Will Siri finally be “Giri”?

 

Multiple sources suggest that Apple is likely in talks with Google to use their foundation model to power the new Siri.

To be honest, it is not a surprise. Before the release of iOS 18, there were already some rumors around this.

Currently, neither Google nor Apple has confirmed this, but if we assume it’s going to happen, there are so many messages we can derive from this, not only what will Siri be in the upcoming update, but more importantly, what kind of company Apple is planning to be.

In this article, I plan to start with the current status of Siri, talk about the problems they’re facing, the impact, and my personal projection on their future.

But before that, I want to discuss a topic that people care the most about.

Will Google’s model save Siri?

None of us has seen the working version of the “smart” Siri yet, and it’s very likely it does not even exist. Without knowing the system’s architecture, it’s very hard to make a yes or no conclusion.

Google has undoubtedly the best resources for AI. It’s reasonable to expect a leap in the performance of the model.

But it does not guarantee Siri will be suddenly better. Even for a smart model, it still has to be well-integrated with the system to work effectively, and this requires deep involvement from Apple itself, who made the OS.

Based on the functionality of the Siri they showed on WWDC 24, there are a lot of technical difficulties to overcome. For example, how can the model index users’ entire context while preserving power efficiency? How can it retrieve the information when needed? How does the model evaluate its moves and prevent unrecoverable mistakes? And whether Apple’s own infrastructure can handle the frequent requests from millions of iPhone users effectively.

Google’s participation will probably remove the bottleneck that was caused by low cloud model performance, but the overall problem that cannot be removed just by a better model remains significant. I would suggest not expecting it to solve anything outright; the final result still mostly depends on Apple.

The current status of Siri

Over the last decade, Siri has been constantly criticized for its limited intelligence and capabilities, especially on social media. Siri is consistently ranked as one of, if not the worst, voice assistant.

Apple has made a promise to make Siri better. In their Apple Intelligence announcement at WWDC 2024, they showed a revamped Siri that was much smarter and more capable than before.

The revamped Siri Apple showed in WWDC24 is capable of retrieving personal context.

A lot of users expected AI to fix the long-existing Siri problem. But Apple unfortunately disappointed them.

Apple ended up failing to deliver that smart Siri in people’s expectations. When the actual release arrived, the only thing Siri was doing better was routing some of the problems it couldn’t answer to ChatGPT, and that’s obviously not enough to invert its reputation.

From my personal experience, I didn’t see any improvement in Siri’s intelligence; it is as bad as if not worse than before.

What is stopping Apple from building its own model?

Apple’s unique business model has several natural disadvantages in the AI-related race.

Apple is a special company in FAANG, with over 50% of its revenue coming from iPhones, which is a single hardware product line. Its services are more concentrated compared to Google.

Unlike others, Apple is very conservative in spending resources on new, unknown fields. You will notice that Apple rarely does foundational AI research; their experience and human resources in AI are mostly in applied ML.

When ChatGPT launched, the AI race began suddenly. Apple was not prepared at all; they historically have had less resources in the AI-related field.

And it’s worse, because Apple seems to have relatively much less interest in the internet compared to others. At the same time, Apple’s products are very privacy-focused. These 2 factors combined mean it has inherently less data available for training at hand.

While people are shocked at Apple trying to use Google’s model, they might not know the current Apple Intelligence model was actually trained on Google TPU.

Google, Microsoft, and Meta already have a strong compute foundation at the first place because of their needs for large-scale cloud services.

In comparison, although Apple has a lot of online services. Those are much less computationally intensive than what Google or Microsoft are doing. Apple’s focus on hardware and local software makes it inherently poorer in server clusters and GPUs.

All core factors for AI model training are missing at Apple. They do have a great excuse for their absence in AI competitions.

Apple’s future

Apple will keep its role as a solution integrator, and it will not turn into a research company just because of its failure on Siri.

Apple is objectively putting more effort into hardware, while losing its focus on software.

The number of visual bugs and glitches has notably increased since iOS 18, and in iOS 26, it’s everywhere.

Today’s Apple can design chips powerful as M5, but at the same time, surprisingly lost the ability to handle the radius of a tiny round corner or alignment of a label.

Apple is no longer a company that pursues perfection like they used to be, and it does make sense. Once the software and hardware get more complex, the difficulty of perfection grows exponentially. The expectation of Apple to be perfect is not realistic anymore.

Apple is a for-profit company; their mission is not spiritual but financial. The first priority for them is to maximize revenue, not respond to wishes.

Just when I wrote this line, Safari crashed, while I’m using the official release of macOS with basically nothing running in the background.

Safari crashed when I was writing the draft of this article.

My conclusion

Using Google’s model doesn’t transform Siri into a Google product. Apple is still the one who determines the final experience. Although this revealed Apple’s potential weakness, their fault-tolerance infrastructure gives them plenty of time to make up for the problem.

It is a mature move, safe and cost-efficient. It also reminds us that Apple has left the Steve Jobs era.

In the foreseeable future, Apple will likely remain at the top of consumer-level electronics. It will still be one of the biggest and most profitable companies in the world.

But we can’t help but ask, 10 years later, when its halo eventually fades, forgotten, will it still be loved by the next generation when they view it objectively? And finally, at that time, will the next “iPhone” still be created by Apple?