Did Apple simply outthink everybody in AI? A chat with tech analyst Ben Thompson

• Apple was not a pioneer in AI. However it could be successful the AI race, anyway.

• Analyst Ben Thompson says that is due to Apple’s leverage: All of the AI firms need entry to Apple’s customers.

• And meaning Apple’s personal AI efforts could be smaller.

The tech world has had a number of days to digest Apple’s big AI announcement, and there are nonetheless a number of questions unanswered. Like, for starters: How large a deal is it for Apple to get into AI? And the way dangerous is it for Apple? And who has the ability right here — Apple, or its AI companion OpenAI?

When I attempt to get my head round large, difficult tech questions, I typically depend on Ben Thompson, the influential tech analyst who writes the Stratechery e-newsletter. So I referred to as him as much as speak it by way of.

Thompson’s principal takeaway is that Apple is utilizing its leverage as a telephone maker to make AI firms — on this case, OpenAI — do a lot of the costly and troublesome work. And that Apple will get the upside of AI with out incurring lots of the prices and dangers different tech gamers are taking up.

However it’s value studying Thompson’s feedback at size. The next is an edited excerpt of our dialog.

You seemed pretty enthusiastic about Apple’s AI efforts before they announced them on Monday. You seem even more so after the fact. What do you want about what Apple confirmed off?

I feel what’s so compelling is definitely how little they’re doing.

This whole second is about generative AI. And I am unsure folks fairly understand that Apple’s not really doing any generation.

They do have picture era, which is pretty tightly constrained. However [primarily] they’re leveraging this functionality to do fascinating issues that weren’t attainable beforehand, like tying collectively completely different vertical apps.

I assumed essentially the most compelling demo was the presenter getting the message from her mother about her aircraft arriving, discovering the dinner reservation that was in an electronic mail, getting details about real-time flight information, and incorporating that every one collectively. It simply resonates as a result of that is an actual drawback that individuals have.

It felt like Apple at its greatest: “We’re not promoting whiz-bang know-how; we’re promoting considerate options which are tremendous apparent when you see them.”

I assumed that actually got here by way of. And it sidesteps a number of the challenges round generative AI. Like hallucinations, like getting stuff fallacious. They’re mainly simply handing that off to OpenAI. And it will be branded OpenAI. And you are going to be [told] you are going to OpenAI. And if it screws up? Nicely, go speak to OpenAI.

The rationale why that is compelling, and why I used to be optimistic even earlier than the presentation, is that it speaks to their place within the worth chain. They personal the interface the place folks conduct their lives. And that provides them the posh of fixing issues solely they will clear up.

After which having an interface for the oldsters that need to spend billions of {dollars} to do these large giant language fashions, to plug in and form of take it or depart it — it is Apple leveraging their place of being the trusted machine in folks’s lives, and getting everybody to bounce to their tune.

You talked about hallucinations, which individuals perceive comes with AI. The Washington Post requested Tim Cook dinner about them, and he stated he could not promise there would not be any. However it sounds such as you’re saying the hallucination issues are going to be from the OpenAI queries. And so they’ll be OpenAI’s issues.

I feel the hallucination bit is possibly overrated. It is one thing that is very straightforward to form of latch on to, like glue on pizza. It’s totally humorous and sort of embarrassing.

However when you go to ChatGPT.com, you go into it figuring out it would provide you with one thing bizarre. And truly, 99% of the time, it does not.

And [OpenAI] may be very assured, and I feel it is cheap to be assured it will proceed to get higher in that regard. And the actual fact of the matter is 100 million folks, or no matter it’s, are nonetheless utilizing it regardless of these risks. Which speaks to the great utility that does exist.

Apple doing what Apple does greatest

However again to the airport state of affairs. That sounds nice, proper? However what if — whether or not it is a hallucination or a extra primary error — I depend on Apple to inform me once I’m selecting Mother up, the place we’re going, and so they get it fallacious? Is not the danger there a lot, a lot greater than a garden-variety hallucination?

Sure. I’d say that is the No. 1 danger dealing with Apple.

I feel there are two dangers Apple faces. One is a few large AI breakthrough, such that the telephone turns into the commodity, versus Apple making the AI the commodity. Like: Who’s commoditizing who? That one is TBD and considerably out of Apple’s management. However it’s value mentioning.

The actual danger is execution danger. Apple does have the posh of coming to market later, and so they benefited from an enormous quantity of analysis and enhancements. Like shrinking down these fashions, giving them excessive efficiencies to allow them to run on-device. They’ve had all these advantages.

What they’re proposing to do — to really orchestrate completely different apps and completely different bits of knowledge — nobody has finished nicely, but. Apple’s guess is that they can do it nicely as a result of they’ve the info, as a result of they’re on the machine. However there’s a actual execution danger.

There’s a lot of AI that demos nicely and dies down on the sting circumstances. And there is a gazillion edge circumstances. So to the extent this doesn’t work out and does go badly, [the risk] is there.

Apple is just not a quick follower right here. They’re a pacesetter. And whether or not this can be a place that they’ve the abilities and capabilities to steer successfully is by far the most important open query.

Do you are taking Apple at face worth once they say this tech only works on our latest, greatest, highest-end machines?

100%. The No. 1 constraint in operating these fashions is reminiscence. And each single machine that helps this has 8 gigabytes of RAM at the least. That is simply the lengthy and in need of it.

I feel that is proof, truthfully, about how Apple was late. To develop telephones, the entire course of runs over a number of years. I feel if they might return in time, they’d have made the bottom iPhone 15 have 8 gigabytes of RAM [so it could run the new AI features].

Will Apple spark extra iPhone gross sales?

Do you think this spurs device sales? That is clearly the query Wall Road’s questioning about.

I feel so. And I feel it does not simply spur machine gross sales. I feel it doubtlessly spurs greater ARPUs [average revenue per user]. I am very curious if Apple’s going to really ever begin speaking about RAM. As a result of RAM is the important thing constraint right here. Should you received a telephone with extra RAM, you can theoretically run a greater mannequin. Are they going to speak about that such that you do not simply improve your telephone however you really spend extra, to get a higher-end telephone? There’s positively the chance to try this.

You have been speculating about whether or not Apple is paying OpenAI, or OpenAI is paying Apple. There is no actual reporting on the market concerning the association between the 2 firms.

Does it matter in the long run whether or not Apple is slicing a examine for a number of billion {dollars} or getting a examine for a number of billion?

I do not assume it issues. I feel it is extra fascinating as to what it says concerning the relative energy on this worth chain.

In my preliminary article, I assumed Apple can be paying OpenAI. After the keynote, and [seeing] the diploma to which OpenAI was diminished — it is only a hyperlink, mainly, and they will warn you each time you go there, and there was no Sam Altman onstage, and this was all Apple stuff — that made me really feel like Apple might have finished the very same presentation with out OpenAI. And simply say, “If you wish to do chatting, open an app.”

Due to that, I feel that they are in a stronger bargaining place. And my guess is there’s simply no cash altering fingers in any respect.

(Editor’s word: Following my interview with Thompson, Mark Gurman at Bloomberg reported that, as Thompson steered, neither firm is paying the opposite, although they each think about the deal might generate income by getting iPhone customers to join paid companies offered by OpenAI.)

Learn the unique article on Business Insider