Report: Apple wanted to provide AI basic models source open

In the meantime, Apple is said to have planned to switch on a gear with the help of external developers at Apple Intelligence: The group pretended to release parts of its basic models as open source this year. According to one Report of the Silicon-Valley specialist sheet the information The group also wanted to do this for image reasons-to show the world that the iPhone manufacturer has made progress with regard to Large Language Models (LLMS).

However, this would apparently also have to admit performance losses that happen as soon as the models no longer run on Apple’s private cloud computers (PCC), but locally on the device. For these and other reasons, software chief Craig Federighi is said to have prevented the idea.

Models are considered worse than in the competition

Apple’s AI models are considered to be inferior to the competition. The company cleared this in its latest “Tech Report” to Apple Intelligence with some benchmark results. In some categories you reach similar values GPT-4O, but then the results sink again. Local models work poorly due to the lower available computing power – Apple is more than other manufacturers on this approach because it has data protection advantages. Most recently, Apple also lost important employees from his AI model department at Metas “Superintelligence” team, which attracts with enormous salaries (or share allocations).

The open sourcing of important models was considered a possible rescue anchor. But Federighi, Apple’s most powerful software man, thought that was problematic. In an email to Ruoming Pang, which now works at Meta, he wrote that there are already “many open source models from other companies” that have already allowed research.

In addition to the fact that Apple’s basic models run worse than that of Alibaba or Google, Federighi is said to have feared that there is criticism of the local models. Source -open foundation models may have shown that Apple “has too many compromises,” it said.

Federighi has to hold the team together

In Apple’s AI team, these have apparently been controversial for a long time. The idea of the “On-Device First” holds back development at the group, says individual team members. Even Federighi seems to have doubts whether this approach works in the long run.

At the moment, however, he has completely different problems: he basically has to hold the Apple Foundation model team together after there were said departures. Apple therefore wants to pay AI researcher more money, the company does not come-but the huge amounts that Meta wants to spend, for example, does not come. Meanwhile, the users are still waiting for a finally improved Siri.