Ask HN: Where is Apple? They seem to be left out of the AI race? | Hacker News Reader

Ask HN: Where is Apple? They seem to be left out of the AI race?

9 pointsThinkBeat27 comments

This is probably a case of me not paying enoguh attention to the news, but I would have thought that "AI Assistance" would be square in the middle of something Apple would want and need

View on HN

On the machine learning side of AI, they have CoreML. You can drag-and-drop images into Xcode to train an image classifier. And run the models on device, so if solar flares destroy the cell phone network and terrorists bomb all the data centers, your phone could still tell you if it's a hot dog or not.

https://developer.apple.com/machine-learning/

https://developer.apple.com/machine-learning/core-ml/

https://developer.apple.com/machine-learning/create-ml/

Journalists for several years said Apple is too late with their smartwatch and others capture the marke.t Then Apple delivered strong. It's not a guarantee but the AI market ("race") is still early.

2015: https://www.forbes.com/sites/erikkain/2015/03/09/the-apple-w...

"Consumers have plenty of choice these days, and I have a very hard time seeing how Apple could make itself stand out in such a crowded field to anyone other than Apple devotees. The Moto 360 is a terrific, attractive Android smartwatch. The Samsung Gear 2 is an equally appealing option. "

2023: https://appleinsider.com/articles/23/02/22/apple-watch-domin...

"Apple Watch accounted for 34.1% of all smartwatch shipments in 2022, and 60% of the revenue for the entire market, globally. "

To be fair the actual benefit of AI is more software than hardware driven.

I imagine apple is waiting for the hardware to be such that it’s cheap to do on device, and then see what types of software use cases are best.

I disagree. AI’s true ability shines when supported by hardware. Or rather baked into Hardware as a first class citizen. A chat GPT level Siri would always win against Chat GPT since it’s the default option.

Apple's entire MO (on paper) is being late to the party, but being the best at the party when it arrives.

Its the exact opposite.

They are the first to do a lot of things, and in terms of product design they do a good job, which establishes them solely in the market space even though first iterations of products are generally subpar overall and riding on one feature that makes it stand out. Then when competition catches up, they already have an established user base that is "locked in" to Apple products through their walled garden.

The reason why they don't have any AI stuff is because AI is purely software, and Apple doesn't do software that well because any software it releases had to be wrapped by that walled garden to secure the user base.

AI is both software and hardware, which is why they spend a few refresh cycles improving the Apple Silicon Neural Engine, and are now releasing MLX since there is better adoption.

The thing is, parallel processing (i.e same instruction on multiple data pieces) is easy to implement for the most part in hardware.

The issue is nobody wants to write low level software for a particular chip. Cuda is why NVIDIA is ahead despite AMD having comparable hardware.

If you look at what its like to develop for MacOS or iPhone, its pretty clear that apple doesnt do this well. Even for their neural engine, I think you have to do some sort of entitlement thing to get code to run.

i don't think apple has been first to do something since, like, the iphone. their business model is definitely rock up a little later than the others and combine both good quality and a giant marketing machine to whatever maket they enter, the days of apple doing firsts is over i believe

You mean like being the last to build a custom world class desktop/laptop/phone CPU for their computers and phones?

Or like being the last to produce actually usable VR headset?

you think the vision pro is the first VR headset? it doesnt even have the most specs or features of current headsets, its just apple so you know about it. as for the CPU, uhh, its a particularly good ARM chip? its definitely not the first, it just does a very good job

What other actually usable VR headset exists?

The Quest, as an example that isn't tethered to a battery.

Even the iphone was'nt the first of it's kind. There was blackberry, Window Mobile, Symbian-Phones. Heck, not even Smartphone as a term was coined for the iphone. Apple was just the one who created a pleasant experience without compromising for price or history.

AI is in it's infancy and most products leveraging gen-ai do not provide a good experience. The tech demos are really cool and make grand promises, but in real-world work, the cracks show almost instantly. It's also really hard to have a super controlled AI experience, and Apple loves to control the experience for the user as much as possible.

Apple is going to take a slow approach to AI, so I wouldn't expect anything major from them in this space any time soon.

This is the truth. Generative AI is new and really really cool but it also sort of sucks right now if you want to actually use it for anything. I know plenty of people disagree and can find uses for it in its current state, but if the value was real and general it would be much more self-evident. The UX is just pretty meh across the board.

Ai is overhyped now.

It is only smart for a powerful mature company to wait for dust to settle and take what they want while first comers lie exhausted by competition.

Apple would want an AI assistant to run locally on your phone. Not to be dependent on sending your data to the cloud to give you results.

I don’t see how you think they are behind in general? They’re shipping a fairly powerful neural engine with their processors and they’re using them more and more.

They’re building out their CoreML framework. I don’t know how good this is.. is there an equivalent framework on other platforms that run efficiently on phone, tablet and PC?

If they haven’t made an LLM based AI assistant that runs locally it’s probably because it’s not possible to get good results yet. Is there anyone else that has managed to do it yet? I’m not aware..

Why not make a good cloud based assistant? I suspect one reason is they don’t have an advantage to leverage in cloud computing. I think they’d much rather try to leverage the advantage of the billions of chips they’re designing and selling themselves than being dependent on building a big cloud compute platform with other companies hardware.

1. Apple has been working on Siri, it's AI assistant, for years. Yes, it's terrible.

2. Apple has been integrating AI/ML into it's software and hardware for years.

Perhaps you mean "generative AI"? If so, Tim Cook has made public statements in recent weeks that theirs is coming soon, so probably Summer 2024 or so.

Apple seems to prefer to enter a space late, and then do it well after they've seen what works and doesn't work for others.

They went with a small generative network for autocorrect, but I don't think they've shipped anything else. I mean they have some pretty advanced image editing AI, but that's probably not what you're asking about. LLMs are cool as a tech demo, but what polished actual product could you put them in now?

Local Siri with function calling for dozens of on-device use cases around contacts, phone settings, interpreting web results, timers, calculator, calendar and phone manipulation.

It already has everything it needs - Apple Shortcurs already has tons of triggers and actions.

Siri is embarrassingly bad. Sprinkling magic LLM bits on top of it would make it less bad. That much is clear, even if there are implementation details to get right.

When it comes to AI, Apple have been far behind for last 10 years. Not sure what their strategy was. It seems like they've decided not to compete

Except that they have the best consumer-usable OCR, best consumer-usable speech to text, ...

Some people are building clusters of iPhones just to take advantage of that OCR...

They are doing AI, you just dont know about it yet.

Reply on news.ycombinator.com