- Apple wasn't a pioneer in AI, but it might be winning the AI race anyway.
- Analyst Ben Thompson says that's because of Apple's influence: every AI company wants access to Apple's users.
- That means Apple's own AI efforts are likely to be small-scale.
The tech industry has spent the past few days digesting Apple's big AI announcement, but many questions still remain: First, how big is Apple's foray into AI? Second, how much risk does it entail for Apple? And third, who holds the power here — Apple or its AI partner OpenAI?
When trying to understand big, complex technology questions, I often turn to Ben Thompson, the influential technology analyst who writes the Stratechery newsletter, so I called him up to sit down and discuss this issue.
Thompson's main point is that Apple is using its clout as a phone maker to offload much of the expensive and difficult work to AI companies (in this case, OpenAI), allowing Apple to reap the benefits of AI without bearing many of the costs and risks that other tech companies bear.
But it's worth reading Thompson's comments in more detail. Below are edited excerpts from our conversation.
Peter Kafka: You seemed pretty enthusiastic about Apple's AI efforts even before Monday's announcement, and even more so after the announcement. What was your favorite thing about what Apple showed?
Ben Thompson: I find it very telling how little they actually do.
This moment is about generative AI, and I don't know if people realize that Apple isn't actually doing the generative thing.
It is certainly possible to generate images, but it is quite limited. [primarily] They are using this capability to do interesting things that weren't possible before, like connecting different vertical apps together.
The most compelling demo was the presenter getting a message from his mother about a flight arriving, finding a dinner reservation in email, getting information about real-time flight data, and combining it all. This resonated with me because it's a problem people have.
It really felt like Apple at its finest. We're not selling flashy technology, we're selling thoughtful solutions that are very obvious once you see them..
I think it's worked really well. And it avoids a lot of the challenges with generative AI, which is hallucinations and mistakes. They're basically just handing it over to OpenAI. And it's going to be branded OpenAI. And you [told] You're going to OpenAI. And if that fails? Well, talk to OpenAI.
The reason this is compelling, and why I was optimistic even before the presentation, is because it speaks to their position in the value chain: they own the interfaces that people live through, and that gives them the luxury of solving problems that only they can solve.
And then people who want to spend billions of dollars building these giant language models have an interface to plug in and choose to accept or reject it. This is Apple's way of using its position as a trusted device in people's lives to get everyone to do what it says.
Peter Kafka: You mentioned hallucinations, which people understand comes with AI. The Washington Post asked Tim Cook about it, and he said he couldn't promise it wouldn't happen. But it sounds like you're saying the hallucination problem will come from OpenAI's queries. And it will be an issue for OpenAI.
Ben Thompson: I think the hallucination part is overrated. It's something you can easily get into, like putting glue on pizza. It's very weird and a little embarrassing.
But when you visit ChatGPT.com, you are doing so with the expectation that something strange will happen. And 99% of the time, nothing will.
and [OpenAI] I think it's fair to be very confident and confident that it will continue to improve in that regard, and the fact is that 100 million or more people still use it despite all these risks, which speaks to the enormous utility that exists there.
Apple is doing what it does best
Peter Kafka: Going back to the airport scenario, that's great. But what if I was relying on Apple to tell me the time and destination to pick up my mom, and Apple got it wrong, whether it was a hallucination or a more basic mistake? That's a lot riskier than a run-of-the-mill hallucination.
Ben Thompson: Yes, this is arguably the biggest risk facing Apple.
I think there are two risks facing Apple. One is a massive advance in AI that doesn't result in Apple commoditizing AI, but rather in phones commoditizing it. So who commoditizes who? This is up in the air and outside of Apple's control. But it's worth mentioning.
The real risk is execution risk. Apple had the luxury of being late to the market and benefited from a ton of research and improvements, like making these models smaller and more efficient so they could run on their devices. Apple benefited from all of this.
What they're proposing is to actually integrate different apps and different data, something that nobody has done well yet. Apple's bet is can This works well because the data is stored on the device, but there are real risks to doing so.
There are a lot of things in AI that go through demos and then fall apart in edge cases. There are countless edge cases. So if this doesn't work, if it doesn't work, [the risk] do you have.
Apple is not a follower here. They are a leader. And whether this is a place where they have the skills and capabilities to lead effectively is the biggest open question so far.
Peter Kafka: Do we take it at face value when Apple says that this technology only works on our latest and greatest high-end machines?
Ben Thompson: 100%. The biggest constraint in running these models is memory. All devices that support it have a minimum of 8 gigabytes of RAM. That's the gist of it.
Honestly, this is something Apple was It's slow. The whole process of developing a phone takes years. If I could go back in time, I would have had 8GB of RAM in the base model of the iPhone 15. [so it could run the new AI features].
Will Apple boost iPhone sales even further?
Peter Kafka: Will this boost sales of the device? That's clearly what Wall Street is wondering.
Ben Thompson: I think so. And I think that will not just help drive device sales, but also increase ARPU. [average revenue per user]I'm very curious to see if Apple actually starts talking about RAM, because RAM is a key constraint here — if you get a phone with more RAM, you can theoretically run a better model — will they talk you into actually spending money on it, rather than just upgrading your phone? more, Do you want to get a more premium mobile phone? There is certainly an opportunity to do so.
Peter Kafka: You're left guessing as to whether Apple is paying OpenAI or OpenAI is paying Apple. The arrangement between the two companies hasn't actually been reported.
At the end of the day, does it matter whether Apple writes a billion-dollar check or receives a billion-dollar check?
Ben Thompson: I don't think that's important. I think what's more interesting is what it says about the relative power in this value chain.
The original article assumed that Apple would pay OpenAI. [seeing] How much OpenAI has been weakened. It's basically just a link, and every time you go there you get a warning, Sam Altman wasn't on stage, and it's all Apple stuff. I thought Apple could have done the exact same presentation without OpenAI. If you want to chat, just open the app.
So I think they have more negotiating power, and my guess is that no money is changing hands at all.
(Editor's note: Following our interview with Thompson, Bloomberg's Mark Gurman Neither company is making payments to the other, as Thompson suggested, but both believe they can generate revenue by getting iPhone users to sign up for paid services sold by OpenAI, the report said.
