Apple recruits AIGC talents with an annual salary of one million. The goal: let the iPhone run locally on a large model

Original source: Qubit

Image source: Generated by Unbounded AI‌

Apple has begun to recruit AIGC talents on a large scale.

According to the latest news from the Financial Times, branches in California, Seattle, Paris, and Beijing have released dozens of positions, all of which require large-scale technical personnel.

Among them, the end-side large model is the focus.

According to the recruitment information, Apple wants to compress the large model to the terminal, and in the future, the core products such as iPhone/iPad will directly run on AIGC technology.

At a rough glance, the salary that can be given is basically at least 170,000 US dollars, which is about RMB 1.2 million yuan.

At the same time, we have also noticed that Apple's R&D expenditure this year has increased by US$3 billion (more than 20 billion yuan) compared with the same period last year. Cook revealed that part of it was brought about by AIGC investment.

The industry giant who has been very quiet in this trend is finally going to "branch" this time?

Apple launches AIGC to engage in end-to-end large models

According to statistics from the Financial Times, relevant recruitment information was released from April to July this year.

The main recruiters are Apple's Machine Intelligence and Neural Design (Machine Intelligence Neural Design, MIND, part of Apple's AIML) and other teams.

The recruited talents are related to large models and AIGC, and judging from the information disclosed in the recruitment details, Apple is going to focus on large models on the mobile side, which is different from the current direction of giants such as Google and Microsoft.

As shown below is a senior software engineer recruitment notice issued by Apple on July 15, which requires engineers to "define and help implement the functions of accelerating and compressing large language models (LLM) in Apple's next-generation inference engine", which refers to The most important thing is to be on the mobile terminal instead of the cloud.

Another message, published on July 28, was even clearer, asking for candidates who could bring “state-of-the-art base models to the iPhones in our pockets, enabling the next generation of ML-based experiences in a privacy-preserving way.” ——

Yes, a large part of the reason why Apple wants to focus on the end-to-end model of its core products is for privacy.

This has become a major publicity point for Apple in recent years. Just like the offline version of Siri, running AI software directly without the cloud can not only make the program run faster, but also process user data more securely and privately.

It is worth mentioning that as early as 2020, Apple spent nearly $200 million to acquire a Seattle-based artificial intelligence startup: Xnor, which specializes in running complex machine learning models on mobile devices. It beat products from major manufacturers such as Microsoft, Amazon and Intel.

This undoubtedly provides a lot of basic support for this effort.

In addition to running large models on the mobile terminal, Apple also recruited some talents in this wave of recruitment mainly for the multimedia functions of AI software, including "image/video content generation" and "for visual actions and multi-round interaction".

According to people in the field, Apple's Paris laboratory is more actively recruiting AI talents than other large technology companies, including but not limited to poaching many researchers from Meta.

In fact, compared to other technology giants, Apple has been making quite slow progress in generative AI technology.

During the 2-hour WWDC session in June this year, the keyword was not mentioned even once.

Some analysts questioned Apple's hesitation.

It wasn't until last month that it was lightly revealed that it wanted to engage in AppleGPT.

In last week's conference call, Cook officially talked about this topic.

Apple's third-quarter earnings report showed:

As of now, Apple's research and development expenditure this year has reached 22.61 billion US dollars, an increase of 3.12 billion US dollars over the same period last year.

Cook pointed out at the meeting that the biggest reason for the extra "3 billion yuan" (about 22.4 billion yuan) is actually facilitated by generative AI, and in the next few months, the company will increase investment in this area**.

Generative AI technology is "absolutely critical to us," he said.

Yahoo News said that this is the first time Apple has admitted that it has entered the hot field.

Regarding the cautious attitude shown before, Cook replied:

Apple's consistent practice is to announce products as soon as they are released.

Finally, back to this latest recruitment move, some netizens also expressed high expectations for Apple's plan to deploy large models on the end side:

Edge computing is the future!

I think only Apple has the inference engine and silicon that can do this.

So, what do you think of Apple's plan to stuff big models into iPhones?

Reference link: [1] [2] [3] [4]