Craig Federighi ignited Apple's AI efforts after using Microsoft's Copilot
Despite years of working to develop AI systems, it wasn't until Christmas 2022 when Craig Federighi played with Copilot that the company truly got behind the idea.
Apple's Senior Vice President of Software Engineering, Craig Federighi
The conventional wisdom is that Apple is far behind the rest of the technology industry in its use and deployment of AI. That has seemed nonsensical since Apple has had Siri for almost 15 years, and a head of AI since 2018.
However, a new report from the Wall Street Journal claims that despite all these years working on Machine Learning, and despite having ex-Google AI chief John Giannandrea, it could be true that Apple is substantially behind. Reportedly, Giannandrea and his team have struggled to fit in with Apple, and to get AI implemented.
A disconnect from a highly deadline-focused Apple and the more Google-like relaxed team run by Giannandrea, prevented the company forming a cohesive, coherent AI plan. Giannandrea was also reportedly held back by a lack of resources given to him within Apple, to the extent that his team regularly used Google's cloud services for its work.
Across Apple, different teams also continued working on AI features independently. That included Craig Federighi's team, which according to unnamed former Apple employees, invested in building up the AI behind its image and video capabilities.
But ultimately it was Federighi who has changed Apple's approach to AI. He is said to have spent Christmas 2022 playing with Microsoft Copilot, powered by ChatGPT, and that made him an AI convert.
Subsequently, his team of software engineers were given resources to pursue AI and specifically generative AI. Federighi is reported to have said in Apple meetings that he had now come to appreciate its benefits, and that it will be incorporated into all of Apple's software.
It's not clear whether this work is being done solely by Federighi's team without Giannandrea's. However, the two men were both involved in meeting with OpenAI CEO Sam Altman, which appears to have led to a deal that will be announced at WWDC on June 10, 2024.
Read on AppleInsider
Comments
If so Craig played with it 6 months ago, not a year and a half back, and thus Apple didn't get serious until earlier this year, perhaps the last 4-5 months.
EDIT: Verified that Copilot wasn't announced until March of last year. But apparently Mr. Federighi was playing around with pre-release Github code on the project.
The real competition Apple faces is in „use of AI features“. Foundational models are hyped upon because all of this is new, but in the long run there are not really useful as end user products. There are already lots of very good Open Source foundation models. Llama 3 is very comparable to GPT 4 and there are MANY other options. Finetuning those models for specific purposes and building actual product features around them is the real thing to be done.
its pointless to compete on foundation models for Apple.
Perhaps they are stretched a little thin on the required resources and the strategic decision making process might not be as good as it could be.
Cancelling the car project doesn't mean they couldn't do it, only that they concluded doing it doesn't make sense. It would also seem that autonomous driving is a much harder problem than first imagined.
I think you're right that the strategic decision making process is an issue. I'm not sure it's correct that it could be better, but I think it is correct that it WAS better under Steve Jobs, just because he had an uncanny ability to realize early in a process which ideas were worth pursuing and which were not. I recall him making the point that R&D is only expensive when you spend money on a bad idea. If you only spend money on good ideas, it's not very expensive at all. And so the key is to quickly identify the bad ideas and shut them down. So Apple is missing Steve, but the good news is that nobody else has him, either.
The software and hardware (tools) foundations have been in place for many years and there have been many pieces of software designed using what Apple has already released. The existence of a new button called CoPilot or calling something a AI computer won't affect the back of house infrastructure encoded by Apple the ground has already been laid the PC crowd and some of the Apple crowd just slept thru it because Robbie the Robot didn't hold their hand.
Recall and CoPilot will bore the PC public in a hurry remember the Gamer Boy/Tech/YouTube Crowd attention span is very short they miss the foundational work all the time.
https://en.wikipedia.org/wiki/Apple_A11#Neural_Engine
https://developer.apple.com/machine-learning/core-ml/
https://www.macobserver.com/tips/deep-dive/what-is-apple-neural-engine/
It is/has been normal for Apple and other companies to design build and engineer prototypes that never see the light of day so what? 10 billion dollars is a far cry less than Microsofts 69 billion dollars for a game content company or Googles 12.5 billion dollar debacle acquiring Motorola. Apples largest acquisition on record is only 3 billion dollars.
Jobs made Apple the most valuable company in 2011. Cook has moved Apple down to #3. It has been a decade of iterative upgrades from Apple with AirPods and M-series being the stars.