To most, Apple’s lack of artificial intelligence (AI) in its products has been a glaring oversight.
Many have argued that Apple is far behind on employing AI.
None of this comes as a surprise.
After all, Siri — Apple’s voice assistant — has limited utility. It’s almost hard to believe that it was launched in 2011.
It’s even harder to believe how little progress has been made since then. While Siri can answer simple questions that can be queried easily on the internet, the typical response is to send back a list of web search results to choose from. Unimpressive.
But what we see at on the surface is different from what’s happening behind the scenes.
Apple’s approach to implementing these newfound “powers” of generative AI is no different than its normal product strategies.
Take 5G wireless, for example. Apple’s first 5G-enabled iPhone was released in the fall of 2020 with the iPhone 12.
But Samsung released the 5G-enabled Galaxy S10 in most Western markets in the spring of 2019. It looked like it had almost an 18-month lead.
And 5G-enabled phones in certain markets in Asia were available even sooner than that.
Was Apple late? It certainly wasn’t the first to market.
But also, did it matter?
It didn’t.
In 2019, 5G network coverage was very limited. It was still early in the 5G infrastructure build out phase. Galaxy S10 owners may have had a 5G-enabled phone, but so what? They were almost entirely running on 4G networks due to the spotty coverage of 5G, since 5G-enabled phones also work on 4G networks.
Apple intentionally waited until 5G network infrastructure coverage expanded enough for consumers to experience 5G speeds on a more regular basis.
I could even make an argument that Apple could have waited another year to launch its first 5G phone. Even by 2020, most 5G networks were operating over radio frequency spectrum that resulted in 4G-like speeds.
Apple understands that most consumers won’t understand the differences in radio frequency spectrum, the status of 5G network build out, or the difference between Verizon’s, AT&T’s, and T-Mobile’s approach to 5G networks. That’s why it waits.
Apple’s modus operandi has been the same: It wants to maximize the likelihood of a consistent and fantastic consumer experience with its products.
That’s why Apple tends to be 18-24 months late incorporating new technologies into its mobile devices. The same was true for 3D-sensing technology used for augmented reality apps, and near field communications (NFC) technology, which is used for contactless payments…
The same is true for a technology like generative AI.
It’s easy for us to be anxious and excited about applying a technology like generative AI to a great consumer product like an iPhone. But considering that the first version of OpenAI’s ChatGPT was released in December of 2022, it really hasn’t been that long for the technology to mature.
And as we learned in Outer Limits — Google’s Dystopia, where we saw how “buggy” and biased Google’s Gemini was, it was/is clearly not ready for widespread deployment.
But all of this doesn’t mean that Apple hasn’t been busy in the laboratory working on generative AI.
Quite the opposite, in fact.
Apple develops and iterates with technology for years before commercializing.
And we just got a glance of its developments with generative AI.
Last month, a team at Apple published research on MM1, its naming for multi-modal large language models.
Multi-modal models are all the rage in the last few months. They enable an AI to ingest different kinds of inputs — like voice, video, text, and images — synthesize that information and infer the correct output… kind of like us humans.
Here is a simple example from Apple’s research, whereby the AI is given two related images and asked to both infer the correct answer and explain why it came to that conclusion.
The left image is a table with bottles of beer, and the right image is a menu with pricing for four different kinds of beers.
The images may be small, but the beer on the table is “Magna,” which costs $6 each.
It’s interesting to see Apple’s MM1 performance… as compared to two other multi-modal AI models.
Only MM1 figured out the correct cost for the table of beers — $12, as there are two Magna beers on the table costing $6 a piece, according to the menu. The other two models either couldn’t properly “understand” the menu or kind of beer on the table… or made judgment and reasoning errors, leading to two incorrect bills — $15.99 and $44.
Either way, it’s clear that Apple has been making far more progress than most had assumed.
This explains some major changes recently announced by Apple.
In late February, Apple announced that it was shutting down Project Titan, its autonomous electric car project.
It was a large project with about 1,400 employees that had been under development since 2014.
The press positioned this as a failure. But I see something different…
Apple will be shifting a large group of that team over to generative AI projects at Apple. This tells us that Apple had a resource problem that is not limited to Apple.
Specifically, there is a massive shortage of AI engineers everywhere.
Elon Musk was recently quoted, saying “it’s the craziest talent war I’ve ever seen.”
Apple CEO Tim Cook was faced with the stark reality that he needed to make a choice. Despite having $73 billion in cash on hand, Apple didn’t have the AI engineers to execute on Project Titan AND its plans for empowering its consumer products with generative AI technology.
For Apple, this was the right choice.
The return on invested capital for autonomous EVs is much farther out compared to consumer devices. The capital expenditures are massive, and profitability only comes after years of building scale.
With consumer devices like iPhones, iPads, Apple Watch, MacBooks, etc., imbuing generative AI can be as simple as a software download. The high gross margin returns will come quickly.
Given the release of the MM1 research, which is promising, I’d be willing to bet that we’ll see some major announcements from Apple at its Worldwide Developers Conference June 10-14th concerning generative AI, as it pertains to iOS18… which will be released this fall with the forthcoming iPhone16.
I’m confident the announcements won’t just be limited to the iPhone, as well.
Making matters even more interesting was recent news that Apple would be using Google’s deeply faulted Gemini on the iPhone.
Given what we’ve already seen from Apple’s MM1, that doesn’t make a whole lot of sense.
What would make sense is if Apple allows for an improved version of Gemini to be used for web search, not generative AI applications.
After all, Google pays Apple about $18 billion a year to ensure that Google is the default search engine on iPhones. As Google incorporates AI into its search engine, it makes sense that would extend to the Apple deal.
By this fall, the tech industry will be almost two years into this latest movement in generative AI, specifically large language models (LLM). Given the improvements seen since December 2022, the timing feels right for Apple to push the tech into its products.
And that means that Siri, finally, is about to get a life-changing makeover.
If it were me, I’d make sure Siri was positioned as the default personalized digital AI assistant for all iPhone users.
And akin to Tesla, I don’t think that Apple’s decade-long research and development on autonomous technology will go to waste at all.
A few days ago, it was reported that Apple is developing home robotics technology as potentially a new product line. A mobile personal robot would utilize the autonomous software in the same way that Tesla’s Optimus leverages Tesla’s full self-driving software for its cars, trucks, and semis.
Apple’s (AAPL) stock has pulled back about 15% since its December highs. Long-term, I still see plenty of growth ahead.
We always welcome your feedback. We read every email and address the most common comments and questions in the Friday AMA. Please write to us here.
