Apple's Push Into AI Allegedly Happened After Craig Federighi Tried Microsoft AI Coding Tool

[JosephAW]

Someday in the future will be able to ask AI to create a personal iOS branch to fix battery life and broken “features” from the original GM release.

 

[torartc]

Oof that's way later than you'd expect. Not a good sign.

 

[npmacuser5]

I agree with Craig Federighi. Using the Edge Browser to write this with CoPilot as a search options. The scary part comes home real.

 

[contacos]

I am sure important people at the board or shareholders asked them inconvenient questions about their stand on AI and they realized they needed to catch up.

The worst part about all of this is that Apple easily could have been a synonym of „AI“ just like everyone says „iPhone“ not „my Smartphone“ instead of being a very late follower. They introduced Siri on iPhone 4S and did nothing but a UI change with it since!

 

[fakestrawberryflavor]

There's about 18 months between December 2022 and May 2024. M4 was probably well into the design cycle but it shouldn't have been too late to make changes to focus on NPU.

If Apple had stepped on the NPU gas earlier, Apple should be occupying the bright orange bar.

View attachment 2385825

I think Apple could crush all of this pretty easily soon, if they forecasted this well a few years ago (although I'm not so optimistic)

They could have like a 24 core NPU in the M4 Pro and a 32 Core NPU in the M4 Max, and even increase the clock speed of the NPU slightly in those parts. Something like 1.15x the clock speed of the NPU and more cores could equal something that looks like:

M4 Pro: 66 Tops
M4 Max: 88 Tops

I don't think this is going to happen, but it's not out of the realm of possibility given these other chips have much more die area to work with, and much better cooling opportunities, etc.

 

[I_Sheep]

This story is planted to hide the fact that Apple is far behind all other tech companies and they’re worried this will impact their share price

 

[redbeard331]

I don’t believe this either, a company as big as Apple with employees all over the world? Don’t they have their own internal CIA that knows everything?

 

[@Brett]

Craig is probably secretly using a Windows Copilot+ PC.

Faster NPU than M4 at 45 TOPS. Also, it comes with minimum 16GB RAM which likely blew his mind and hair.

Would pretty much be an Apple chip then. Qualcomm bought Nubz’s which is a bunch of ex Apple chip engineers and designers. It seems Qualcomm was not able to do it by itself. I also find it funny MS bragging about the chips when they do not make them.

 

[@Brett]

This story is planted to hide the fact that Apple is far behind all other tech companies and they’re worried this will impact their share price

How are they far behind? Everyone is rushing out half thought out non AI (AI) products and trying to charge people for them. Apple has also been in the space for awhile with machine learning and neural processing.

 

[rp2011]

Craig is probably secretly using a Windows Copilot+ PC.

Faster NPU than M4 at 45 TOPS. Also, it comes with minimum 16GB RAM which likely blew his mind and hair.

Let’s remember that this is the FIRST windows laptop with a NPU, the Mac has had a NPU since the M1 and the iPhone since 2017. And Windows Copilot needs all the ram it can get with all the screenshots it’s taking to..um keep track of the websites, mail and sensitive information it’s collecting on you, I mean for you.

 

[ghanwani]

Interesting. And here I was thinking that it was because they saw how $MSFT was getting pumped and didn't want $AAPL to miss the fun.

 

[betabeta]

It's simply the fact that things are changing so fast, money is pushing tech beyond what we imagined. Crazy but the M4 will look dated by the time it's fully launched. Nvidia has already outdone its own Blackwell before it's fully launched. We have no idea what is going to happen in 5 years, we could plan 5 years out set roadmaps. Unimaginable, but Google or an Apple could be Sears, as in nobody ever thought Sears would go away. Really if you look into AI and robots, it's not like it was even 2 years ago, and experts admit they don't even know how this is possible with LLM's.

I think this is exactly what Craig is seeing, yeah I think Apple was caught off guard, the top experts in AI were caught off guard. So Honestly if Apple didn't collaborate with OpenAI, I think it was lights out. Even now odds are pretty good the entire OS will still need to be overhauled. Apple has a HUGE hill to climb when it comes to people using Macs for work, art or music, because businesses will need very powerful computers or powerful cloud computing, that is why they are moving to change server side chips so fast. Today you have tons of people doing art or music, but people/business will have to train all their back catalog on AI, AI training takes a ton of compute.

Apple along with everyone else was not expecting this, Microsoft is ahead now because of a gamble, a bet, the bet paid off. Now Sam Altman is riding giants, will he tame them and overcome them? Or will the giants throw him off?

I side with Craig, he is 100% right in working with OpenAI, Apple is doing big things with AI, but OpenAI is years ahead so the risk is worth it.

 

[User 6502]

This claim seems a little suspect, at least in terms of the timing. Are they suggesting that Apple designed and built M4 with a heavy AI emphasis, based on stuff that happened in 2022? For a chip shipping in volume in actual products in early 2024 (meaning production of the chip starting in 2023), one would have expected them to have started before 2022.

Nobody is suggesting that. The AI emphasis is of course only marketing quickly put together closer to the announcement and it has little to do with the actual chip capabilities or what it was designed for.

 

[apparatchik]

There's about 18 months between December 2022 and May 2024. M4 was probably well into the design cycle but it shouldn't have been too late to make changes to focus on NPU.

If Apple had stepped on the NPU gas earlier, Apple should be occupying the bright orange bar.

View attachment 2385825

A chart straight from AMD’s marketing department comparing very different thermal envelopes, tiers and even price points.

Apple designs their chipsets according to their needs (and, BTW, they pioneered including neural engines in their SoC starting with the A13), so, for example, none of the chips on that chart could perform as stated inside an iPad, fanless, with the exception of the M4.

 

[Geoffb9]

I kind of believe it.
I’ve been posting here about the sad and pathetic state of Apple’s software development. They’re out of touch and their development pace is pathetically slow.

Did they actually need AI to develop Mail App features. For example. They don’t have Smart Folders on the iPhone/iPad Mail App like the the Mac. Seriously!

 

[heystu]

Siri will also be overhauled with generative AI, making the personal assistant smarter and more capable.

Proof read it for your MR. You’re welcome.

 

[Geoffb9]

It’s sad and pathetic to wait for a VP level executive to make a push for AI. Have the rest of engineering teams been living under a rock? Or is Apple a hugely micromanaged company?

 

[Ethosik]

AI just needs to be stopped. Windows Recall and now Adobe's TOS mess we don't own anything anymore. I am NOT looking forward to WWDC this year, Apple is going to be next to do something horrible.

Its funny how companies suddenly threw out copyright when it fits their agenda, scanning everyone's source code/images/creative works. But if someone uses a Zelda asset in Gary's Mod its an issue.

 

[Geoffb9]

Who else is disappointed in Tim Cook?

It seems his management style is based on fluffy stuff. Just look at who announced the last few years products.

 

[applepotato666]

It’s sad and pathetic to wait for a VP level executive to make a push for AI. Have the rest of engineering teams been living under a rock? Or is Apple a hugely micromanaged company?

I see your point, but the thing is - they’ve always had it and it’s really not hard for a company at that scale to develop generative AI.

Apple is a consumer-oriented “luxury goods” brand with little focus on enterprise. They’ll want to ship features that those everyday people using their products actually care about, perfect the interface around them, not bloat the OS with things nobody uses.

Microsoft and Google are able to go so quickly because they don’t have a consumer and UX focus like Apple does. Apple simply can’t ship Safari that looks like Edge with all its confusing sidebars and 20 ways to do the same thing. It would confuse everybody. Microsoft can do it because Windows is for enterprise and you can ask the IT department.

Apple has been shipping the best and most useful ML around when it comes to day to day intelligence. “Siri Suggestions” is essentially that. Telling me it’s someone’s birthday because they told me when they’re born in a message some time ago and suggesting I give them a call, showing me a playlist I usually listen to at around this time of day or turning on Do not disturb because it knows I’m in a meeting - that’s the type of stuff the iPhone does that people care about, and it was the first to do them (if not still the only, my Samsung does none of that). Execution is the easy part for Apple. Nobody is going to buy an Android over an iPhone because their iPhone can remove subjects from photos. All of this you can do with an app that’s probably already on the App Store. And that’s the stuff Google has been shipping to Android, so it’s not anything that special.

 

[adrianjagielak]

Wasn’t GitHub Copilot X (the one based on GPT) announced in march of 2023?

 

[Wieslawo]

There is no need for the position of Senior Vice President Software Engineering to understand that large language models have a future. Apple has always been in its own bubble, but this bubble was above everyone else and ahead of the competition. Now the bubble is at the end, but the people in the bubble haven't noticed that they've been left behind.

 

[fl010]

I can't wait for the WWDC 24 and how they will try to counter it.

 

[NEPOBABY]

GitHub Co-Pilot has plenty of complaints and of course it's not Swift friendly.

I also have Ollama plugged into Visual Studio Code using the big local coding LLMs. Janky but whatever. I prefer to scratch my head and do things myself. I'm not lazy and I enjoy the pain of making things myself. When I'm old and retired my brain cells will still be healthy size and I won't be living the Brawndo life.

 

[hans1972]

Kind of pathetic. It's like great grandpa being introduced to a microwave while it has already become people's daily tool for a long time.

There weren't many people using the general Copilot at this time since it hadn't been introduced to the general public.
ChatGPT had only been released about 3-4 weeks before.

This was right at the beginning of the public at large being introduced to tools based on LLMs.