill people spoiler alert yeah seriously i've never seen it sorry uh when are when are computers going to be smart and talk like people that'll be so cool it's all in all the science fiction let me learn about where computers are today and let me tell you when i was a kid in the 70s uh you'd see uh stories on pbs and books that i'd check out of the library and stuff that would they They would say, like, look, here's this computer they made, and here's what it can do.

44:30And people were super excited. Like, look what we've made this do. I know it doesn't seem that smart to you, but I can talk to it and ask it questions, and it understands the sentence, and it knows what the different parts of the sentence mean, and it can reason about things and make inferences and respond to me. And people were so excited about that. And they'd say, and you know what? Look at computer chips. It's 1974, but in 1975, computer chips are going to be even better. In the 1976, they're going to be even better. And if you extrapolate on this graph, you can see that in the year 2000, computer chips are just going to be massively more powerful than they are now.

45:04And then you look at this AI thing and look what we've done now. And it's kind of dumb now, but we're on the right track, right? And so if we just follow this graph, by the year 1988, computers should be smarter than any human on the planet. And this is what I read about when I was a kid. And when I got a little bit older, you kind of realize that since the 60s or the 50s or whenever they first tried to do this, everybody thought how 9000 is going to be like, you know, at most 20 years in the future. Because just look at this graph, right? I mean, I didn't know the term Moore's Law then or whatever.

45:33But I was experiencing the advance of technology, which was my computers were getting faster and faster all the time. And the video games I played in the 70s were so much worse than the video games I played in the 80s and 90s or whatever. So I was living the progression. and yet artificial intelligence before the current AI trend, that is in a computer that thinks and talks like a person, like, again, Hound 9000 but not murderous, or any other things you see in sci-fi, very smart people really believed that, you know, although what we have now is not great, in five to ten years it will be better, and in 20 years it will be the most intelligent thing on the planet Earth.

46:12And as part of my, you know, I talked about this in Rectus ages ago, discovering the parameters of the world. Can you move things with your mind? How big is the universe? Are we going to make computers that think? One of the parameters of the world was people always kind of think break-even fusion and artificial intelligence are much closer than they actually are. And as excited as I was about it, as the years went by, and you saw the same stories, and everyone said it's probably five to ten years in the future, you figured out that these people hadn't yet found the right approach to this problem.

46:45We found the right approach to flying, which is like fixed wing aircraft with engines. We found the right approach to doing math because we're getting really good at that with our computer chips. That approach, we just like that. That's the way just start scaling it, you know, and that's served us well up until today. But we haven't yet found the right approach to computers that can think. And so with the car project, Apple, like like all those excited scientists in the 70s that I was reading about and like so many other companies, in the world decided that cars that drive themselves are probably right around the corner.

47:22Look what we can do now. It's pretty impressive. And if we just extrapolate and say, if we just continue along this path and our computing resources get better, like in, you know, five to 10 years, all of our cars are going to have no steering wheels and they're going to be driving ourselves because look at our progress and look how, look, look what we're already able to do in such a short time surely by insert year that has already passed here all cars will be self-driving and driving will be obsolete uh and that's why apple embarked on this project because if that's going to be a thing that is a big disruption to the transportation world and apple should probably be in on that and they were focusing on making a drive that making a car that was autonomous in some way and if you're wondering why apple thinks they should make a car It's not because, you know, they just wanted to make something in an industry like in a low margin industry they wanted to get a part of.

48:12They're like, well, this this industry is about to be disrupted. In fact, we'll put a link in the show notes to Tim Cook, an interview in 2017 where the interviewer is asking him essentially about the car project or about automotive stuff. And Tim Cook flat out says, I think there's a disruption coming and it's because of autonomy. And, you know, Apple wants to be in on that and whatever vague way he said it to not, you know, mention unreleased products. a lot of people thought that uber thought that you know we we hire drivers now but kind of like netflix sent people dvds for a while uh that's just how we're going to start eventually we're not going to need the drivers the cars will drive themselves so many companies looked at what they had and just like the ai people said this is amazing and in five years it will be so much better turns out they did not have the right approach no one no one has the right approach yet no one knows how to make a car that can drive itself as well as humans can drive in all conditions uh it's not like it hasn't gotten better it's gotten way way better waymo has cars that can totally drive themselves in limited circumstances in limited areas doing a pretty good job but even waymo is approaching the limits of the techniques that it has been using so So Apple failing on this project is just a consequence of us collectively as humans having not yet found the correct approach to make a car that drives itself.

49:34We can make a car that helps you drive a little bit better and the jury's still out on whether it helps you or not based on, you know, you actually paying attention or not. But we haven't found the right approach where it's like, if we just do this for five more years, we'll be there. Unlike, for example, you know, adding numbers together in your CPU. We found the right approach for that, right? Or planes or whatever. And that's disappointing for everybody involved. It's if you're wondering why Apple spent$10 billion in 10 years doing this, important people at Apple like Tim Cook, or if you read the New York Times story, Johnny Ive had convinced themselves or had been convinced by the hype or the other people they respect in the industry that this was a thing they thought they could do.

50:18like they said look look at look at what tesla's for example been able to do now we're smarter than them and they're already almost there let's get on this maybe we can beat them to it because surely one of us is going to crack this problem in the next five years and the answer was no no one's going to cry in the next five years no one's going to crack it in the next 10 years when are we going to crack it i don't know is is the current approach the right one seems like not you could but maybe it is and we just have to do it for 20 more years but like talk to the ai people everyone thinks what we've done is amazing what we've achieved is amazing extrapolate and we'll get there and this is yet another disappointing instance in my adult life of lots large portions of smart people in the world thinking that a thing was right around the corner and it wasn't which is weird because most of the time if you listen to a tech podcast or you're interested in tech most of the time tech does sort of fulfill our dreams even more so like computers really did get way better video games got so much better like so much if you if you could take like child me and show them like to mention again the last of us part two it would have exploded my brain right that really did happen just like everyone thought it do in fact maybe even faster the internet the internet look at that happened so much faster than people thought it wasn't people underestimated that but sometimes if you think something is around the corner sometimes you're going to be wrong because it turns out you're barking up the wrong tree and i think all of humanity so far has been barking up the wrong tree and getting cars that can drive themselves as well as humans in all conditions and that's sad for uh the little versions of me who are seven years old checking books out of the library about self-driving cars now and they're gonna have to learn they guess they just didn't figure it out doesn't mean we're never going to figure it out we might eventually but apple let's say apple is taking a break for a few decades before they another run at this because if it does become plausible they should probably dive back in and participate in this industry but right now i think even waymo had a big thing where they were like yeah we're kind of giving up on the whole self-driving everywhere thing and we're just concentrating on limited circumstances everyone who has tried this has as they say in dune they tried and failed they tried and died well they didn't die but you know the projects died so I think this is unsurprising, but disappointing for little kids everywhere.

52:32I think the car project, we've been saying for a long time now that we didn't really understand why they were doing this. Obviously, it's very hard to judge with something this secretive. We don't really know how much progress they were making, but it sure has seemed like over time, this project has not been going well. We've heard of a number of reboots and reorganizations and changing directions. There's been so much change in the project trying to I guess get it to go in a good direction. Well, on that though, in the 2017 video, Tim Cook flat out says, we think there's going to be disruption in the automotive industry and it's because of self-driving and automation essentially, right?

53:14But then the most recent reboot that we heard about and talked about on the show was Apple said, yeah, forget about self-driving. We're just going to do like we're still going to do a car, but it'll just be like you know current sort of driver aid type stuff and that is not the that was not the vision according to tim cook in 2017 it's not like oh we just want to make a car because we want to make a car he thought there was going to be disruption related to automation and that basically turned out to not be the case in the time frame that they cared about and so changing the project like oh let's reboot it but now we're just going to make a plain old car that's like everyone else's car It's not an exciting project, and I'm very glad that Apple chose not to do it.

53:52Even if they had gotten the self-driving part to progress further and that was still the plan for the project, even then I still don't think Apple is the right company to do that for lots of reasons. It doesn't really fit a lot of their business shape and style and principles. We discussed recently how it seemed to be almost impossible to fit that into their new carbon goals, for instance. Like, you know, their goal with carbon neutrality is to have their products be carbon neutral end to end over their entire lifetime, including the energy they use during their useful lives. That's really hard for a car.

54:31Well, the rumor is it was going to be like 100 grand. So maybe they'd sell so few of them that they could carbon offset every single one they sold because the volumes would be low. But yeah, I do really wonder, I mean, maybe when they made that commitment, they kind of knew that the car project was going to get canceled. because you either don't sell a lot of cars and do a lot of carbon offsets or you can't meet your goals because they can't transform the world's energy infrastructure to be carbon neutral to power all their cars. No, and so that I feel like the carbon plan was always at direct odds with the car project.

55:04And I don't know how they could have possibly ever resolved that. And the other thing is like, we've heard through a few different channels about Tim Cook really believing very strongly, allegedly even saying directly apple products don't kill people period like apple does not make products that kill people at least you know directly obviously you can text in your phone and crash but they feel very strongly about that and i think that's that's a very good thing both for the world and of course on a smaller level for their brand and it's a luxury it's a luxury for their because they're a company that currently doesn't sell things that can directly kill you right and so they're getting they're thinking of getting into a business where you can't avoid that cars people die in car accidents right so i don't know how he would square that stance or that's just like a soundbite really but like the idea is we don't we don't want to be a company that does that but we also want to get into the car business well i have some bad news for you about cars there's no such thing as a car that doesn't sometimes kill people due to car accidents at least not yet maybe they thought they could make one again with the self-driving thing but you know that's that's kind of like you're making that choice like you could just keep making phones and computers and they're not going to like crash into you and kill you right but you're choosing to try for a car and that's you know that's changing things up again we've we've been wondering for years why this project was still going on it seemed like it was a pretty long shot from the start and it seemed like it was never going super well so i'm glad to hear that they finally axed it you know obviously for all the tumult this is going to cause the actual people working on it, including probably a bunch of lost jobs.

56:38That's not great. But for the company strategically and for the leadership to have chosen a direction, I think they made the right call. I kind of don't know why they didn't make it years ago, but I'm sure they had their reasons, whatever the reasons were. Now they've made the right call. This project is seemingly really finally dead. And I think it's better that they don't continue to waste all of this experimental money and opportunity cost on a project that's very unlikely to ever be very good for them because now they have freed up that capacity. Maybe not the exact people because it's different specialties, but they've at least freed up like that organizational capacity.

57:15That money. How about that? Sure. You know, the money is a big part of it, of course, but also just, you know, focus, talent, you know, like there's other limited resources they have. Now they can focus that on products and, you know, improvements to their other products that are much more likely to actually benefit them and their customers. The car project was always kind of this massive moonshot. It was very, very expensive. And so if they're now redirecting that towards things like better Siri, or whatever we're going to call all their new AI efforts, that's way more likely to both A, succeed, and B, to actually benefit their customers in useful ways and to benefit the rest of their products together.

57:58like we see what apple does they they share a lot of tech they share a lot of advancements they share a lot of software between all the different platforms they share parts they share techniques that like they share manufacturing you know strategies and things between all the different platforms they share supplies and suppliers with the car it was always kind of like does this have any chance of benefiting their other products not much you know how would it even fit into their retail strategy who knows i mean it's not even clear from the outside how much car play was shared with project it seemed like not at all um so like except maybe like some dashboard design concept yeah maybe some underlying stuff like it's it's so hard to tell and on this topic by the way not to get back into vision pro but you know this is something that's come up with all the demos like when you see obviously ar kit and stuff we saw all that coming out ios and mac os it's like oh that's going to be for the headset right and yeah yes there's a connection there right but now that uh vision pro is out and you see universal control it's easy to just retcon it and say, I bet they did universal control for the headset and it just came out earlier on Mac and iOS.

59:02That could be true or it could be the reverse. They did it on Mac and iOS and the headset team adopted it. But that's the nature of things inside Apple. They will share things between their projects when it makes sense to get the most value out of it. And from the outside, it's very difficult to tell. Did, for example, universal control start with the headset team and trickle down to iOS and mac os or was it the reverse it doesn't really matter either way they're getting the most bang for the buck out of that technology they developed they developed universal control and it has broad applicability across their products and so now you look at the car project what things on in mac os ios ipad os or whatever have broad applicability to the car and vice versa and you know you go for carplay it's like well carplay that's got to be a thing but because we know so little about the car and because CarPlay is such an odd duck, it's like maybe not even that was shared.

59:54So the best you can come up with is maybe whatever real-time operating system and tooling and stuff they came up with for the car they'll find a use for. But it's so hard to tell from the outside and the synergies are not quite as obvious as like universal control or ARKit. Yeah, and I think there's probably just fewer of them in general. And that's why I think axing this project is the right move strategically for the company because now whatever the special projects group can work on can be things that the company might have actual other use for in much larger scales. Because we see with Apple, they historically were very bad at multitasking.

1:00:31They've gotten a lot better in recent years. But still, you don't really want them, as a fan of their products and as a customer of their products, you don't really want them to have massive distraction projects that don't benefit the products you use. with Vision Pro, we actually see that there is a decent amount of, you know, shared innovation between Vision Pro and their other products, probably that we as customers were able to benefit from. With the car, it seems like whatever we would benefit from that would be much more limited with the other products. That was kind of an advantage, though, because you felt like, well, they hired a lot of new car people to work on the car.

1:01:05So those people weren't even at Apple to begin with. So you're not stealing that many people from other teams. And because there's so little crossover, I think for the 10 years that this project ran, It was a money pit and maybe it did suck in some important talent, but it was less of a drain than, for example, iOS and the iPhone were or where jobs came out on stage and said, yeah, you're not getting a new version of macOS because we put all those people on the phone because it's more important. Sorry. I'm sure that did happen, but to a lesser extent, just because so many more people had to be hired from the car industry.

1:01:36So that was that was one blessing. I think that's part of the reason this project lasted so long, like the open secret that they were doing this and didn't, you know, it didn't sideline the whole company for a decade. Right. The company continued to do what it was doing. Yes, there was some impact, but like the iPhone shoved the whole company off of its previous track and eventually it found a new track. Project Titan didn't do that. Project Titan was just just like a$10 billion quiet little leech off in the corner. But Apple continued to be able to function in its main business. I mean, and it's a very different company, you know, in this in that decade than it was before that as well.

1:02:11So obviously, like, that's, that's a different scenario. But I do think it's interesting, though, to consider, like, for years, we had heard about these two giant Skunk Works projects, the car and the headset. Well, now the car is dead, and we have the headset. What do you think is the next big Apple Skunk Works project that we that isn't out yet? Like, I guess, you know, AI, in general, could be something like that. and I'm sure we're going to see more of that as the rumors indicate probably this summer. And honestly, I'm very much looking forward to whatever this ends up being. But do we know of any other large rumored Skunk Works projects that seem at all plausible that Apple has been working on?

1:02:51I think this is the time maybe to talk about this, which I think will very clearly probably quickly not happen and we don't have to think about it. But because of this announcement, people who really want Apple to make a car have been saying, you know, Apple, if you've decided that disruption in the automotive market is not imminent due to self-driving, which I think is the correct conclusion. But you still want to be in the automotive market because you think people in your team know how to make cool dashboard things like the new CarPlay or you just want to you want a new platform that you can own.

1:03:25Like you want to own the infotainment stack or whatever. there's still a way for apple to be in the car business and part of the thing when we've talked about the car project in the past we're like apple's good at making electronic gadgets and if you squint you can call a car an electronic gadget but honestly there's a whole bunch of parts of that that are not electronic gadgetry like suspension and brakes and apple just doesn't really make anything yeah that's that's kind of like saying like my oven is an electric gadget because it has a control panel on the front of it like yeah things that you sit inside that travel like the whole the the skill set of how to do that and that's before you even get to stuff like the batteries which i know apple has some uh knowledge about from all of its gadgets but it's a different scale entirely there and electrical motors which you know there are there are some in the vision pro i guess but not apple's area of expertise for sure uh but if apple still wants to be in the electric car business.

1:04:20They could buy lucid and rivian and have some cars suvs and trucks that would you know suddenly sport carplay yay uh and they would have you know instant expertise lucid has the best powertrains in the entire industry and the best ev packaging in the entire industry oh we're gonna get so much email please don't do that john you're just tempting them now i this is this is not a questionable i think is 100 defensible oh god you're making just based on the numbers how big is the engine how much power does it produce how efficient is it lucid has the best electric powertrains in the industry they might not have the best batteries but their batteries are pretty good rivian is really the only game in town for sort of non-mainstream fancy apple style pickup trucks and truck like suvs because you know the mainstream makers make them but rivian is the froofy kind of apple-ish brand and the important thing about apple buying both of those one it will probably cost less money than they already spent in project titan two everything those companies do is stuff that apple as far as i know does not know how to do like all the rumors of like they're going to build their own car like oh here's who they're going to build it for them or whatever i'm like do you think apple's gonna get the kind like the people who who started uh lucid were all like the lead engineer for the tesla model s started lucid like they they cut their teeth at tesla and then they came over and did lucid apple has done none of those things so they think out of the out of the gate they're going to come up with a an electric car that they you know solve all these manufacturing problems before everyone else has such a head start on them i know they said the same thing about the cell phones you know the pc people aren't going to walk in but let me tell you a cell phone is much closer to a personal computer than a car is it's a big gap there right so if they wanted to be in this business because they think i mean i don't know they i think they shouldn't because it's low margin business and apple products don't kill people and yada yada if they're not going to be self-driving there's no real point but if they really want to a much faster better way to spend that money is to save these two companies that are currently having some difficulties harvest the the all the people and technology that they have and voila now you have a line of apple cars if some if tim cook decides he's the ceo and if he and the board think that apple should be selling evs that is i think the only reasonable way to do it in a reasonable time frame if not again let's just revisit you know making a car in 20 to 50 years and maybe the self-driving thing will have uh will be more plausible then but yeah like i'm i'm with marco that i'm relieved that they're not doing this because once they're not making a self-driving thing which has been obvious that they're not going to do for a while now to everyone except for apparently apple um just making a car is not something i have any faith that apple would be any good at at all like nothing they've done makes me think boy if only apple would make a car they'd be great at it nope absolutely not and making car is ask rivian ask ask tesla ask lucid making cars is very very difficult like it's not something that you get right on the first try or the second try or the third try car companies have been around for a long time they've had time to work out a lot of this stuff and even they get stuck in their ways which is why people can innovate around them but like just the first the first several years of Tesla, the first several years of Rivian and Lucid, manufacturing big, important things like that, complying with regulations, making them safe, reliable, like just, it's really hard.

1:07:53And Apple has never done anything like that at that scale. And it is so different than cell phones and computers that I don't think they could pull it off. So if they really, really want to do this, make some acquisitions by, by the, the struggle. They tried to buy Tesla. You know, that's, you can read the story. They tried to buy Tesla way back when, so it's clear that they were open to that idea uh but if the if the car dream is still alive at apple uh they should buy lucid and rivian i hope the car dream is not alive at apple and they should just ignore this for another 50 years well i think if anything first of all i think you're right that would be probably the best way for them to get into it at this point but if anything that shows why they shouldn't get into the car business because suppose tomorrow apple buys lucid and rivian okay now what like why why did they do that and then how do they how do they integrate that into the rest of their business well it's still a massive question of like now apple's a car company great now why do they want to be a car company and now now they have to deal with being a car company right but they already had that would give them two car companies that are already car companies and they're car companies with problems and you know there there's a lot of growing pains but they they jump start it like they don't have to worry like will we make a car that people want will we make a good car those two companies have already made cars that people want that are good they're just not making money making them so apple could you know apple has the runway to has a little bit more runway than those two companies do to figure out a way to start making them profitable and obviously they would integrate with them and i think jeff johnson had a good snarky tweet about this whole thing which is snarky uh but also i think there's kind of a point in that why apple might want to be in this historically low margin business uh he says i bemoaning the uh the the end of the car project but i so wanted a car that demanded 30 of every shopping trip and refused to travel to destinations unapproved by the manufacturer got him but like that's the thing that's kind of the reason apple was interested in in this thing is like oh the automotive market is being disrupted and if the automotive market is being disrupted that is an opportunity for us to find another place for one of our platforms and modern apple when they find a place for one of their platforms they're like we we need to have a platform which we will then leverage to extract you know 30 from everybody right that's their mo um hasn't been going that great but they got they hit it out of the park once with the iphone uh and so they've used that model for every other new platform they've tried to roll out apple tv vision pro ipad i know those all kind of look like you squint and like oh that's just an extension of the iphone right but the inside of the car the inside of of any sort of inside of cars are becoming more looking more and more like a platform that apple could participate in the problem is if apple doesn't isn't literally a car company it might be difficult to pull a microsoft and become the plat the software platform for everybody else's hardware apple's trying um but there's a little bit of pushback contention there so why would apple want to be in the car business at all i think cars will continue to be low margin but i think there is a potential future to you know using the terminology extracting rents from the plat the software platform that lives inside cars certainly every car manufacturer wants to do that just ask gm that's why they don't want carplay in the car they want to be the one you know skimming off the top of everything that happens inside the car that they can i'm not sure there's a business plan there apparently it's not renting heated seats to people like BMW tried.

1:11:23But I think that's, in the absence of self-driving disruption, another place for a platform is, I think, what would still have a chance to attract Apple to cars. But who knows? Maybe the CarPlay project is that and they don't need to make a car. It's too early to tell. I'm going back to what you were saying about Lucid and Rivian, and the co-host of Neutral in me says, Heck yeah. That sounds amazing. Because I think Rivian is arguably more spiritually similar to Apple than Tesla. I think that Lucid, to your point, has some of the best and most advanced technology. Please don't email us Tesla people.

1:12:03We don't care. Um, so I, I, I agree with you as a car enthusiast, as someone who at least tries to watch and understand Apple, I can't imagine them saying about a business, as you've noted is very low, low margin, which they can't be too keen on to begin with. I can't imagine they're going to say, let's spend a sum total of roughly$20 billion. Cause that's the sum total of the market cap of the two companies today. let's spend$20 billion on two companies that are barely keeping themselves afloat. Like what? You wait a little bit longer. Lucid will get cheaper, I think. Okay, sure. And so might Rivian.

1:12:40But then, I mean, so, so we're buying these beleaguered, you know, failing companies. Yeah, we're getting them on the cheap, but if they were good, they wouldn't be beleaguered and cheap, you know? No, that's not true at all. Like their problem is they don't like their, their problem is they need to get through like the, the scary growth period. Like the Tesla barely made it through. like Tesla almost went out of business several times too right but there it takes so much capital to sort of get to get the ball rolling to work out all the kinks and both of them are kind of on the bubble like they're probably gonna make it but Tesla just like barely made it due to like some savvy fundraising and government mooching from Elon Musk at the right time right like it's it's a tough gig in the car business so I don't think they're beleaguered because they have bad products uh they have they have great products uh they're just beleaguered because making a car costs a lot of money and they're not entirely sure that they'll get past the point where like they're just plowing and plowing in money and then they're selling a smallish number of cars for a very low or negative margin and it's like just kind of like amazon's like just give us another decade to not make any profit but eventually we will and amazon did eventually and they had that runway uh rivian and lucid remains to be seen if they will be able to find the funding to give himself that runway but you know who can give him that funding apple i mean that's true but i just don't think apple would have that kind of patience especially after having been burned by being 10 billion dollars in having not a lot to show for it like i just well they did have the patience for that 10-year boondoggle project they went well so i mean that's true i don't know it doesn't ultimately matter but it is an interesting thought exercise of nothing else yeah i mean honestly i feel like they'll apple will not do this because they should they're just i think they're done with cars for a while right but yeah if if someone inside apple just demands it like just like no we need to be in cars which i don't think is true because again if tim cook's whole reason was that there was it was about self-driving disruption i think he's now dissuaded of that and he's fine but if someone demanded it that would be the only feasible way to do it because if they didn't and they tried to make a car on their own i think it'd be embarrassing for them because their car would they wouldn't have to worry about losing money because they got so much money but they'd make a car and that car would have to compete with beleaguered rivian lucid and then not beleaguered tesla right and i don't think it would fare that well because i don't think apple is good at almost any of the things you need to be good at to make a good car so yeah i just you know i just i think they should leave it alone but if they don't want to leave it alone they should buy a reviania tesla and heck even if they don't want to be in the car business at all apple should just you know give an investment to lucid and reap it when the company uh comes roaring back uh two decades from now yeah i don't know we'll see where it goes i i think i am certainly in support of of apple letting go and stop trying to make fetch happen.

1:15:23I echo what I think it was Marco who had said, I feel terrible for all these people that are going to now have to find either new jobs than Apple or new jobs entirely. That stinks. But I think on the whole, it's better for Apple to go this way. I think it might be interesting seeing people who surely are very good at computer vision and artificial intelligence and things of that nature working on other projects and problems inside of Apple. I think, I suspect a lot has been learned from this. And I feel like I made the same, I shared the same theory a while back and I feel like somebody debunked it.

1:16:00But here I am again thinking that, you know, I wonder if the real-time OS or some of the things that they learned building a real-time OS for the car were used in the R1 for the Vision Pro. And I don't know, maybe, maybe not. Yeah, that was debunked. But like, I think you'd have a good point with the computer vision aspect because Apple does now have a platform that could benefit from computer vision expertise. And so those people on the car project who are probably super isolated from the headset project, because why wouldn't they be? Bring those people to Vision Pro now. Because that is expert.

1:16:29Even if it's not technology that they're currently using that crosses over, that expertise for sure crosses over. Well, and now also, with this talent and budget and executive focus freed up from this project, if you think about what else could they do, even without having good uses of shared technology with their other products, they could do something like pour more money into the idea of a glucose monitor for the apple watch yeah i think they're already doing that but yeah freeing up the money is good for everything across the board of course if you told me what they should give the money to it's just fixing bugs in mac os but i don't think that's probably going to be high on their list but you know just just take a little bit of that 10 billion or you know whatever it is 1 billion a year that you just freed up but a little slice that into mac os bug fixing please i don't think it would even take that much yeah in terms of what the next big thing is like it's you know we said this in the our first episode in january 2024 we did like what do we expect this year to be for apple uh and i think one of the things that we got right based on all the stories that are coming now uh 2024 the year when apple sprinkles ai sauce on a bunch of stuff uh i talked in a recent rec tips episode about this it's going to come out soon what are they going to sprinkle it on what will actually ship this year i don't know but there's going to be ai sauce and it's going to be sprinkled uh and that is i don't think that's a you know to marco's question what's the next big thing i don't think that's it because kind of like self-driving cars at this point i would say that um the like large language model ai stuff has proved its utility kind of like you know smart lane holding better cruise control things have proved their utility in cars but i also think that we're pretty close to well i think the jury's still out but i i would put money on the fact that the current approach to LLMs, you can't just do more of that faster and get to Hel9000.

1:18:18Maybe it's a piece of the puzzle. Maybe it's totally off in the wrong direction. But I don't think the next big thing equivalent to the car project is Apple should make a computer that thinks because nobody knows how to do that. And I don't think the current approaches are going to get us there. But all that LLM stuff has proven utility. You can do useful things with it if you use it the right way. And Apple needs to use that to make its stuff better in all the ways that everyone else is making their stuff better. Make Siri better. Make image search better. Like we know, like there's no question it does that.

1:18:53Forget about HAL 9000. You can use this technology, Apple, to make your current products better. And that is not the car project, the headset or whatever. But I think getting a bunch of people put into that part of the organization is a good thing. So like rather than, you know the car project and the headset are both kind of like pie in the sky we have a great idea we think this is the future we're going to try to do it one of them they shipped one of them they didn't but they're big pie in the sky things this is not pie in the sky this is keeping up with the joneses this is hey have you noticed what everyone else is doing apple you should do that too because you got a lot of stuff that could benefit from things that other people are proving the utility of so that i think is going to happen some of it this year hopefully more of it next year and if that's all that came out of this like that apple never came up with the next new big thing i think it'll be fine partially because i think you know oh we got the headset but we're still waiting for the glasses right like you can extrapolate from vision pro and say where is this going to go and unlike self-driving and hal 9000 you can see a fairly straightforward path of like just keep iterating on the vision pro with increases in technology that we expect to happen and you know screens get better cpus get better like just we can see that progressing we don't quite know how to get all ready glasses because some of the display stuff is questionable at this point but even if it's just vision pro but better than the current vision pro there is a path to travel there and i think that is a big enough project a brand new platform that is very much more ambitious or different than competing platforms and the other things that it's done.

1:20:32There's 10 years of runway right there anyway. So I don't think Apple should be out there saying we need to start something new. And that's setting aside all the health stuff that we know they're already doing. I don't think it should be like hovercrafts, like, or we're going to start an airline or like just chill out. You got just, can we just do the Vision Pro for 10 years and see if that works out? Because there is a runway there. You have so many other platforms that you're working on. You're still trying to do all the health stuff. You got all the services. you got all the antitrust things you got the app store like i don't think apple should be looking to you know have a rebound uh project after project titan hey me again we are still sponsored exclusively by atp members please consider joining and becoming a member today you will get a version of the show that does not have these membership ads it's stuck in the middle of it uh you will also get the bootleg feed with the unedited live stream as mentioned earlier and you get the bonus episodes also, as mentioned earlier, that are exclusive member content for us doing all sorts of tech stuff, non-tech stuff, even some feeling stuff here and there.

1:21:34It's a lot of fun. We also have a couple of those small bonus things as well. So for instance, when we do our time-limited big sales of new t-shirts and stuff in the ATP store, you get 15 % off that. If you can support us, this is by far the best way to support the show. So we would really appreciate it if you consider it atp.fm slash join and it's eight bucks a month or 88 bucks a year or a couple other different currency options here and there again we support gift memberships now thanks to john's hard work on that i didn't do any of it so john gets full credit for the gift membership system but it's really cool you can do that as well so check it out today atp.fm slash join once again ad-free version unedited raw bootleg feed bonus episodes the discount during our sales so eight bucks a month atp.fm slash join it is the best way to support the show thank you so much for your consideration and let's go back to the show joe writes what email clients do you all use these days curious to know how each of you handle this on both ios and mac os and what you feel are the best clients i feel like this is one of those as john often describes annual or sometimes semi-annual questions that people always want to know what email clients we use and i don't know that it's changed, like properly changed for any of us since we started recording the show in 2013, 2014.

1:22:54Creatures of habit. Yeah. But for me, it's Mail App. I was briefly using whatever that one is. Is it Mail Plane that everyone loves? Something that everyone loves for Gmail. And I was using that for a while. And then when I switched to Fastmail, I put that aside and went crawling back to Mail.app. And that's what I use everywhere. And I thought that was the same for Marco, isn't it? I have actually really never gone for many other mail clients. Sorry, I just meant that you're all in on mail.app. Yeah, I'm all in on mail.app, and I am mostly fine with it. I wouldn't say I love it, but I don't love email.

1:23:30And so I treat email as a very kind of functional thing that I have to deal with somewhat reluctantly, and I deal with it as little as possible. Apple's mail apps let me do that on their platforms with lots of various nice little integrations. I appreciate all of that first-partyness of them. My problems with email have nothing to do with the client. None of the email apps that are out there would do a meaningfully better job at making me hate email less. I just kind of deal with it the way I do. The only thing that I... I literally just turned into this earlier today. my one big feature request for Apple's mail apps, having used them now for how many years?

1:24:12When the iPhone first shipped, and the iPhone had its version of mail.app on it, it mostly worked, it mostly satisfied my needs, except for one giant area, search. Mail search on iOS has always been really rudimentary and really crappy, and many times I end up having to go search on my Mac mail app instead and it will find things the iOS app will not find, including this literally just happened today. I assume the reason why is because mail on the Mac downloads everything off the IMAP server and indexes everything locally, whereas mail on the iPhone appears not to do that. I think mail on the iPhone has always done server-side IMAP search, not local search.

1:24:56I assume that's why we're getting different search results. But that trade-off made sense in 2007. That trade-off does not make sense in 2024. I understand that people's email boxes can be really big. So maybe make it an option. Give me a switch in the settings app that says download all mail and let me search all mail in the iOS mail app with the exact same search characteristics and results that I would get if I did it on my Mac. It's 2024. I'm pretty sure Apple can make that happen. John, you still using the Gmail web client? yeah the main sort of performance characteristics that i'm looking for in an email client uh date all the way back to you know my the early days my very first email clients on the mac i was super heavy into entourage and uh later when that uh morphed or clarice emailer which morphed into entourage um and they what i loved on those things is they had like rules that you could apply to mail that would ferry your mail into various folders or whatever but it was so frustrating when i'd have all those rules set up on my home computer and then i'd like try to check my home email while i was at work and so i would like install the same email client on my work computer but then the rules would be on my home computer and this was before the days of ubiquitous cloud sync and it was never quite the same and a lot of times i was using pop so if you popped it from one location if you popped it from a second location the second location wouldn't have it because it didn't pop it so now you have the same message appearing in two different places they had to be routed the same We should probably explain that just a little bit more because everything you said was accurate, but it sounds bananas to people of today.

1:26:30So most email clients or most email servers, excuse me, from, I don't know, 15, 20, 25 years ago, the way it worked was you used Post Office Protocol or POP. And generally speaking, granted, you could tweak this, but generally speaking, what would happen is your email client would go to the server. It would say, OK, what are the new messages? It would grab the new messages and then it would delete them off the server. So like John was saying, let's say you have a machine at home that is checking your email. Generally speaking, the machine at home would check your email, delete the email, and then it would be living or alive.

1:27:01Those emails would only be living on that one computer. If you went to any other computer, because you wouldn't go to a cell phone at this point, if you went to any other computer and went to get mail, well, tough noogies, it ain't there because it was deleted off the server. And this was normal. But I was smart enough not to do that, but that introduces a second problem, which is if you say, okay, don't do that. leave it on the server now my home computer would run run a little you know using pop it would check for mail it would see there's a new message you would download it right now my work computer would run two seconds later and it would see that same new message and it would download it which is why both computers needed to have the same set of rules because as far as they're concerned it's like the home computer downloads the message and files it away the work computer downloads that same message now these two computers are downloading the same message you know and you know pulling it from the server, they better route it to the same place.

1:27:50Otherwise, home and work will slowly get out of sync. Like they're not even connected at all, but it's like a one way process. But if they don't have the same rules, as things flow in, they won't go in the same boxes. And by the way, if I respond to something at work or market it read read at work, home has no idea that that happened because their only function is using the pop protocol protocol to see if there's new messages and downloading them. There's no reverse direction where when I mark something as read at work it sends a message up to the pop server so that when i go home no no that doesn't happen it's one way right so trying to deal with that just home and work being able to check my my personal email at work which is as all working people know is an essential part of your sanity is being able to do that was made very difficult and i and i do like rules and i do like routing and i do like everything to be the same everywhere right and imap you say well imap solves this well imap especially in the early days was a little bit of a nightmare it was trying to solve this problem by having two-way communication and everything but the clients for it weren't great and there were lots of different imap servers and depending on your isp and where you email from it was very confusing and annoying but imap was an improvement but still kind of janky but anyway i was using like clarice emailer and eudora not eudora i know eudora is great but i wasn't ever a big fan uh entourage and clarice emailer back in the day when they did support both pop and imap but then gmail came along and gmail said forget about your stupid client the server is the source of truth i have all your email that your only email only exists in one place folders rules those are all here on the server oh you can look at what's on the server however you want say in a web browser try chrome it's nice um but don't be fooled gmail google has all your email and all your rules and all your settings and all your preferences and when you mark something as read or unread or reply to it or you know your sentiment send mail into a sent mail folder that is all happening on google servers and what it did was gave me blessed freedom to say wherever i check my email from a phone an ipad an internet cyber cafe anywhere i go where i don't care whose computer it is i don't care if it's at work at home on any computer that i have when i load up gmail.com my email looks exactly like it did the last time i left it if I marked a thing as red if I replied um by the way search Google's really good at that I can search my email really easily and it also in the back of the day lets you import all your old email so this is a long-winded way to say Gmail I use the Gmail Gmail for my email and I use the Gmail web interface I like the Gmail web interface I miss you know the days of Entourage and Clary's emailer but i've moved on since then i don't like apple mail i don't like the interface on any platform so i use the gmail web interface on my phone i use the google's gmail app which is not very good but it doesn't really matter because all my mail is there you know i can do what i want in there if i don't like it i just put the phone down pick up literally any other computer log into gmail and there's all my mail big proponent of gmail it's been one of my favorite products over the years and I'll be very sad when Google eventually cancels it.

1:30:59I don't think they would cancel that, but I definitely take your point. No, I think I'll be dead or retired by the time they do that. Too much valuable data goes through, you know, people keep Gmail around. Warren Bacay writes, you recently mentioned Marco's love of the Microsoft ergonomic keyboard, which I've always liked. Why hasn't Marco constructed a do-it-yourself mechanical keyboard in the ergonomic layout he prefers with the switches he wants, et cetera? This seems like exactly the kind of project he would love, and he'd end up with a keyboard customized to his personal specifications. You know, I get this, and obviously my name is not Marco, but I mean, at least for me, I actually like Apple keyboards, and in your case, you like the Microsoft Ergonomic keyboard.

1:31:38And I don't know, as I get older and more crotchety, I feel like there's less space in my life for fiddly things. And I'm assuming that's probably where you are too. I'm glad that that whole world of keyboards exists. If I ever want to dive into it, I will have no hesitation of diving into it. But right now, every few years when my Sculpt flakes out, I can spend$60 and have that problem just disappear. and the reality is i like the sculpt ergonomic keyboard there's not a lot about it that i would change it solves my needs very very well and for very low effort and if i were to customize something with the world of like custom keyboard enthusiast kind of stuff i think what i would end up creating would actually be worse for my own preferences so for instance even though i know there's different key switches with different volume levels all of them are way too loud for me i've tried them all yes even the you know the cherry mx brown or you know whatever every few years there's a new one and people say oh this one's like this one's quieter or whatever and i try it and it's loud so that is problem number one is that i i don't really like the the giant mechanical key switch uh sound i also don't have any problem with the feedback of the relatively standard scissor mechanism of the sculpt ergonomics keys like that's not a problem i'm looking to solve i understand why people like those things it's just not that important to me and i don't like the noise trade-off secondly the sculpt is actually a very compact keyboard for an ergonomic keyboard and if i would actually create an ergonomic layout that has kind of the two key elements of what makes it help, which is the split gap and then the kind of curved and reverse tented shape of it.

1:33:31To actually create that with mechanical keycaps with that kind of keyboard would be probably a larger overall finished product. And I, again, to create that, it's solving problems I don't have to make a result that would be worse for my preferences. That's why i don't do it things might change over time my preferences might change or the availability of sculpt ergonomic keyboards as we've discussed it's kind of in flux right now um so that that could be getting better with like the the matthias recreation or the um was it in case who took over that business yeah i think so yeah so um so you know their recreations could be fine or could be ruined you know the matthias recreation could be fine or could be ruined i don't know yet i've pre-ordered both and we'll see what happens.

1:34:17And I still have a few sculpts left from my stock, my personal stock. So we'll see. But right now, this is a problem that I don't really need to be solved. Every time I use the sculpt, I'm not thinking like, God, I hate this thing. I can't wait until there's a replacement. No, I actually like it. I prefer it. And whenever they die, I just pop a new one out of the closet and move on with my life. Anonymous writes, as tech enthusiasts, how interested are you in LLMs and generative AI? Is it something that you follow from the sidelines or do you actually dive into these topics and do some of your own research and follow-up question for which use cases have you been using chatbots like chat gpt if at all and what were your experiences uh for me i'm interested certainly this is the new hot thing and unlike crypto it doesn't make me want to vomit all over myself so that's an improvement um i am trying to keep up with it and get at least a nominal understanding of how it works and in a very broad and basic understanding of how it works.

1:35:13In terms of how I use it, I don't typically do anything with generative making images or anything like that. I've tried a couple of times. I think when I was first trying to come up with a placeholder icon for call sheet then called Flickup, I was trying to come up with an icon that was at least passable, not to ship with, but just to ship a beta with. I couldn't even get something that looked even remotely like what I wanted at the time. And that very well could have been a user error on my part. But nevertheless, I do use ChatGPT every once in a while, typically to solve some sort of programming problem with something that I'm not familiar with, or a great example of this, a wonderful use of ChatGPT is how do I do this particular FFmpeg incantation?

1:36:01Because I know a lot of the things that I do regularly. and I like to think as people go, I know a fair bit about FFmpeg, but there's a ton of things that it does and almost infinite amount of things that it does that I don't know how to do. And I do have a folder in Apple Notes, speaking of Apple Notes, John, I have a folder in Apple Notes with individual notes of like different recipes, if you will, for FFmpeg. But there's certainly times that I'm like, well, how the heck do you do that? And ChatGPT oftentimes will either get it right or get me close enough that I can get it right quickly. And that's actually applicable to a lot of programming problems.

1:36:33So that's how I've been using it. I feel like I picked on Marco first a lot recently. So, John, what are you doing with this? I'm interested in these things. I'm interested in understanding how they work. I've tried to learn as much about them as I could without actually having any reason to make anything like this. It's difficult to follow sometimes. It's not my area of expertise. But I think I have somewhat of a handle on what their approach is. um but the most important thing like the distinction between crypto is that these these things these large language models and generative ai have practical utility you can use them to do useful things it's very obvious to everybody that's why people are excited about this again unlike crypto where it's like the useful thing is make a bunch of people a lot of money you know like okay but like does this help me with my work no not really um i use these things kind of in as a uh augmentation to google search right so anything that i'm doing where i would be using google uh i will throw in these various uh large language model things into the competition just like i may i uh you know put all my little voice cylinders in the house against each other and i ask them all the same question to see how they do i'm trying to see who's going to be the most useful so sometimes you know when i'm programming is a great example on a programming very often you google for things right oh i don't remember what that is how do you do this um you know what is the what is the order of arguments to this function i can't remember right um you can use google for that and you find a stack overflow question or you find the reference documentation or you know if you're in xcode you can go directly to it with just a right click or whatever um but yeah i'll throw in chat gpt uh google's bard thing the bing thing to see how they do um because sometimes even though i know the answer is in the reference documentation sometimes reading the reference documentation like if you're trying to look like the select system call in pearl and you want to know what what are those bit what are the constants that you use in the bit masks and is it the right bits then the read bits or the read bits and the right bits and you know you could look this up in the reference docs it's going to tell you the order of the things you're going to be able to find the constants then you can look them up and just like you'll get i know exactly where the answer is and you could google for it and it will probably point you to the reference docs that you can read to get that answer one of the utilities that things like chat gpt provides is you could just ask them for the call that you want to make and it will essentially not do that work for you but summarize uh that it'll say okay uh use this bit mask for the first argument use this bit mask for the second argument based on what you asked me so it already has found the the what order the arguments are in and it found the correct bitmask things stored together in the right things and it gives me the result faster than if I had done that work myself.

1:39:24One of the things I don't use any of these large language model things for is anything based on facts. Because unlike the line of code where I can just run it and find out if it does the thing that I want it to do, I can step through it in the debugger, I can see, yes, it's working, no, it's not working, right? I can tell whether it compiles or is it made up a function or whatever. using it for facts I don't find useful you know what year did this movie come out it'll give me an answer and the answer will be plausible but the problem is if I knew what the answer was already I wouldn't have asked and so now I'm faced with the answer it says oh this movie came out in this year is that right maybe looks like it could be right but now I have to go check somewhere to see if it actually is right and if I'm gonna go check somewhere anyway I might as well just start in google and end up at the wikipedia page or the imdb page or whatever now is wikipedia right is imdb right i don't know but at least i have a foundation for setting a level of trust how much do i trust wikipedia how much do i trust imdb when it comes to the release date of movies right how much do i trust you know this result from the new york times or the verge or whatever with large language models when it just gives me the answer for fact-based things it's useless to me because I mean I can just stare at that and go that might be the answer but I don't know where that answer came from and if it's right or if it's wrong like there's no the the LLMs don't know don't care like that's not that's not their job they're just smushing together a bunch of stuff and spitting out something that is plausibly an answer to the question asked and so that makes them entirely useless for me in that area and i would say also dangerous because you might think you're getting the answer and even like when google does the generative results at the top of its search results i tend to ignore those two for the same reason kind of for the same reason i've been ignoring the thing that google's been doing for many years before generative ai which is like don't worry about the search results we've extracted the answer for you i'm like yeah but like i kind of need to see where you got that answer from so i just ignore it and i go to the plain old search results and then i have to make the judgment myself like do i trust this seo spam page that's in the the first 10 results are all like that or do i want to actually go to something more authoritative or if i go to the wikipedia page is this page constantly being in edit wars and the talk page shows there's so much controversy about this or does everyone agree that this movie came out on this date right so i do use it i think there is there is a bright future of utility for things like this and the the image generators the same type of thing because setting aside the legal and ethical implications there uh people do agree that being able to type in a description of an image and get an image back is a useful thing to do if we can get the rest of the stuff worked out um so yeah i'm i'm definitely incorporating it into my into my normal workflows and i have found things uh kind of like steve jobs said about the ipad like there should be something where the ipad is the best device to use it there are things that i do when you know programming for example where chat gpt is better than google for getting me what i need faster there are also things where google is still better there are things where right clicking and looking it up in the reference docs is faster but very quickly things like chat gpt have become a sibling to direct lookups and reference documentation blind google searches and now right alongside their ask a large language model marco so personally i have used generative AI relatively little.

1:43:01I did make that icon for my Fire Island driving app. I made that with generative AI because that's an app that had zero budget and any time spent on it was time I shouldn't have been spending on it really. And so I needed something quick and dirty that would be seen by about 70 people and it was perfect for that. I don't really have a lot of need to generate a bunch of BS on a regular basis. So I tend not to use it that much. But it's a tool that I'm glad it's there when I need it. I'm more looking forward to, and some of this is already out there, tools that use generative AI functionality and techniques in ways that I actually do use on a more regular basis.

1:43:42So for instance, like if I'm editing a photo of myself, and I want to, you know, remove like the pimple on my forehead, you know, we've had various techniques to do that in image editors for a long time now but if you involve generative ai in that kind of feature you can usually do a better job of it or other things other kind of image manipulation things like you you have a low res image and you want to make a high res version of it and you want to you know do whatever you can to extrapolate whatever the details might have been in a higher res version of that image great that's a great use of a lot of these generative ai type techniques to do things like that so these features by themselves in like the current you know standalone products that we see them in, they have utility to a lot of people, no question, but I don't see a lot of that myself.

1:44:25But I'm looking forward to, and I'm already still currently benefiting from, some of the implementations of those techniques into tools that I use. So again, other instances, like if Apple ever realized that there's five of us out there who use Logic to edit podcasts, they can probably add a whole bunch of really interesting, useful features to Logic to make editing podcasts better, some of which would be AI based. And then another example of that is like the world of audio transcription algorithms has been radically improved by the OpenAI Whisper model. It doesn't get as much attention as the other OpenAI stuff, but Whisper has been a massive transformation, no pun intended, in audio transcription.

1:45:09I'm looking forward to advances like that, you know, that being integrated into more things. Like one of the big hopes I have for whatever Apple's AI sauce sprinkling process looks like this summer is Apple has a built-in API to transcribe audio. It's not that good. I'm hoping that this year it gets substantially better because maybe they're using some AI type models to, you know, improve their transcription. They've already been going in that direction in, you know, small steps here and there. So that's the kind of thing I'm looking forward to is, you know, yeah, there's going to be fun, new problems that arise that we didn't even think about that they can all of a sudden solve and it revolutionizes everything, etc.

1:45:50But most of what we do in our computing lives is boring tasks that we've done for a long time. Things like, as mentioned before, answering emails, you know, doing various media manipulation tasks and stuff like that. There's lots of opportunities in tons of computing for those features and those tools to get better within the capabilities of what we already know AI can do. We don't need it to get so much better that it can answer all of John's questions with factual accuracy. That'll be nice if it goes in that direction. It should. But even with what we know it can do today, there's tons of possible applications for the existing AI techniques, even the existing models that have already been built, in ways that we haven't seen yet, either because They kind of can't get built without somebody okaying it, like in the case of Apple being a gatekeeper over their platforms.

1:46:46Like, for instance, I would love the option to have ChatGPT respond to my Siri questions whenever they're about knowledge. Because even though it is a BS generator, it tends to generate more accurate answers than Siri does the vast majority of the time for informational questions. Well, I wouldn't say it's more accurate. You're more likely to get an answer, and Siri will just be like, sorry, I can't help you, or go look on the web. but honestly, that's the writer answer. I don't want something just giving me plausible BS when I ask a question. If you can't answer it, don't answer it. Well, just ask a man.

1:47:19Wow. Yeah, no, I do not want plausible BS. When I ask him when a movie was released, Alexa should tell me the real answer, not the plausible BS answer. Anyway, so I'm looking forward to just the integration of these awesome new AI techniques and models into the rest of our boring computing lives, making them a lot better. That's where I see the most promise. And yeah, we'll have some cool products along the way that are more exciting than that. But in the grand scheme of like, how is this going to impact most people? I think it's going to be more like the former, more just making our everyday tasks better and making the tools that do them better.

1:47:57And to be clear, Apple has been doing that for years and years. Back before the AI buzzword, they used ML, right? the magic thing where you can i don't even know if this would even qualify for quote-unquote ai today but like subject detection where you can drag the picture of your dog out of the photo apple's photo search where you can search for like you know tin can or whatever and it has a set of like that was ml and not ai but that's exactly the type of stuff we're talking about and what we're saying is hey uh state of the art has moved on you could make all those things better i i really because in photo search if you ever try this in apple photos you can search for things like dog or cat or bed or spoon or car or parking lot, right?

1:48:36But if you try Clipboard, you might see that, oh, the autocomplete isn't bringing up Clipboard. Well, can't I just type Clipboard? And the answer is no. Apple has decided there's a fixed number of things where you can search for. Well, technology has marched on and you can do text-based photo search a little bit better than Apple's doing it with quote-unquote AI technologies instead of the quote-unquote ML that they used for that before. Just make your existing features better. mark out the transcription is a perfect example they're already doing that with quote-unquote ml just do it with ai and it will you know but we're just saying like keep up with the joneses people are doing cool stuff and apple is being left behind and by the way as a i saw someone toot this today this is a website fast uh no it's yeah it's fast sdxl.ai it's a demo of another way these things could be interesting it's just a you know you type words and it makes an image let's go it's like an image generator like stable diffusion or uh mid journey or like oh you know this all sorts of things that can generate images for you uh but there is this subgenre of them it says we're not as interested in generating every image really perfectly or whatever our thing is speed this thing tries to generate images for you as fast as you can type so you just start typing and it's generating images and you add another word another word and it generates it generates it generates and you might think oh that's a fun party trick but i'd rather have i'd rather let you can i just type a whole sentence then you generate a really really awesome image i don't don't need you to do it too fast but being having the thing react to you in real time with like a smaller faster model does change how the interactions feel and this is just another example of like you know when i think about apple doing stuff on device or uh you know what's the advantage of doing it locally especially if your model needs to be smaller sometimes you want that trade-off Sometimes doing it quickly, locally, not as well is better than doing it in the cloud expensively requiring a network round trip or whatever.

1:50:33It just depends on the thing that you're trying to do. So I'm very optimistic that Apple can find many, many places in its existing products where it can use these technologies, which everybody knows Apple has been working on to make all of its products better in, I guess, what Mark would describe as boring ways. but honestly it's not boring to me if my image, my photo search starts working better. Thanks to our members who supported us this entire episode this week. You can join us and become a member at ATP.FM slash join. We will talk to you next week. Now the show is over. They didn't even mean to begin.

1:51:11Cause it was accidental. Oh, it was accidental. John didn't do any research Marco and Casey wouldn't let him Cause it was accidental It was accidental And you can find the show notes at atp.fm And if you're into Twitter You can follow them At C-A-S-E-Y-L-I-S-S So that's Casey Liss, M-A-R-C-O-A-R-M-N-T-M-A-R-C-O-R-M-N-S-I-R-A-C-U-S-A-C-R-A-C-U-S-A-C-U-S-A-C-U-S-A. It's accidental. It's accidental. They didn't mean to. Accidental. Accidental. Tech Podcast. So long. I'm looking forward to WWDC this year. like all these rumors coming out about you know maybe a slight updated design but also like tons of ai stuff i am looking forward to this yeah but what i talked about in the rectives episode was like what stuff will they have ready for this year right because they're you know i'm sure tons of stuff that they're doing you know someone's going to come next year someone's going to come out the year after that what can they get done in this year and the big question for me is is this the year where they do the apology siri essentially and say siri it doesn't suck now because that seems like the hardest one like siri is the hardest one making uh like uh you know transcribing audio better i think they can ship that this year right making image search better they could probably ship this year but the siri one is the big one and honestly it's kind of a distraction but marketing wise and pr wise like because i would be perfectly happy if it's like this isn't the year we make siri better siri still sucks but everything else is a little bit better autocomplete we already changed that to a transformer model last year but guess what now it's a little bit better and searching your mail it's a little bit better like there's so many places where they can apply this to just make it a little bit better it's not as exciting as saying siri 2 now it doesn't suck but i'll be perfectly happy with the wwc this year where every single piece of every single os has some of that ai sauce sprinkled on it and just gets a little bit better even if siri still isn't any good because i have faith that in a year or two or three siri will get what it deserves, which is a complete replacement.

1:53:33Last year was the year that we all were saying, well, all this new AI stuff is breaking on the scene, but it's a little too new for Apple to do much with it. They've had a while now. Generative AI stuff and LLMs have been really broken into the scene now for, what, about two years now or a year and a half? It's been enough time that Apple has had time to get caught off guard, realize they have to put a bunch of effort into this as everyone passes them by, start dumping money into it, and to start to see results. Now, this is the summer that we're going to start seeing the first, probably bigger scale efforts from Apple in this front.

1:54:14Yeah, like I said, they did do the LLM based autocomplete in, what was it, iOS 16? 17? 17. So, despite the fact that they were caught off guard, they've already shipped some stuff that says hey we found a way to use this tech to make something in our os better it's just it was just so tiny but this is going to be the big coming out i think exactly and it's not going to be the last big coming out of this stuff like we're going to see this over the over the following few years i would expect i do expect to see a whole bunch of those like you know as you mentioned like little stuff around the system features already existed maybe that are just now better that's what i want to see and honestly as a developer like yeah i guess i probably should have gotten into this with my answer.

1:54:52I guess we're doing this now in the after show. I've been hesitant to adopt much of this stuff in my app so far for two big reasons. Number one is that I just don't have the time right now. I'm in the middle of this rewrite laying the groundwork for it to be easier for me in the future to make bigger changes. But during this time of relaying groundwork, I can't make those bigger changes. I don't have the time. But in the future I will, hopefully. But number two, most of the AI features I could add right now would require me to do the work on the server. That is something that I could invest into, but I don't think it would be a great use of my time, and I think it would be very expensive for me.

1:55:35Even if I wasn't running the most cutting edge models or the biggest models or whatever, it is a very high cost for a one-person operation. It's probably not going to be worth it to me. When I look at what I would want out of some kind of server-based AI stuff or any kind of AI-based stuff in my app, what I most likely want would be transcription and recommending systems, like the recommending algorithms. That would be the two major areas of value for Overcast. But those are actually, like, relative to what people expect out of my app, those are not massive percentages of the value, so to speak.

1:56:15I want to have transcription and I want to have a more sophisticated recommendation algorithm. But if I never do those things, it's not the end of the world for my product. And so therefore, it is not worth investing a huge amount of time and money on the server end to try to make these two kind of ancillary features if it's going to be that expensive. Whereas if Apple is building in a lot more stuff into the device side of things, I would expect much of that to be exposed through APIs. For instance, the speech transcription algorithm has been part of the public API now for a while, I think something like six or seven years.

1:56:53If they improve the on-device transcription algorithms with AI stuff, I would expect I could probably use that locally on device. For all the same reasons that I don't want to run a whole bunch of AI stuff on my servers, Apple doesn't either on their servers. because the scale they would be dealing with with every iPhone owner having access to feature XYZ, that is a massive scale to deploy in Apple's data centers and massive costs to pay. I don't even know if there are enough GPUs in the world to handle that kind of load. Well, they're already doing a small subset of this with podcast transcription on the server side, right?

1:57:30But that's not in response to customers' requests for service. They just transcribe all their own podcasts once and then every customer gets the benefit it as opposed to essentially letting users say, take this image and make a description of it for me or something. Yeah, you don't have to transcribe everyone's videos they have recorded on all of their iPhones forever. Like that's server side, at least. And so, you know, what Apple has, Apple has this massive asset of all of these iPhones out there in the world that are pretty capable hardware devices that they don't have to pay server time to run.

1:58:04so Apple has a huge interest for multiple reasons in pushing as much of this as possible to the device side because they have this fleet of computing capacity that's out there in the world that is just, you know, the value there is immense. I also have a smaller version of that, a much smaller version of that with Overcast. Like as I lean into AI-based features in the app, I would so much rather do things client-side than server-side and right now, the client side versions of these things either don't exist or require me to bring in custom models like for instance the openai whisper models and that is possible there are a few apps that integrate whisper in certain you know features but it's a big thing like whisper is a huge model it comes in different sizes but even even the small ones on the phone are pretty substantial and a pretty heavy lift computationally but you know if apple got in there and optimized things with their optimization techniques to run on their hardware with their models, they could do a much better job of it if they chose to.

1:59:07And I hope they do. So anyway, I'm hoping that as I finish my rewrite over hopefully this year, hopefully that gives Apple enough time to deploy iOS 18 with some really nice features that I can use for my first feature update after the main rewrite release. And I would love to be able to do stuff like live transcription, transcription search if possible. There's so many things I could do if it was on device that I would probably never really have the prioritization and the budget and the time to do server-side.