Who’s building the Apple MM1 ?

On March 18, Apple released an AI model called MM1 to join the LLMs race after canceling its electric car project at the end of February.

It seems that Apple, the once-missing giant, is waking up. But, no leaderboard, no one cares whether it’s 30B or 60B, like a small yawn, it seems to have been overshadowed amidst the AI news storm.

In the paper MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training, the same familiar names appear in the list of authors.

According to the article “Apple Boosts Spending to Develop Conversational AI” by The Information, published on 2023-09-06, Apple had a team of 16 members at that time, including several former Google employees.

After searching these authors on LinkedIn, there are some interesting insights.

  • Of the 30 authors, nearly 2/3 have joined Apple within the last two years.
  • In 2022-2023, major authors emerged from Google and universities. In the old blog ‘ChatGPT Effect: Part 4, Competing for AI Talent‘, Google AI talents were depicted as a hidden weapon flowing to OpenAI and start-ups. However, what went unnoticed was Apple’s involvement in talent competition. Another interesting observation is that none of the authors came from OpenAI, ANTHROP\C, cohere… Why is everyone eyeing Google’s talent?
  • Meanwhile, many of the authors are graduates and interns. Interestingly, the interns all left Apple after completing their projects, with one going to OpenAI. Perhaps hardware products are Apple’s highlight, lacking an AI atmosphere. However, Apple is striving to attract young students and cultivate its own AI talents. On the Apple Machine Learning Research page, there are two programs: Internships in Machine Learning and AI, and Apple Scholars in AIML.
  • The vast majority of First authors, Core authors, and Senior authors joined Apple in 2022. Among them is Zhe Gan, who joined as a First author. Before joining Apple, he served as a Principal Researcher at Microsoft Azure AI, where he focused on building large-scale general-purpose multimodal foundation models. Additionally, in 2022, four former Google Engineers joined Apple – two as Core authors and two as Senior authors. Although unable to discount the contributions of other employees, one cannot help but speculate that Apple’s AI drive largely stems from external personnel.

Hopefully, in the future, there will be more AI talent joining, and Apple will make some breakthroughs in AI. Under Cook’s leadership, Apple’s products have become quite boring.

Data collected from LinkedIn

https://dongou.notion.site/b16a59a6b76e442cb35540ad17535d89?v=2aa02288f90a4101902797d542a28ed1&pvs=4

Tags:
AI APPLE MM1 Talent