Distribution Runs Thicker...
Salvador Rodriguez, Aaron Tilley, & Miles Kruppa
Has AI frozen the gates of hell?
In its hustle to catch up on AI, Apple has been talking with a longtime rival: Meta.
Facebook’s parent has held discussions with Apple about integrating Meta Platforms’ generative AI model into Apple Intelligence, the recently announced AI system for iPhones and other devices, according to people familiar with the matter.
To me, this has little to do with the "hustle to catch up" and more to do with how Apple has chosen to do AI. That is, using their own models (both on-device and in-cloud) to handle the tasks they believe most of their users will care about and actually use and to outsource the rest – namely, at first, the chatbots. ChatGPT is the most prominent of these, of course. And as such, OpenAI is the first partner Apple has announced. But the company also went out of their way during WWDC to make it clear that they were just the first. Others would be coming.
Google's Gemini, Anthropic's Claude, and Perplexity have been the other models that have been most talked about/rumored to be coming next, but given the rumored "deal" for these services – that is, largely no deal, at least not a monetary one – it shouldn't be surprising to see an option for almost all of the third-party AI offerings to show up here. That includes Meta, despite the history between the two companies. If Apple thought Meta AI was more unsafe or more unreliable than the others, they may not do it, but there's nothing to indicate that right now.
What about from Meta's perspective?
Meta and other companies developing generative AI are hoping to take advantage of Apple’s massive distribution through its iPhones—similar to what Apple offers with its App Store on the iPhone.
Meta clearly wants their own AI products everywhere. And Apple is providing a potentially huge canvas here. Also, this is not to be confused with the Llama models, on top of which Meta's AI is built, of course, but is an open source LLM.
In its talks with other AI companies, Apple hasn’t sought for either party to pay the other, the people said. Instead, the AI companies can sell premium subscriptions to their services through Apple Intelligence. As it does on its app store, the iPhone maker would keep a cut of subscription revenue from its devices.
This is the first report I've seen which would seem to confirm that like the deal with OpenAI, Apple won't seek direct monetary arrangements. At least not right now. Again, this points to why it makes sense for all parties to get involved here.1
One more thing: is the numbers-pulled-from-thin-air department:
While ChatGPT usage is expected to double with the Apple partnership, OpenAI’s infrastructure costs are expected to grow 30% to 40%, said Gene Munster, a longtime Apple analyst and managing partner at Deepwater Asset Management. Munster expects 10% to 20% of Apple users will opt into paying for a premium AI subscription to a product like ChatGPT. That could mean billions of dollars for AI companies that integrate successfully with Apple’s new platform.
Double? I mean, maybe. But who knows. Certainly usage will rise, but double feels like a lot. It also just feels like these are complete and utter guesses. But I do believe rising infrastructure costs may lead to some sort of alteration on the Apple/OpenAI deal, since it's Microsoft infrastructure, of course.
Update June 24, 2024:
Mark Gurman & Kurt Wagner
While Meta and Apple may have had "preliminary" talks about integrating Meta's AI technology as an offering in 'Apple Intelligence', Mark Gurman and Kurt Wagner report:
Apple Inc. rejected overtures by Meta Platforms Inc. to integrate the social networking company’s AI chatbot into the iPhone months ago, according to people with knowledge of the matter.
The two companies aren’t in discussions about using Meta’s Llama chatbot in an AI partnership and only held brief talks in March, said the people, who asked not to be identified because the situation is private. The dialogue about a partnership didn’t reach any formal stage, and Apple has no active plans to integrate Llama.
This report is slightly confusing as it seems to be conflating 'Llama', which is the name Meta gives to their models, which they also open-source, and 'Meta AI', which is the name they give to their actual AI products, which are not open-sourced. Still, the reporting notes that:
Apple decided not to move forward with formal Meta discussions in part because it doesn’t see that company’s privacy practices as stringent enough, according to the people. Apple has spent years criticizing Meta’s technology, and integrating Llama into the iPhone would have been a stark about-face.
This would seem to be in line with my caveat above:
If Apple thought Meta AI was more unsafe or more unreliable than the others, they may not do it, but there's nothing to indicate that right now.
Now there is something to indicate that, apparently. That, or just the ongoing bad-blood between the two. While even Apple and Google partner on things – notably search, but soon, it seems, Gemini as a part of Apple Intelligence as well, Apple and Meta are at odds perhaps more than any two big tech companies right now.
Still, that aside, I still would not be shocked to see Meta AI as an option if it proved to be popular with users (and yes, Apple determined it could be trusted) because of the system Apple has built here. May the best chatbot win.
1 Though I imagine we won't be seeing Microsoft's Copilot here -- though perhaps mainly because it's already built on OpenAI's tech.