The Tech Press and rank and file computer user continue to diverge in ideology. AI is important but for consumers the “right AI” is more important.
Ternus, who reportedly has a reputation for maintaining Apple products rather than innovating new ones, will be tasked with a tall order in leading the world’s first trillion-dollar company into its new AI era — not only playing catch-up, but trying to get ahead of its competitors, which are already moving at breakneck speed.
They don’t need to “catch-up”, this is based on the assumption that Apple needs to do what every other company that sees Enterprise as their client base. What Apple needs to do is cull the vast offerings of AI down into palatable chunks for their user base which is consumer. Consumers aren’t feeling rosy about AI. It’s caused products to become more expensive and hard to find. For many consumer use cases the benefits are dubious.
I’m excited about AI but only from the standpoint of what’s important for the consumer power user. Apple can certainly find this niche and exploit is effectively. I see it as a potential services bonanza with perhaps an Apple One Premier AI bundle which gives power users excellent access to Apple Intelligence without worrying about tokens and caps within reason.
I agree. Apple continues to beat the “on-device”, “private and secure” drum but I don’t see business & enterprise crunching their data on client computers. Just managing user accounts and data access for employees used to be a handful at times, and now companies are going to have to manage the same for AI agents. Business AI will be running on servers and/or the cloud for the foreseeable future. IMO
Apple needs a competent digital assistant for the iPhone, iPad, and Mac as well as any future smart glasses, and table top “robots”.
I think Apple will be cheered, in hindsight, for licensing AI from Google and avoiding the capex bloodbath that the rest of the industry has spent trying to avoid AI FOMO.
Let everyone else invest tens and hundreds of billions building AI data center capacity that may or may not be needed.
No shame in Apple’s AI being “pay as you go” to Google or others. Sure, it will not be cheap, but it is value paid for value used, rather than huge gamble on the unknown future.
Every company on the planet is obsessed with creating agentic AI - Apple seems to be asking what AI can do with the existing tools and resources we already have.
I’ve grown to prefer this “slow and steady wins the race” approach as of late, as my disillusionment grows with AI companies (ChatGPT specifcally, Gemini secondly).
I’ve landed on Claude as justifiably useful for my own needs.
I think this is accurate, although I suspect for some (many? most? Who knows…) consumers paying attention, the ‘right amount’ is approaching zero.
I’m lucky in that I don’t have an employer who is forcing me to use AI, but I have yet to see an application that would meaningfully provide any benefit to me – certainly none that come close to offsetting the considerable moral and practical objections against AGI.
Of course, people have different experiences and needs and therefore may feel differently, but the fact that Apple hasn’t jumped on this particular bandwagon is a point in its favour, as far as I’m concerned.
As an aside: I am utterly baffled by the attraction of agentic AI. I cannot think of anything more risky than letting an algorithm with a built-in ‘hallucination’ rate loose on my data.
My favorite AI story-of-the-week is Allbirds, the once-golden shoe company, that crashed in value, sold its remaining assets, and announced it was pivoting to become an AI company.
No joke – that’s the epitome of the bubble we’re in. I’m glad Apple hasn’t pivoted. As a normie long-term user of Apple products, I’m very enabled by my Apple hardware/OSes to use the AI tools I’ve chosen to use at this point. I don’t want or need more.
Certainly there are plays on the enterprise side Apple could and should consider, I imagine, but I’ll bet Cook and nowTernus have been watching the billions poured into model-making and data-center-making, relieved that Apple decided to wait for the deflation and shakeout. The tech and business press’s insistence that Apple is missing out is just fluff.
I’ve seen Tesla without a driver go crazy and other auto pilot errors. Agentic AI to me is like the race for Autopilot …it’s dreamy until something cataclysmic happens and the finger pointing starts.
Also companies that dream of this AI future where computers running AI are doing all the work have lost one of their most valuable resources and that is human. Once your company requires AI employees the pricing that AI vendors can ask/extort begins to approach inelasticity.
@KVZ I had seen Allbirds in the news and had no idea what it was about. What a strange pivot. Kind of reminds of when Apple found success with iTunes Music suddenly it was in vogue for companies to “claim” they were working on their own music stores.
People are expensive. And benefits used to be the way one company could make itself more attractive than another, that didn’t offer any. In the 1990’s benefits could cost a competitive company around 33% more than payroll alone.
Today, with states and the Federal government requiring certain benefits, I can’t even guess how much they cost a company. I can see the desire for the efficiency that AI promises, but that comes with its own set of challenges.
(Probably better discussion in the AI section) - As someone exploring AI, and not taking sides in the larger issues, I have found one has to really expand the context to understand the uses of AI.
There is a lot of amazing uses for agents where acting on quantitative data isn’t the ultimate results.
A few examples from my own areas of interest:
An agentic AI for network operations and management can pull logs, data, and activity reports from a wide range of network devices and build analysis of typical versus atypical traffic and flows to create alerts and recommendations about network problems that would otherwise require drilling into megabytes and gigabytes of data.
An agentic AI for marketing can A/B test (evaluate alternatives) and analyze the performance of marketing communications, blog articles, video scripts, and ad copy to help optimize the creative workflow.
(Note that both of these examples are using AI to evaluate and recommend, not to create or act on data autonomously. I know some folks are having tremendous results with agents actually creating and publishing marketing content in some contexts, but I believe that is much more controversial and needs very careful controls to avoid AI slop results.)
Interesting, thank you! But I think that we’re probably using ‘agentic’ in slightly different ways (probably unavoidable in such a rapidly expanding field).
Perhaps technically incorrectly, I was meaning ‘agentic’ precisely for those apps which claim to be able to create or act on data autonomously: ‘an agent’ is after all someone who ‘acts’. Your examples seem to me to be analysis, which seems far less risky.
I think it’s a matter of nuance. The agents involved in the tasks I discussed are actively logging in to devices and services. In the first, to collect data, but in the second, to actively control marketing “campaigns” - posting ads, blog entries, changing website headlines/text/appearance, etc.
It is up to the humans to determine the level of access and autonomy of actions allowed, but I think there is utility is going beyond allowing 0 (nothing) agentic actions and 1 (allowing everything).
I think Apple were right to avoid trying to create their own large scale LLM. They didn’t have the right infrastructure to do that.
However we’re into well phase 2 now, where companies are figuring out what to do with these tools beyond chat, including in the prosumer market that Apple often targets - small businesses, graphics, coding, audio, etc. I think this is the time that Apple need to start taking a lead. Its early days, but Apple needs to define its approach and start to ship before long.
I think collaborating with Google was the right move. Apple is building their Private Cloud Compute to support their devices when on-device processing isn’t practical. They don’t need their own AI data center to develop their software.