Apple’s Approach to AI Should Rewrite the Industry’s Expectations for the Technology
Apple’s recently-announced but belated embrace of AI is being called “boring” by some fans of ChatGPT. My perspective is somewhat different — because AI helped me learn how to bake pie.
Last Thanksgiving, I was looking forward to taking a break from my computer and being present with family while making a pumpkin pie. But I had questions about preparing it, so I turned to the web. As anyone who’s searched for a recipe online knows, I was instead overwhelmed by sites full of pop-up ads and long irrelevant stories by the recipe website’s author.
Frustrated, I opened up ChatGPT and asked it how to bake a pie from scratch.
Within seconds, clear, concise instructions appeared. (And no “add glue” hallucinations — that’s a Google thing.) The final output was delicious. And because I didn’t have to trawl the web for instructions, I could focus on cooking while chatting with family; ironically, ChatGPT brought a more human experience back to making pie.
Apple’s approach to AI is pretty much like that, translated across its many apps and devices, and roughly follows the Calm Tech philosophy: Fitting the tool to the task, rather than the other way around.
Apple design philosophy is not only valuable for consumers, but will hopefully re-shape how we collectively understand AI — a topic that’s all too often been dominated by bold prophecies of a future completely transformed by artificial intelligence.
Here’s what I mean:
Cognitive Load & the Limits of ChatGPT
Excitement over ChatGPT, coupled to its deceptively open-ended chat interface — implying it can answer any prompt equally well — has led to numerous applications that actually require more cognitive burden, not less. Someone might use ChatGPT to help write a report but then spend so much time fact-checking the output, it becomes more burdensome than helpful.
A lot of disappointment with artificial intelligence stems from people using GPT for tasks it’s not ideal for, getting mediocre or incorrect results, then negatively judging all of AI. Generating unique essays on a topic where the prompter has no original thesis, or telling GPT to design a website without the prompter understanding the fundamentals of web design, are examples of that. The general GTP interface is like giving someone a Swiss army knife when a screwdriver will do. More complex tasks, like asking GPT to check a legal paper, might require that the user is a lawyer qualified to review the results.
AI applications are good when they reduce our cognitive burden; they’re not good when their use only increases our mental load. But there are many many examples where using GPT has less cognitive burden
- Cooking recipes: As eaters of my Thanksgiving pumpkin pie might attest, ChatGPT shines in comparison to recipe websites bloated with ads, personal stories, and other chafe. It’s not perfect, but it reduces the amount of web browsing time and attention required to complete that task.
- Technical issues and code reviews: Having to go to websites to trawl through years of outdated support issues for specific code or technical issues takes time and a lot of mental energy. GPT can help by surfacing possible answers, reducing cognitive burden before having to hire a specialist.
- Possible interpretations and translations of text: I often need help communicating with people who are neurodivergent — even as someone who is neurodivergent myself. Sometimes I get so confused when someone writes an email or text, I run it through GPT to help understand possibilities as to what it might mean.
Use cases like these may not seem groundbreaking or revolutionary, but they’re quite valuable to many people — and help explain how Apple designed its approach to integrating GPT.
How Apple AI Reduces Cognitive Burden
Apple’s belated integration of AI with its devices has been criticized for being slow off the mark, but the company is smart to roll it out incrementally over time. A lot of thought has clearly gone into what GPT is good at, and what it is not, while giving people more affordances within iOS apps they already use.
Essentially, Apple is taking a Calm Tech-like approach to making AI solutions that are genuinely helpful: Compressing action and bringing tools where people already are in specific apps, reducing cognitive overload while using them.
For instance:
- Siri: Apple’s voice assistant is super limiting and kind of dangerous when you try to rely on it to perform tasks while driving, but being able to have GPT help you while on the road — for instance, opening otter.aa to take notes, recording and summarizing phone calls — is very helpful.
- Notes: Currently, this much-used app is a basic notetaking program — that will soon get a powerful upgrade with GPT. Enabling people to search and bring key content into Notes while within individual notes — instead of having to exit out and search, and copy/paste across multiple apps — is incredibly useful. It also makes for a far less attention sink of an experience, and far more likely that people can keep their attention span while working on a project.
- Text/Email: Writing Tools enable re-writing for tone, coherence and readability. For those of us whom English isn’t their first language, or have dyslexia, neurodivergence, and other challenges, GPT is transformative. I already know artists who already use ChatGPT for grant writing, when they don’t have the emotional bandwidth to dry out their art and soul into flattened text.
Some critics have called Apple’s AI announcement “meh”, as if they were expecting Siri to become Her of Her. But that reaction totally misses the purpose of “artificial intelligence” at its best: Amplifying what’s best about AI and humanity working in tandem — and not against each other.
Hopefully the AI industry as a whole will follow Apple’s lead. If it does, we’ll see more AI applications designed for our specific, immediate needs — and less grandiose projects promising to transform all of humanity. Few of us are prepared or even interested in a revolution on that scale — while most of us really would prefer better instructions for making pumpkin pie.
—
Amber Case is the founder of the Calm Tech Institute, a design/research firm dedicated to buildings frameworks and standards for products that make better use of attention. Previously a fellow at MIT’s Center for Civic Media and Harvard’s Berkman Klein Center for Internet & Society, they are also author of Calm Technology: Principles and Patterns for Non-Intrusive Design.
Follow our Calm Tech Institute newsletter for more updates!
