Apple Has Helped Pop The AI Bubble

You’ll never pay for third-parties when the new Siri is this good.

“Pioneers get the arrows, Settlers get the land.”- Matt Silverman commenting on my previous Apple AI article

Apple is a company of almost infinite patience. Letting its rivals run ahead by almost years at a time lets them better survey the consumer technology landscape and see which way the winds eventually blow. This cautious approach doesn’t earn them many fans in the influencer mindspace but it does mean that when they launch a product you can be sure it has been thoroughly well-designed. This happened with the Laptop. It happened with the Phone. And its happening right now with Artificial Intelligence.

The invention of the Large Language Model, your Claude or ChatGPT was a natural evolution of the machine learning algorithms that had dominated the 2010s. Perhaps their greatest innovation has been the Natural Language Interface or NLI, a revolutionary step forward in computing UX. Rather then needing to know jargon or revise complex commands, you can just ask at automonous AI ‘Agent’ to do something in plain language and it will obey in kind. The rest though has had people scratching their heads.

The problem AI companies have been trying to solve is largely one of their own making. It costs obscene amounts of money to keep a server-based LLM running and at the same time nobody really asked for them to be built. The likes of OpenAI, Anthropic and Google have been searching for a way to make the general public reliant on their applications, transforming them from curious users to paying customers as they go. Their CEOs have been involved in a circular hype cycle where they make lots of promises, bring in the funds, buy lots of chips from companies who then use their profits to buy equity in the AI products. This pumps up both their share prices as they go and thus the ‘AI Bubble’ was inflated.

Largely under pressure from shareholders sold on the hype the bubble brought with it, Apple decided to go against its patient nature and rush a series of products to market to try and compete. Gathered under the Apple Intelligence umbrella brand they overpromised and underdelivered and so, licking their wounds, Apple went back to R&D. Realising the data harvesting required to train an LLM was beyond their principled stance on privacy, they instead licenced a copy of Gemini from Google to the tune of $1bn a year to train their own models.

If you’re unwilling to get your hands dirty you can always pay somebody else to do it for you, right?

Launched in beta form alongside the bug-cleanup job of OS27, Apple’s new Siri AI set out to correct the mistakes of Apple Intelligence and by all accounts it has succeeded. Siri now provides the sort of in-depth answers that users expect from modern AI products combined with the now standard Apple-levels of privacy nobody else offers. Anything incapable of being processed on-device is sent to Apple’s Private Cloud Compute (PCC) servers. PCC keeps no online record of user requests and are not used to help train further models. Because Siri has private access to your iCloud data, you can talk to it as a NLI (with slang and colloquialisms along the way) about your day or string together commands and it will respond to in kind.

You’d never get these sorts of responses from the old Siri (Images by Author)

Unlike the rambling answers of competing platforms that are designed to promote engagement, Siri is clear and concise. Having secure, private access to your local data gives it a contextual awareness rival platforms lack. Apple’s hasn’t reinvented the wheel here. The new Siri doesn’t ‘do’ anything that Gemini doesn’t already on Android. But its principled stance on privacy has been flipped into an advantage that none of their rivals can offer: confidence. People are quite rightly fearful of the Mother Brain LLMs infiltrating their workplace, effecting their jobs and using their data for who knows what. Siri isn’t here to replace any humans but to make the things your iPhone already does easier to access.

The new Siri isn’t launching in the EU for this very reason. The DMA wants AI choice to be a level playing field yet Apple doesn’t want to grant rival platforms such deep platform integration. If users want a rival AI though they’ll just buy a rival device, the duopoly not extending to LLMs.

As any small child will tell you, bubbles cannot inflate forever. For a kid playing with dish detergent the results of a finger prod are instantaneous. Tech companies reliant on a hype cycle things can be less obvious, with large amounts of money involved often masking any problems. For example a now infamous LLM product from a social media company became so costly they had to merge with a sister corporation. They later floated on the stock exchange to finance a stealth bailout because doing so publicly would have tanked their share price. During a recent Gemini keynote the presenters were having to will the audience to applaud, the developers in attendence largely disinterested. Early corporate adopters are now questioning their decisions and stock values have fallen accordingly. These are all signs that the market is starting to deflate.

The new Siri did not have any immediate effect on AI stocks and actually lowered Apple’s own share price! This was based on the feeling of Apple playing catchup and newer features requiring expensive memory chips (themselves overinflated due to AI buyups) that will eat into margins. Two weeks on from its launch however and an industry-wide stock selloff has begun with shares in Google dropping by as much as 5%. It doesn’t sound like much but when your company is valued in the $$$ trillions it is a massive chunk. Due to many complex reasons, investors are losing confidence in the overspending and overextension of an AI market dangerously addicted to debt.

Where Apple now fits in is perhaps the most important part. For AI companies to continue their meteoric rise in value they need to convince investors they are the future. Its all very well getting (now cautious) corporate customers on board but they have always needed to turn the general public into subscribers. Now that the new Siri offers an NLI for industry-standard Apple products, parity with other platforms and deep personal context they may never get there. All NLI interactions with Siri are free with your iPhone with only repeated image edits requiring payment. For the average Apple customer, subscribing to a competing platform would be a waste of money and there are a lot of Apple customers.

Apple has not outright popped the AI Bubble as some users might like, but they have definitely sucked some of the air out of it. LLM products are likely here to stay with us, but perhaps not in the ways their owners might like. By concentrating on enhancing the operating system rather than replacing the apps, the new Siri offers the average user all they might need in a private, secure wrapper. And may yet give skeptics like me a reason for AI to stick around.