Apple’s weak AI efforts were highlighted this week with an announcement that some Apple Intelligence improvements to Siri are being delayed until 2026. This failure to deliver on features presented in the WWDC 2024 keynote has seen noted Apple commentator John Gruber serve a blistering takedown of Apple’s efforts to date.
He’s not the only one.
As the competition roars off into the AI-powered distance, what challenges is Apple facing, and what options are open to it?
Apple Intelligence’s Catch-22 Challenge
First, let us acknowledge that artificial intelligence—specifically generative AI—is not just a buzzword used to sell the latest technology but a technique that has imbued itself into the landscape. Sometimes, it is front and centre, such as Microsoft’s Copilot and Google’s Gemini. AI is more subtle but clear in other areas, such as backfilling images and crops in Photoshop. In more areas, it is almost invisible yet accepted; the Grammarly tool is a good example.
If AI is present across multiple ecosystems in countless modes, not having proficient AI is a fundamental weakness. Apple is already being left behind; without finding proficiency in AI, it may never catch up.
Then you have the cornerstone of the majority of AI tools in use today… the large data sets used to train the various models so they can offer sufficiently robust outputs that meet expectations. Companies that have been gathering data on their user base (for whatever reason) have ready access to these large data sets.
Apple has long been a vocal proponent of the sanctity of its users’ personal data. It keeps as much of it as possible on a user’s handset, encrypts data it draws into the cloud so it is out of reach of even its own staff, and will fight to preserve those principles in courts against National Governments if necessary. Altering that principle would rock Apple’s ecosystem to its core.
Apple Intelligence’s Neverending Loop
Apple lives in a world where AI is going to be table stakes; AI needs large training sets that Apple does not have access to; Apple won’t use user data to develop and train AI; and Apple lives in a world where AI is going to be table stakes in any software or hardware moving forwards.
How Can Apple Intelligence Break The AI Loop
What options are available to Apple?
It’s too late to do something radical such as dropping its efforts to implement AI and use its marketing power to highlight issues such as the rapacious need to gather training data, the personal information that would need to be traded, the hige environmental result, and the potential inaccuracy of the results. That could have been a great sell two or three years ago, potentially changing the discussions around smartphone AI.
Make that move now, and it would be written off as “You can’t do it, so you’re trying to change the conversation; it won’t work.”
Tim Cook and his team could work on a different approach, putting AI into the handset with smaller LLMs. Running exclusively locally would ensure data remains private, and a smaller environment should be easier to code for. It’s a model that many Android smartphones are using to great benefit. But those Android devices will quickly turn to an army of servers in the sky to access cloud-based AI models for more accurate results.
The final answer would be to outsource the AI efforts and work to bring the same AI benefits that Android can offer on mobile and both ChromeOS and Windows offer on more traditional computers. It has started down that route, working with OpenAI to boost Siri’s results, but the answers so far don’t feel up to the typical Apple standard.
Apple Intelligence Reaps The Seeds Of Yesteryear
Apple’s decision to focus on privacy has left it ill-suited to the rise of AI. The path they have taken the ecosystem down differs from the competition, and the divergence is growing.
Put simply, Apple needs to do this big data thing with no data. If you believe the legends, Apple can deliver AI in a way that only Apple can. Frankly, if it can do the big thing with nothing, it will have earned that legendary status.
Now read more about the hardware improvements Apple was bounced into to support Apple Intelligence…
Read the full article here
