Yesterday, John Gruber posted on his site Daring Fireball a blistering commentary, “Something Is Rotten in the State of Cupertino” on Apple’s recent announcement that the “Advanced Siri” would be delayed further than expected. While I disagree with some of the specifics, in all I agree with him on this. This is completely uncharacteristic of Apple, and unfortunately it is an area Apple is not great at.
To be clear, I love Apple and the products they’ve made, but there are several things they are just not great at. Online services is one of them, and I would argue that’s very closely linked to the “AI” craze going on at the moment. Sure, they’ve improved since the MobileMe disaster that Gruber mentions in this post (for the record, I worked in an Apple Store when that launch happened, and it was just awful. Our phone activation servers also went down… with a new phone, and a new cloud syncing platform all in the same day…). iCloud is a decent platform for most people. I’d even say it is really good at a few things, but we just kind of ignore the other things because we know they won’t change.
Compare this to other cloud services. What AWS can do, what Google Cloud can do, what Azure can (sometimes*) do, iCloud is extremely basic. Dig a little deeper and you realize that iCloud is actually running on these other cloud services for most of the hard work. Apple does a lot of work to keep things private and portable between platforms, which is great, but if you’re talking about a ~$4 Trillion company? It’s kind of embarrassing how little they do with the cloud.
You could easily argue that this doesn’t matter. Apple is largely a hardware company, and the software that runs on that hardware. As long as they continue to do that well, and lately I think they’ve been doing that extremely well, then people will keep buying their hardware, even if a lot of the work they do is to connect to a SaaS platform to do some sort of work. That might be true, but it’s potentially a little shaky. Now that AI is seemingly making a huge impact across the industry, it could be a bigger problem.
You can make arguments either way on whether AI will actually be revolutionary, or if it’s mostly hype and won’t live up to the expectations, but that isn’t really the point here. All of the leading AI companies / models right now have something in common - Huge amounts of cloud infrastructure to run these extremely computationally expensive models. OpenAI - Essentially unlimited access to the resources in Azure. Gemini - Google has probably the most advanced cloud infrastructure to run all of this on, despite missing (or squandering) their enormous AI lead, they have the infrastructure to at least stay relevant if not catch up or even win. Amazon - Huge amount of infrastructure, but kind of struggle with focus. Still, Alexa+ is supposedly coming soon, they have their own large models that are actually available and being used, and they partner with Anthropic to run much of their model building as well.
Apple of course has essentially unlimited money, so they of course could just throw that at one of these clouds and run models that way, but so far it kind of seems like they aren’t. Now the competency of modern cloud, and everything that comes along with it, seems to become a problem. Private Cloud Compute is interesting, and seems to have some cutting edge privacy features, but it’s hard to tell how much it’s actually being used right now, or what the plan is in the future. Almost feels like too little, too late. More importantly - they seem culturally incompatible with the way successful companies have built out similar products.
Still, I think Apple would be fine being the hardware that everyone wants to run AI models on. If they had thrown huge amounts of money at OpenAI, or Anthropic, and run those models more natively, Advanced Siri would likely be in a demo-able state, if not already pretty good. There are definitely arguments against doing these kind of partnerships, but from a technology standpoint, it would get them closer to what they promised. And that’s the big problem, as Gruber points out. The promise.
Can they pull this out? Of course they can, but it’s not guaranteed. They told everyone very specific things at WWDC about Apple Intelligence, and to Gruber’s point it is seemingly just vaporware to this day. That’s not ok. They advertised heavily on features that just straight up don’t exist.
The question unfortunately remains - Can they do it at all? It’s hard to say. This is something they are historically bad at. How many times have we been promised better Siri over the years? It’s fundamentally broken, and I’m not sure that the culture is right in that part of the company to truly do what it needs to do to fix it. They also clearly focus on running things on-device. This got us a long overdue increase in base RAM across the product line, which is fantastic. But is that enough? We obviously aren’t running what they promised, so what if they guessed wrong on the hardware requirements to even do that? That would be a huge misstep, and one of the big risks of pre-announcing things you aren’t completely sure is even possible…
*I couldn’t help myself…