Apple has struggled with AI for a while now. The company had been working on a revamped Siri under the Apple Intelligence umbrella, which it unveiled at last year's WWDC, but things are not going as planned. Now, a new report from Bloomberg claims that the iPhone maker is considering using outside technology from OpenAI or Anthropic to power its voice assistant.
This is a pretty massive shift for a company obsessed with controlling every last bit of its hardware and software. At this year's WWDC, Apple was reportedly worried about letting people down from an AI standpoint, and given the fact that its in-house models are trailing behind the competition, those concerns seem justified.
According to Bloomberg's Mark Gurman, Apple is in active talks with both OpenAI and Anthropic to potentially use their advanced AI models to power a future version of Siri. The company has even asked them to train special versions of their models that could run on Apple's own private cloud servers for testing.
Internally, there are a couple of key executives you should keep track of: software chief Craig Federighi and Siri head Mike Rockwell, who wants to explore using outside technology to fix Siri's long-standing issues. Rockwell, who previously ran the Vision Pro project, apparently started evaluating third-party models as soon as he took over the Siri team.
On the other side is John Giannandrea, Apple's top AI executive, who disagrees and has championed developing everything in-house. His influence appears to be shrinking as more of his responsibilities, from Siri to robotics, get handed off to Federighi.
After running multiple tests, Rockwell and other leaders concluded that Anthropic's Claude was the most promising for Siri's specific needs. That finding led Apple's head of corporate development, Adrian Perica, to officially start discussions with Anthropic. It is trying to find a way to make Siri competitive with assistants on Android phones, many of which use Google's Gemini AI, without completely giving up on its privacy principles.
This change, however, has unsurprisingly reportedly tanked morale within the internal AI team. According to Bloomberg, the engineers on Apple's foundation model team, some of the most sought-after talent in the field, feel like they are being blamed for the company's AI problems. Some have signaled they might just leave.
And based on recent reports, companies like Meta are ever ready to scoop them up, reportedly throwing around salary packages worth millions to poach top researchers.
As for the talks with Anthropic, it has hit a snag over money. The AI startup is supposedly asking for a multi-billion-dollar annual fee that would increase every year, a price that Bloomberg says Apple is reluctant to pay. This disagreement has left the door open for Apple to move back to OpenAI or another partner if a deal with Anthropic cannot be reached.
1 Comment - Add comment