[varunnrao]

imo, Apple are actually ahead when it comes to the hardware side of the whole thing. Their vertical integration gives them an edge not many can match, when it comes to running ML models. It's a no-brainer for Apple to reduce the barriers for devs to build really cool native Mac/i{Pad}OS applications and incentivize them to leverage the built-in AI/ML abilities to a greater extent. The iPhone in part took off _because_ of the whole app ecosystem that got built around it. Sure they might take a loss in their services revenue in the short term but they get to be _the_ AI platform for at least the next decade and half - both on-device and server side with their new Apple silicon servers.

It's just that most Apple software seems to suck in some fundamental way right now. I don't know if it's a technical issue (SwiftUI being meh when compared to UIKit for example) or a culture issue or the money coming in insulating management from accountability. Software execution has been lagging behind the excellent hardware execution for almost all of the Tim Cook era. They desperately need someone like Scott Forstall to come in, kick butts and get stuff going again.

They ideally have a couple of years while waiting for Moore's Law to catch up to turn around their software side. Otherwise, it's a real shame that all that great hardware is just being used to run Electron B2B SaaS apps.

[krackers]

They also had (but squandered) the potential to be ahead on the software side. macOS is the only platform I'm aware of that has every app wired for scripting (AppleScript/Apple Events). And not only that, they already solved the issue of adoption since almost every well-behaved (read: non-electron) application has decent support for AppleScript.

It would take very little effort to put an LLM frontend on top of this, and yet they've not only abandoned applescript (or the underlying apple events) at a time when they could expose it to the masses, but have gone in the exact opposite direction with "Shortcuts".

Oh and the icing on the cake is that apple events can be sent over the network as well, and this infrastructure has existed since the early days of OSX.

[varunnrao]

I agree. The AppleScript/Apple Event Manager thing is an example of treating macOS like a second class citizen in the ongoing iOS-fication of the Apple Ecosystem. The point of macOS for me is that it's simple to use for most beginners but allows more advanced users to add complexity through tools like Apple Script and Automator and the underlying Unix base.

Like you say, an MCP server integrated with AppleScript/Apple Event Manager would instantly hook up any LLM with virtually all Mac apps. More Mac devs will then be incentivized to support these features. For people who find AppleScript un-intuitive, JavaScript is also supported. And in my view, this is a revolutionary way to use my computer - a very Star Trek way of using the computer.

[chippiewill]

> Their vertical integration gives them an edge not many can match, when it comes to running ML models.

They have an advantage when it comes to running them locally, but it feels like they're trying to push it onto consumer hardware before consumer hardware is at the point of actually being able to run useful LLMs.

[varunnrao]

You're right. The hardware right now can't run useful models.

But that's why I think they have a couple of years to sort out their software issues. When useful models can be run on their devices they have to be ready. The hardware advantage can only be an advantage when they have the software to run useful applications. Hopefully they don't get stuck in the typical big company bureaucracy and ego matches and instead can make a change for the better.

[pzo]

They made it less useful because they were greedy. Before M1 we used to have laptops that had 16GB RAM as a base. With M1 the made base back to 8GB. In PC world before M1 and even before apple started soldering everything you could have easily laptop extended up to 64GB RAM for a cheap price. Those ram sticks are not expensive in retail price and should be even less expensive if wholesale and not full sticks as bill of material but just memory modules.

If last year they would make each macbook air as a standard have 32GB RAM and iPhone 16GB RAM + 250GB SSD as a base they would have the most capable hardware with big user base.

Sure they loose some money of people having upgrade models but they would sell much more Macs. As a reference they sell each year only ~20M Macs comparing to 60M ipads and 240M iphones. Macs are having only like what ~10% market share worldwide? They could easily double it but they protect their profit margin like virginity.

[wtallis]

> Before M1 we used to have laptops that had 16GB RAM as a base. With M1 the made base back to 8GB.

You're getting product segments mixed up. From what I can tell, the 13" MacBook Pro in the Intel era never started at 16GB; the last model still started at 8GB. That's what was replaced by the M1. The 15/16" Intel-based MacBook Pro models didn't get a proper replacement until the M1 Pro and M1 Max, which started at 16GB and 32GB respectively. The only regression I can find there is that the last 13" Intel MacBook Pro could be configured with up to 32GB, which wasn't available from the base M series chips until the M4 last year.

[kelseydh]

A lot of people forget that it was only recently that the Photos app on your iPhone could run OCR text search on pictures in your phone. Google had that feature on their phones many years before Apple.

[visarga]

Apple's TTS voices still run on 10 year old technology. Pretty disappointing, at one time the had the best system voices.

[jldugger]

The ML blog seems to disagree with that take: https://machinelearning.apple.com/research/on-device-neural-... (2021)

> Recent advances in text-to-speech (TTS) synthesis, such as Tacotron and WaveRNN, have made it possible to construct a fully neural network based TTS system, by coupling the two components together. [...] However, the high computational cost of the system and issues with robustness have limited their usage in real-world speech synthesis applications and products. In this paper, we present key modeling improvements and optimization strategies that enable deploying these models, not only on GPU servers, but also on mobile devices

[microtherion]

Having worked on Apple's TTS for more than a decade, I can state with confidence that this is utter bullshit and you don't have the slightest idea what you are talking about. Both in terms of quality, and of the underlying technology used, Apple's current TTS is in no way comparable to what existed 10 years ago (at Apple, or anywhere else in the industry).

I challenge you to find a 2014 recording that is on par with a contemporary Siri voice.

[pzo]

I have been playing recently with those enhanced TTS model and they are of similar quality like piper TTS models to me - not that good. StyleTTS 2 like kokoro sounds so much better for me and also run realtime on their devices. And when you compare their online models to not even what OpenAI have but some small recent startups like Sesame or open source models like Orpheus, Apple TTS sounds (pun intended) really behind.

[visarga]

I don't dispute your claim, just that I still find Alex voice to be the best, and it's been the same since over 10 years ago. The other voices have issues, they don't sound too good at 1.5x.

[microtherion]

Ah, that's more specific.

Alex was developed when VoiceOver (the screen reader) was the primary use case for text to speech. Consequently, it was optimized for low latency and robustness under rate changes.

The Siri voices sound much more natural at 1x and have a higher signal quality, but rate changes were a lower priority for this use case.

Fun fact: when we worked on Alex, many VoiceOver users stubbornly hung on to Fred (which is mostly using late 1970s technology). Screen reader users are not fond of switching voices; it appears their hearing locks in to a particular voice, so switching is costly.

[resource_waste]

>imo, Apple are actually ahead when it comes to the hardware side of the whole thing.

It surely is just your opinion. Nvidia is king and Apple has found a way to market an integrated GPU and CPU RAM as something magnificent, rather than something that has existed since the dawn of computing.

There is a reason Nvidia is king. There is a reason corporations buy Nvidia and not Apple for their LLM uses.

[wtallis]

You seem to think that the AI market consists entirely of the segment that NVIDIA dominates (datacenter training and inference) and that the segment where NVIDIA is absent (inference on battery-powered devices) doesn't exist.