There is still some way to go before there’s clarity about how exactly Apple plans to roll out Apple Intelligence, once iOS 18, iPadOS 18 and macOS Sequoia roll out in what I expect would be late September (it’s usually after the iPhone keynote, in time for the availability of the new iPhones). For starters, it may not be a one-shot rollout with the new operating systems, with suggestions that availability may be enabled in batches as part of an extended test period. I find that rather unlikely – it is not Apple’s way of having users sign up for a waiting list for software functionality. There however might be chapters for Apple Intelligence, Siri improvements and the ChatGPT integration. Availability is something Apple hasn’t spoken about yet (all you read about on the internet is pure speculation, packaged with seriousness to sound well, serious). For me, Apple’s biggest challenge isn’t acceptability. It will be to ensure the habitually unpredictable generative AI models don’t go haywire.
(Premium): Tech Tonic | If Apple Intelligence doesn’t annoy anyone, that’ll be a big win
The series of conversations I had with Apple executives on the side lines of the WWDC keynote made one thing clear – Apple on-device and Apple server models are the tech giant’s in-house efforts. Not OpenAI’s GPT-4o, which in fact is just one parallel, optional element. In my understanding, that doesn’t impact how Apple’s own AI models will work, learn and respond. If there are instanced when your queries may require some specialised learning for more detailed responses, you’ll be prompted to check if GPT may be invoked. You’ll have control over whether to say yes or no. That said, quote a few things for Apple to be worried about, considering its dealing with something that’s fickle and springs surprises. Here’s a checklist.
- Generative AI in particular, only really works if it has a lot of data to work with. Even then, there’s no guarantee it will be error-free – something we’ve seen time and again.
- The tightrope Apple is walking with AI involves success with the complexity of tasks they promise, layered context to hold, and personalisation without compromising data privacy, whilst minimising hallucinations, false positives and generating content that may be offensive to human beings.
- Apple executives made clear to me that Apple will not use user data to train models, but training and context will be built with broader activity trends. One of the ways to do that is to try and compute as much as possible on the device.
- No user query will be retailed by Apple’s models. So much so, that they’ve got OpenAI to toe the line for the GPT-4o integration for the optional call-up by users for certain tasks. To be specific, OpenAI confirms that when accessing ChatGPT within Siri and Writing Tools, requests will not be stored by OpenAI, as well as users’ IP addresses will be obscured.
Quite how this turns out, time shall testify to it. This does give Apple its biggest challenge in years, with the unpredictable generative AI to tame. This isn’t just a chatbot. Apple Intelligence’s systemwide integration magnifies a user’s touchpoints with AI. No pressure, then!
Analysing Apple’s AI chapter, as it is being written…
Deep dive into Apple Intelligence models, optional ChatGPT and privacy approach
Beyond Apple Intelligence, iOS 18 is changing and so are iPads and Macs
Point to note: Apple executives confirmed to me that the company’s agreement with OpenAI for GPT-4o as one part of Apple Intelligence is limited to your interactions with ChatGPT as part of Siri. If however, you sign in to your OpenAI account with a paid subscription for ChatGPT Plus, you’ll be out of the purview of that agreement which otherwise ensures OpenAI will not store your data and hide IP addresses so that no usage can be tracked back to users.
SECURE?
Not so much, if you’re trusting one of the many free Android VPN apps. As Simon Migliano, who is Head of Research at Top10VPN.com explained to me, the situation is worrying. Here’s a snapshot – as many as 88% of these free VPNs suffer some kind of data leak (IP, DNS, WebRTC; of this, around 17% are affected by multiple leaks), around 71% apps share your personal data with third parties including Facebook or Yandex and controversial data brokers like Kochava, as many as 15% contain Bytedance tracking code, while 53% contained at least one function in their own source code that posed a potential risk to user privacy and requested permission to run it. I’d urge you to read his detailed study here – you’ll know the worst offenders, and it ought to make you take corrective measures. This is quite worrying, since VPNs tend to otherwise cost a fair penny – and users by default tend to look at the more cost-effective alternatives (read, free). Think about it, they’re free because they can sell your data to the highest bidder.
MEETINGS
One (among the many) lingering follow-through from the pandemic era, are virtual meetings. Mind you, that forgettable time in our lives did teach is a valuable lesson – work is very much possible without having to be in office every day; remote work, if done diligently by genuine human beings, actually results in increased productivity. You’re wasting less time in commute, water cooler conversations and (applicable for some) extended lunch. Virtual work’s success rests on the successful implementation of virtual meetings. Thank you, Zoom, Microsoft Teams, Cisco Webex, Google Meet and everyone else that pitched in with an app. But over time, we’ve probably begun to do too much of a good thing. And that’s a problem, something I’ve illustrated in my latest piece for HT Wknd. Virtual meeting fatigue is becoming a real thing, and I’ve cited multiple studies which tend to illustrate how bad things can get if you’re spending too much time in virtual discussions. Perhaps AI can provide an answer, as I’ve pointed out – Zoom’s AI companion summarising calls and Microsoft Copilot transcribing Teams calls, can help. But do enough people know enough about these tools at their disposal? Think about it.
(Premium): Zooming in…and zoning out: Can AI help with video call fatigue?
GENERATION
The 2024 iteration of the Apple iPad Pro can only really be classified in one way – a return to form, and Apple’s renewed pursuit for thin devices. It should, ideally, be a glimpse into what the future holds. As I wrote in my review, it is a pivotal moment for the iPad Pro line, a chapter that’s defining the vision of a powerful computing device. You’d expect nothing less, considering it is introducing the M4 generation, the first time an iPad has had that luxury with Apple Silicon. New OLED screens (its actually two OLED panels fused into one), may be the visual highlight that gets your attention (as it should), but there’s something else that could hint at broader influence. The slimmest ever iPad Pro, could lead us to even slimmer MacBook Pros, MacBook Airs and the iPhone, in the next few years. Apple, and no one expected this, managed quite successfully with the iPad Pro 11-inch and the iPad Pro 13-inch. Core to this change would be the Apple Silicon, since the company has found the elusive balance in the performance per watt metrics. Everything that’s helped slim down the iPad Pro, may be relevant for the MacBook Pro and the MacBook Air in the coming generations. It may need more for the iPhone though, since physics will pose more challenges – a bigger screen for the “Pro” phones should open up the space for slimming everything down (this year’s iPhone 16 may be too soon, I suspect). I’m intrigued by the prospect. Let us wait for it to unfold, as it should.
Everything about the newest iPads, from our viewpoint…
Apple iPad Pro (2024) defines vision of a powerful computer, without being one
Apple iPad Air (M2, 13-inch) is more iPad than you expect, and that’s its charm