Apple Fell Behind in the AI Arms Race
wsj.com
wsj.com
They seem to try stuff internally that other big players seem to jump into wildly. They had a self-driving project when Google/Thrun started out but didn't push it. They never really got into the cloud space when MS/Google jumped into the pit with AWS.
For merits and faults, Apple seems fairly laser-focused when it comes to priorities and are unlikely to go head first into The New Thing.
AI becoming commoditized will probably validate that. There doesn't seem to be much of a moat for big AI companies at this point, it's just a data arms race and Apple is as well-positioned as anyone to take advantage (when the time comes).
For the longest time they chased "thinner, lighter" only to have people complain that the products are too thin and would rather they had more battery. They made a big song and dance about dropping all that, did so for a bit and then went straight back to it with the ipads now. While we're at it, it's worth mentioning Vision Pro which is also fundamentally something no one asked for.
The reality is that their products are no longer built for their users, they're building for the shareholders. Built to make line go up which I think is going to be their slow death.
As a hardware company, there's a constant need to produce the next iteration of device. Same with software. So you get pretty lackluster year-on-year improvements, but over a 2-3 year cycle, it's pretty impressive. They seem to be refocusing on software quality, which should show some innovation on that front. Apple Vision Pro was their attempt at "the next big thing" and it just didn't quite get there. They probably believed VR had soaked in the market long enough that there was a majority waiting for the best version of it. They sort of missed here, but sort of didn't. There's potential for spatial computing, but the hardware isn't quite there yet.
On the topic of AI/LLMs, they've always had machine learning in their stack, but they haven't highlighted it as something revolutionary and they certainly haven't referred to it as "AI" (because, well, it really isn't artificial intelligence). Right now, most AI products are just wrappers around chatbots. Chatbots are an okay experience, but not ideal. That particular interface, in my opinion, gives an impression of infallibility to these products when it couldn't be farther from the truth. They're nifty, but the current generation of AI products is primitive, rough, and in some cases dangerous.
Apple's caution on AI is actually refreshing. I'm optimistic that the focus won't be on providing AI as first-class product but rather as an augment that makes their existing technology better. Case in point is the improvement in iOS 17's predictive keyboard. They shipped a small language model that improved predictive text leaps and bounds (though, still primitive).
It’s not about defending Apple, it’s about fair journalism. This headline is a journalistic fallacy, as anybody who is actually an expert about the company would understand that determining where Apple stands in a software development cycle is simply impossible.
Maybe the Wsj is right this time after being wrong n times before, but I wouldn’t bet on that as much as I would bet that Apple has been working steadily and secretly at its own vertically integrated solution instead of flashing whatever AI research they had like all competitors. Which is exactly how Apple works, in case one is willing to stop and observe and report, instead of running sensationalist headlines to leech views right before a hyped event like wwdc.
So far what we have is companies like Microsoft rolling out poorly thought through "AI Clippy is going to fuck your day up and boost our stock price 10 fold" type campaigns. Apple is not going to do that.
I think the main interesting news here is that it sounds a lot like Federighi has let Giannandrea fail on his own terms and now is moving in to pick up the pieces. If this is accurate, it doesn't bode well for Giannandrea's tenure at the company.
What I see is a lot of flailing and tripping over one's own feet. Google most recently launched the AI Summaries which they've hurriedly backpedaled. Before that it was the embarrassment of their image generator.
Google has also announced various things that actually IMO look quite useful (ex. AI Teammate) but nothing has launched. The little they did let the press play with at I/O largely fell apart as soon as the reporters strayed even slightly from the demo use cases! As far as I can tell just about every LLM-based integration is a tech demo at best that's held together with prayer and glue, and falls apart if you breathe at it wrong. Production-ready these are not!
Microsoft has also repeatedly beclowned themselves, most recently with Recall, which seems to have some very grave privacy and security implications while offering questionable user value in exchange.
There may be in fact be something here, but we're acting like players have found strong PMF and are scaling when it's clearly not the case!
I do trust them to build the best hardware stack to enable a local LLM.
Will be interesting to see how this plays out in the short term.
But in the long run I think they're very competitive.
Personally I don't know anyone that's hyped about OS level local or cloud AI IRL. If MS can't even properly iterate a few hundred installed programs in a start menu search, does that make me excited about rewind?
There's a killer app for local AI out there and I hope it's brought to us by 3rd parties that dogfood their own stuff.
Because that’s where they could leapfrog the others. They care enough to do things on device (even if it’s inferior)
They can partner with AI companies and build expertise for on-device AI (as they have with photography) leveraging their hardware design talent.
Apple Maps, Magic Mouse charging from the base, Butterfly keyboard...
So Apple pushed ahead with their own maps instead, hoping that they'd be good enough to keep people from wanting Google Maps, which ... didn't pan out. Instead, Apple was forced to allow Google Maps in the app store, since their own maps weren't good enough, and people were complaining. So in the end, Google got the data, since lots of people use Google Maps instead of Apple's maps.
I much prefer their directions UI to anything else out there. It’s really well thought out.
It feels like a perfect solution to me if the primary goal is to make sure customers have the Magic Mouse experience Apple wants them to.
I might swap it out for Siri, which still tries to call emergency services when asking simple questions from time to time.
So thank you, Apple, for not running in this arm race. Not every race is worth participating.
Where is AAPL competitive position being compromised? * Android vs iPhone: they just need to continue executing * Windows vs iOS: MSFT is actually telling them how to position
Given their technical moat on integrating hardware & software there is an hidden opportunity for them in being the best supporting local/on-device LLMs
Out of the major players they have less disruption risk, grow depends less on AI meeting expectations and they have hidden optionality. How is this falling behind?
A new important one, but without the ability to reduce hallucinations or trace why the LLM chose the approach, the ascendancy of AI is a bit premature to judge.
It's a common figure of speech but still interesting WSJ framing the headline. Apple is not a defence company. I guess we don't have "tool races" or a "utility race". Not even just an "AI race". No, this technology is definitely a weapon. Am I wrong?
Anyway.
So Apple has all these users just because they had a great pitchman who died over a decade ago??
I don't get this rationale at all. Yes, obviously Apple has built a very "sticky" ecosystem, but saying "Apple's value isn't it's tech" is ludicrous to me. Apple depends on at least the belief that it has the best products to sustain its user base. If that belief starts to waver, so will their user base.
Products are not tech. For example, before Apple Maps, Apple had the arguably best mapping product, but did not have the tech (that was Google's). They are in a similar position with respect to the current crop of AI technologies. They could build their own tech, or they could just go to, say, OpenAI and say: "We have all these users. It would be pretty great if they started using your platform, wouldn't it?"