Key Takeaways
- Apple is reportedly considering outsourcing AI models for its AI efforts, and that could be the best-case scenario.
- Siri’s flaws make for a poor track record regarding Apple’s past AI features, and make it difficult to place confidence in Apple’s in-house AI models.
- Apple could follow a similar AI strategy to its prior chipmaking strategy by finding a partner initially and eventually shipping custom AI models.
Artificial intelligence is part of the conversation when talking about virtually all parts of the tech industry in 2024. Despite that being the case, Apple has been relatively quiet on the AI/ML front. This is all while competitors like OpenAI, Google, and Microsoft are racing to ship AI products. In fact, although the measure did not pass, 37.5% of Apple shareholders voted in favor of conducting a report concerning the company’s AI efforts. People are concerned that Apple may be asleep at the wheel when it comes to AI, but that isn’t exactly the case. Apple CEO Tim Cook has been steadfast in his claim that the company will ship AI features on the iPhone 16 series later this year. To do so, Apple might follow an old adage: if you can’t beat ’em, join ’em.
Apple will launch generative AI features later this year, but nobody knows what that means
Apple has been known to march to the beat of its own drum, but for once, it’s starting on a backfoot.
Apple is reportedly in talks with companies like Google, OpenAI, and others to supply AI models for mobile devices. While that might sound disappointing to Apple fans who have been waiting for the Cupertino-based giant to dominate with a custom AI model, it could end up being the best-case scenario. Apple hasn’t had the best track record when it comes to AI and ML (I’m looking at you, Siri), and teaming up with a proven company might create a winning solution. Let’s unpack the reports, all the considerations, and why outsourcing AI could be the best option for Apple.
What the rumors say
Apple could turn to Google, OpenAI, or others for AI models
Ever since OpenAI launched ChatGPT, and even before that, people have anticipated what Apple might be working on with AI behind the scenes. When shareholders become impatient, Cook reassured them that AI features are imminent. It seems like the plan is for AI to appear on the best iPhones this year, and could trickle down to the newest iPads and Macs later. Given the timing, it’s reasonable to expect that we’ll hear more about Apple’s AI plans at WWDC 2024. The developer and software-focused event is held in June, so an AI roadmap from Apple is likely only a few months away.
Ahead of WWDC 2024, Bloomberg’s Mark Gurman reported this month that Apple was in talks with third-party companies to supply AI models for the iPhone. One of those companies is Google, whose Gemini Nano model already powers on-device AI features on great Android phones today. However, Gurman notes that Apple also had discussions with OpenAI, which has worked with Microsoft to bring AI to Windows via Copilot. A deal between Apple and Google would not be unprecedented, because Google already pays Apple billions to keep Google Search as Safari’s default search engine. It appears like a deal would be mutually beneficial, because Google would be able to power AI on billions of Apple devices.
Google Gemini: What is it, and how does it work?
Google Gemini is the company’s answer to generative AI, but how does it work and what can you do with it?
There’s another factor to all of this, and that is the regulation surrounding artificial intelligence in specific regions. The Wall Street Journal reported last week that Apple was considering a partnership with Baidu for AI-based services in China. This, combined with Apple’s other discussions with Google and others, could create a scenario where Apple partners with different companies in each region for AI. By doing so, Apple would avoid the headaches that come with making the same product compliant with each region’s individual regulations. As the company faces legal troubles in the U.S. and E.U., it may want to avoid potential issues regarding AI altogether by outsourcing.
Why that would be a good thing
Apple hasn’t proven itself with AI and ML, while others have
There’s a common perception that, as a multi-trillion-dollar company with plenty of resources, Apple is guaranteed to eventually succeed with AI. I completely reject that notion, and past Apple projects are perfect examples as to why. The company just canned the Apple Car project, per Gurman, after spending a decade and billions of dollars on the initiative. Think of what EV startups have done with less resources in the same amount of time. Apple’s track record in consumer technology and massive amount of resources couldn’t make the Apple Car a viable path forward in a decade.
The elephant in the room is obviously Siri, which has been the default voice assistant for Apple products since 2011. It’s powered by voice recognition, machine learning, and — you guessed it — artificial intelligence. In the decade-plus since Siri was acquired by Apple and built into the company’s operating system, it hasn’t gotten noticeably better. Siri is much worse than competitors like Amazon Alexa and Google Assistant, for example. As a longtime iPhone user, I know that using Siri will always take longer than just completing an action manually.
5 things Siri should learn from Copilot
A major Siri overhaul is long overdue
Siri’s shortcomings are deeply tied to its inability to complete on-device requests, which is extremely concerning. The basic things that an assistant should be able to do in 2024 are foreign to Siri. The voice assistant will fail to complete these tasks in one of a few ways. The most common is through a message that asks users to open search results on their iPhone to continue, which defeats the entire purpose of a voice assistant. Siri might also reply with something like “I can’t do that” or “On it … still on it.” Siri is arguably Apple’s most prominent user-facing application of AI, and it’s horrible.
The shortcomings of Siri have been well-documented for years, and they haven’t been addressed. For this reason, it’s hard to have real confidence in Apple’s internal AI efforts. By comparison, companies like Google and OpenAI have proven to be successful in developing AI models. Based on Apple’s track record, outsourcing AI would be a better option than shipping an in-house AI model.
Should Apple keep working on AI in-house?
Apple’s recent work with AI models has been promising, leading up to MM1
That isn’t to say Apple should just give up on in-house AI. The company could follow a similar path for AI as it took with mobile processors. Initially, Apple’s mobile devices were powered by third-party processors. A few years later, Apple started designing custom systems-on-a-chip for iPhones and iPads, starting with the A4 processor. Now, Apple Silicon powers everything from speakers and headphones to the best Macs money can buy. Perhaps a partnership with Google or OpenAI could give Apple time to work on AI models and make sure they’re ready for release, while also allowing Apple to ship new AI features immediately.
Apple’s MM1 language model could finally make Siri useful
Apple’s MM1 language model could be developed in part to make Siri useful as part of the iPhone 16 series.
There are a lot of ways Apple could move forward with AI. Research papers regarding Apple’s new machine learning frameworks and MM1 language model show that the company’s in-house solutions can compete with other options, but MM1 isn’t available yet. To be clear, I wouldn’t be surprised if Apple wowed us all with an AI model that rivals the best ones available. However, based on Apple’s track record, I’d be much more confident in a Google or OpenAI partnership than uncertain in-house models.