Latest / In The Loop / EP20 - Apple Is Cooked—Here Are Three Reasons Why

Transcript

EP20 - Apple Is Cooked—Here Are Three Reasons Why · In The Loop

• 0:00
  This week was Apple's biggest conference of the year, WWDC
• 0:04
  2025. And recently Apple's not been in
• 0:07
  the news for many positive things.
• 0:09
  And spoiler alert, it only got worse.
• 0:12
  The conference can only really be described as disappointing at
• 0:14
  best. And what made it even worse was
• 0:17
  just before the conference, Apple's team launched a research
• 0:20
  paper. This research paper basically
• 0:22
  took a swipe at all the big AI labs, saying that large language
• 0:26
  models never reason, they've only ever memorized.
• 0:29
  In today's episode, we're going to discuss the conference, the
• 0:32
  research paper, and the fact that Johnny I've, Apple's most
• 0:35
  historically important designer ever, has just joined Open AI
• 0:40
  and explain why I think this spells a lot of trouble for
• 0:43
  Apple. This is in the leap with Jack
• 0:45
  Horton. I hope you enjoy the show.
• 0:58
  So let's start at this year's big keynote at the Apple WWE C
• 1:03
  conference, where usually they would announce the most exciting
• 1:07
  new updates and features and strategies of the year.
• 1:11
  And you're probably expecting some from VI News or Siri
• 1:14
  updates or something really interesting.
• 1:16
  So with that in mind, what do we think the big headline was?
• 1:18
  In actual fact, their big headline of the year was that
• 1:21
  they've revolutionized user interface with something they're
• 1:24
  calling Liquid Glass. Now before you get too excited,
• 1:28
  let me translate this from marketing speak to human speak.
• 1:31
  It basically took a trend called Glassmorphism from 2022 and
• 1:35
  probably arguably 2016 Windows and slapped a new brand name on
• 1:40
  it just like they did with Apple Intelligence.
• 1:42
  If you've never heard of Glassmorphism, it's a design fod
• 1:45
  where everything looked like frosted bathroom windows.
• 1:49
  Renaming this to Liquid Glass just like their Apple
• 1:50
  Intelligence rebrand. It's certainly we should all
• 1:53
  start to learn from which is branding is more important
• 1:56
  apparently than substance. And if we keep going in this
• 1:58
  direction of liquid glass or frosted glass windows and
• 2:01
  bathrooms that I guess eventually user interfaces being
• 2:03
  visible, which to say the least is not very accessible for those
• 2:07
  that are hard of sight, good luck trying to navigate the
• 2:10
  Frosted Windows App Store. What I find particularly funny
• 2:14
  is that Windows actually did this with Windows Vista about 15
• 2:17
  years ago. And anyone that remembers Vista
• 2:20
  would know how that turned out. Which is probably the worst
• 2:23
  operating system update in history.
• 2:24
  And I found some quite funny tweets about this Tom Warren on
• 2:27
  Twitter or X sorry, said that I can't wait to not be able to
• 2:30
  read anything on the new iPhone. And I think that's someday all
• 2:34
  this might make me seem like an Apple hater, which is just not
• 2:36
  the case. All I've got an Apple Mac, I've
• 2:38
  got an iPhone, I've got an iPad. So I'm actually the opposite.
• 2:42
  But I'm incredibly disappointed in them.
• 2:44
  And I think to be honest, they're in big trouble.
• 2:47
  And I've come on to exactly the reasons I think that is going to
• 2:49
  be true. This is why they're in trouble,
• 3:02
  because they're so far behind on Siri.
• 3:05
  And the fact that they can't have a unified vision of what
• 3:08
  Syria is supposed to be is kind of crazy to me because open AI
• 3:12
  are literally telling them what they should do, which is a
• 3:14
  universal assistant on every single device.
• 3:16
  That's what Sam Altman came on a stage, and that's their vision
• 3:19
  for what AI is supposed to be. And this isn't just a small
• 3:23
  little problem that Apple are facing.
• 3:25
  This disaster's been brewing for over a decade.
• 3:28
  I think it's a master class on how to completely fumble one of
• 3:30
  the biggest tech revolutions probably since the smartphone.
• 3:33
  So if we start with Siri, because that's where this whole
• 3:36
  mess started. Siri was launched in 2011, so 14
• 3:39
  years ago, and somehow it's gotten arguably worse.
• 3:44
  And meanwhile, Google have launched their Gemini where they
• 3:46
  can actually have real life conversations.
• 3:49
  It understands context of the app and the screen.
• 3:52
  You can have your camera holding up to things in the real world,
• 3:55
  and it can tell you what it sees.
• 3:57
  And it integrates in all applications.
• 3:58
  And one of the reasons for this complete fumble is that Apple
• 4:00
  had been at war with itself for years until very recently.
• 4:04
  John G Andrea was head of AI at Apple, and Apple poached this
• 4:08
  person from Google in 2018 to run all of Apple's AI efforts.
• 4:13
  When this happened, the head of software at Apple, So Craig
• 4:16
  Federighi apparently had 0 faith in this person and what he did
• 4:20
  is he started a competing AI team.
• 4:24
  So that's right, Apple had two separate AI teams in the same
• 4:27
  company fighting for the same resources trying to build
• 4:29
  competing products. The team under Federighi
• 4:32
  actually nicknamed that other team as the Aimless.
• 4:36
  So you know that company culture is probably quite problematic
• 4:38
  when employees are creating nicknames that are very toxic
• 4:41
  and negative about each other. And it gets even better.
• 4:44
  When Chachi BT launched, Gee Andrea basically said that LLMS
• 4:47
  have very limited utility in the world.
• 4:49
  This culture sounds very much like the early days of Apple,
• 4:52
  when Steve Jobs was to build this rampaging around Apple,
• 4:55
  doing incredible things but being quite divisive.
• 4:57
  Now this dysfunction got so bad this year that Tim and Kirk
• 5:00
  apparently lost all confidence in Gee Andrea and strict control
• 5:04
  of Siri from them. And I actually handed it to Mike
• 5:07
  Rockwell, who was the person behind the Vision Pro.
• 5:10
  So the Apple Vision Pro, if you remember though, it's the three
• 5:12
  and a half $1000 headset that probably about 1000 people
• 5:15
  bought. And he goes well beyond just the
• 5:17
  culture with their AI efforts, because Apple's internal data
• 5:20
  centres are running on about 50,000 GPUs, many of them were
• 5:24
  five years old. And the AI teams have been
• 5:27
  begging the likes of Google and Amazon just to get enough
• 5:30
  computing power to train the models.
• 5:33
  So that's the same Google and Amazon they're actually trying
• 5:35
  to compete with. And then if we go back in time
• 5:37
  when we remember that Apple was promising all these amazing AI
• 5:39
  features that are going to be ready for the iPhone 16 launch,
• 5:43
  well, it turns out only half of the promised chip budget got
• 5:47
  approved because the Apple CFO told the AI team just to do more
• 5:50
  with less. So when all the other big
• 5:51
  companies see that AI is actually the next big computing
• 5:55
  platform and is spending really Big, Apple cut their AI team's
• 5:59
  budget in half. Now, I'm not saying spending
• 6:03
  more equals better results, but if you're a technology company
• 6:06
  and you're not spending more making big bets and willing to
• 6:09
  RIP up the entire playbook or business model, I think you're
• 6:13
  in deep trouble. The idea of product market fits,
• 6:15
  or when a product aligns with a market and a need and a want is
• 6:19
  something that every company has lost right now because every
• 6:23
  single person's needs and wants are changing so fast because of
• 6:27
  this new technology. And what frustrates me most
• 6:30
  about Apple right now is that Steve Jobs would have cut 75% of
• 6:33
  the fat when it comes to Apple. So he'd have cancelled every
• 6:36
  product that wasn't going to deliver triple digit growth.
• 6:39
  He would have ruthlessly cut teams and made everyone
• 6:42
  incredibly focused and aligned on what was most important.
• 6:45
  And he'd guarantee he would have been in the office of the
• 6:47
  software and the design teams every single day making sure
• 6:51
  this thing happened and this sort of energy with a founder or
• 6:53
  a leader or a visionary. He was in the office with the
• 6:56
  team day in, day out driving those outcomes.
• 6:59
  It's something that Steve Jobs was famous for and I guarantee
• 7:02
  something that has been lost over the last few years.
• 7:16
  And this comes to the second reason I think they're in
• 7:17
  trouble, which is their way of thinking about the world.
• 7:20
  And this comes to the point of that research paper.
• 7:23
  So Apple know they're behind and their answer wasn't to try and
• 7:26
  make big bets. For example, I suggested about,
• 7:29
  I think about 10 episodes or so ago, that if I was Apple, I
• 7:33
  would spend a huge amount of money to buy an AI lab and just
• 7:36
  leapfrog everybody and own our own model.
• 7:39
  So we own every inch of the stock and we can triple down on
• 7:42
  that user experience across all interfaces in every device
• 7:47
  because Apple can own the stock. That's always been their
• 7:49
  strategy. And instead they've not done any
• 7:51
  of that. They've released a research
• 7:53
  paper arguing basically that all these large language models are
• 7:57
  rubbish and they're not actually reasoning.
• 7:59
  Tim Cook must have genuinely been furious because this became
• 8:03
  the backdrop of the conference right around the time of their
• 8:06
  annual conference, the WWDC 2025, when everyone was
• 8:09
  expecting Apple to have some kind of cool breakthrough.
• 8:13
  And instead they dropped a paper called The Illusion of Thinking,
• 8:17
  Understanding the strengths and limitations of reasoning models.
• 8:20
  Now let me break down what this paper actually claimed, because
• 8:23
  it's important to understand. So our researchers tested AI
• 8:26
  models on a puzzle called the Tower of Hanoi, which is
• 8:29
  basically a game which stacks rings and you have to move disks
• 8:34
  from one pole to another without putting a bigger disk on a small
• 8:38
  one. It's a logic puzzle with a known
• 8:40
  algorithmic solution, and the big finding was that AI started
• 8:43
  failing when the puzzle got too complex.
• 8:45
  So Claude could handle about 6 discs.
• 8:48
  It struggled with 7 and gave up at 8.
• 8:50
  And the conclusion was the AI models therefore don't reason,
• 8:54
  they just pass a match really, really well.
• 8:55
  Now, of course, this was a great day for many AI skeptics, namely
• 8:59
  Gary Marcus, who has been quite famous in AI Twitter and social
• 9:03
  media circles for trying to make a career out of basically being
• 9:08
  incredibly pessimistic about AI, and he tried to call it a
• 9:11
  knockout blow for LMS. And obviously lots of people
• 9:14
  were sharing this online too. That being said, my personal
• 9:17
  opinion plus many researchers I'll go into, is they're just
• 9:20
  all revision. It really doesn't matter.
• 9:22
  For example, many pointed out that AI models weren't failing
• 9:24
  because they couldn't reason, they were failing because they
• 9:27
  were hitting token limits, which is the equivalent of running out
• 9:29
  of space to write their answers on a piece of paper.
• 9:33
  One researcher called Lisa and Avgab actually replicated the
• 9:36
  exact experiment and found that models were literally telling
• 9:39
  the researchers they couldn't complete the tasks due to the
• 9:42
  output constraints. So instead of just giving up,
• 9:44
  the models would explain the algorithm for solving the puzzle
• 9:47
  and create the code to implement it.
• 9:49
  In other words, the AI was smart enough to recognize its
• 9:52
  limitations and provide alternative solutions.
• 9:55
  But Apple's researchers said they interpreted this as proof
• 9:59
  that AI doesn't actually reason. And what was also reported is
• 10:03
  that upper researchers actually prevented the AI models from
• 10:05
  writing code to solve the problems.
• 10:08
  And they acted surprise when the models can solve coding problems
• 10:11
  without being allowed to code. And one of the researchers let
• 10:13
  the AI models use their full set of capabilities.
• 10:16
  So reasoning, researching, writing, coding, they solved the
• 10:20
  puzzle really easily even with hundreds of disks.
• 10:25
  So my take here is that Apple is really in trouble if it doesn't
• 10:27
  start to change its thinking fast again.
• 10:30
  This Twitter post sums it up perfectly.
• 10:32
  Be Apple, richest company in the world, every advantage
• 10:35
  imaginable. Go all in on AI, make countless
• 10:37
  promises, get immediately lapped by everyone.
• 10:40
  Then two years into the race with nothing to show for it,
• 10:42
  you'd give it write a paper up how how it's all fake anyway and
• 10:45
  it really doesn't matter. Now I don't think it even
• 10:48
  matters if models reason or not because most of us care whether
• 10:51
  this model does a thing better than the last model and faster.
• 10:55
  Not the academic definition of reasoning.
• 10:57
  Now that's not to say these debates shouldn't happen, they
• 10:59
  really should. But Apple making this their
• 11:02
  issue to focus on is just madness.
• 11:05
  They're a consumer company. They care about what everyday
• 11:07
  people and customers want to do faster and easier.
• 11:21
  And this gets us to the third and final reason I think they're
• 11:24
  in trouble. And this one I find particularly
• 11:25
  interesting. And the third reason is that the
• 11:27
  core business might not be there forever.
• 11:29
  And if it is there, it's going to change.
• 11:32
  As we've seen. I have very little faith in them
• 11:33
  executing and importantly, managing that change.
• 11:36
  A good example of this change is Johnny Oliver and Sam Altman.
• 11:39
  And now if you don't know Johnny, I've other than probably
• 11:41
  Steve Jobs and Tim Cook, there would literally be no Apple
• 11:45
  today without him. He was this legendary designer
• 11:48
  that came up with his team for most of the concepts between the
• 11:51
  iPod, the iPhone, iPod, and basically every iconic Apple
• 11:55
  product that ever made the company what it is today.
• 11:59
  And he just joined Open AI. So let that sink in from Evan,
• 12:02
  Apple's most famous and most important designer ever.
• 12:05
  The guy who literally shaped the modern tech aesthetic.
• 12:10
  He turned Apple from a computer company into a lifestyle brand,
• 12:13
  has now sold his business to Open AI and partnered with Sam
• 12:16
  Altman on creating what they call the next generation device
• 12:20
  of the AI era. Sam Altman aims to ship 100
• 12:23
  million AI devices and thinks that this collaboration could
• 12:26
  add a trillion dollars in value to Open AI.
• 12:28
  And with a number of leaked meeting recordings, Sam Altman
• 12:31
  described their secret device as some of them that will be fully
• 12:34
  aware of the user's surroundings, unobtrusive enough
• 12:37
  to rest in someone's pocket or a desk, and positioned as a third
• 12:41
  core device that a person would put on their desk after a
• 12:43
  MacBook and an iPhone. So while Apple is making the
• 12:47
  apps on their phones look a little bit more transparent and
• 12:50
  publishing papers on how AI doesn't even really work
• 12:53
  properly, their former design genius is literally building the
• 12:56
  future AI hardware with one of their biggest potential
• 13:00
  competitors. Because Open AI is the leader in
• 13:03
  consumer AI. That's where a massive chunk of
• 13:06
  Open AI's revenue comes from. And this is just one example of
• 13:08
  people trying to invent the next consumer device.
• 13:11
  Take Meta. They've actually partnered with
• 13:13
  Ray Ban on glasses that take great photos and videos.
• 13:16
  The audio quality is apparently really good and obviously
• 13:19
  they're stylish enough to probably make people wear them
• 13:21
  because it's Ray Ban and now Google is doing the exact same.
• 13:25
  They're partnering with Warby Parker.
• 13:27
  And if Google started shipping very nice looking glasses
• 13:30
  powered by Gemini with real time, say, translation, vision,
• 13:33
  AI, search and integrate into all of their services, what is
• 13:37
  Apple going to compete with Siri?
• 13:40
  Now, I can't say if any of these potential devices could really
• 13:43
  take over the iPhone, but again, I don't believe that Apple could
• 13:46
  react and change and execute properly on this if it did.
• 14:01
  So I'll give you one final important tape before I say
• 14:03
  goodbye today. I don't think this is about AI
• 14:05
  or design language or corporate dysfunction.
• 14:08
  I think it's about a company that's built its entire modern
• 14:10
  identity on being ahead of the curve and taking existing
• 14:13
  technology and making it so much better and redefining entire
• 14:17
  product categories. The iPhone wasn't the first
• 14:20
  smartphone, but it was one of the few that didn't suck.
• 14:22
  Apple's whole thing was let others fumble around and then
• 14:25
  we'll turn up and do it. Brian, this time they're turning
• 14:27
  up very late, empty handed, relying on other people's
• 14:31
  technologies and are spending time publishing research papers
• 14:33
  that seem to me quite petty. And the tragic part is this was
• 14:37
  avoidable. Siri was around in 2011.
• 14:41
  They had a seven-year head start.
• 14:42
  But instead of building on that advantage, they let internal
• 14:45
  politics and egos and nonsense destroying any potential
• 14:48
  progress they could have made. Almost all of this was
• 14:50
  happening. Open Eye was just busy changing
• 14:52
  the world. I don't want to be completely
• 14:54
  unfair here. Apple still makes good products,
• 14:56
  but as someone that has loved the way Apple have produced
• 15:00
  industrial design at scale and appreciate how difficult it is
• 15:04
  to create such incredible products that so many people
• 15:07
  love and build their entire brand on, think different.
• 15:10
  Well, it just leaves me very disappointed and it's a big
• 15:14
  lesson of what not to do if you're a leader of an
• 15:16
  organization right now. Embrace focus and start making
• 15:20
  bets, otherwise you'll probably regret in a few years.
• 15:23
  Anyway, thanks for today. I hope you enjoyed this episode
• 15:25
  and as always, share this into a slot channel a WhatsApp chart to
• 15:29
  a friend that lives Apple and may want to hear an interesting
• 15:33
  perspective. See you next week.