The reason why Apple has failed to integrate AI successfully into the iPhone isn't because we need an AI-first device, it's because AI is still universally being strategized as a solution in search of a problem. An AI-first device at this stage will fail for the same reason: it doesn't have a mission statement to solve particular problems for real people, it has a mission statement to be an AI device.
The absolute best case scenario for an AI-first device at this stage is that it ends up like the Vision Pro, which had a similar mission problem.
Agreed, you can see this from how consumer-facing AI is advertised. It's never something like "I wanted to do X, now I can thanks to AI". It's always "I used AI to generate a cute image to send my son" or "I used AI to write a silly poem about my boss" or something. If it was truly valuable, they'd be showing off how it can solve problems that existed prior to AI being invented, not these fluff tasks that don't respond to people's actual needs or create any value. The only thing that AI can do that people are willing to pay for is cheating on school assignments, but obviously the AI companies don't want to use that in their advertisements.
> it's because AI is still universally being strategized as a solution in search of a problem
What Apple showed seems quite useful. It is a shame they failed spectacularly at execution. Even the simplest things that should be answerable by an LLM and their data, which is what a lot of people want, should be a very low hanging fruit - so much utility without building a complete experience from scratch.
Why cant I just say 'do I have any notifications from a bank?' or 'show me emails that require my attention'. Those things are simple if done with a combination of multiple tools (e.g. feeding email content somewhere, asking it to classify, show the results), yet a three trillion dollar company, with dedicated hardware release just for this purpose, failed to achieve it.
I might be over simplifying things, but with infinite resources, they should be able to do better.
> I might be over simplifying things, but with infinite resources, they should be able to do better.
I don't think you are. The problem really is execution. I don’t need anything beyond what AI can already do—I just need an assistant that understands what I want and uses the right tools. I’m baffled that we still can’t get a reliable summary of emails, notifications, or appointments. If you give me the data in text format, I can paste it into ChatGPT/Gemini/Claude and have a much better dialogue than with any current phone assistant. Somehow, trillion-dollar companies still haven’t solved that.
Exactly. Even the simplest models that could run on phone could have been used to improve Siri or give some better utility other than setting timers. Yet, it is slightly more advanced than a decade and a half ago.
I think it was rushed. Apple may though have lost "the vision thing" — but I'm willing to allow them another release cycle to mature their initial bandwagon-reaction.
I don't know for all other people, but ai totally can be solution to some of my problems, but the current technological landscape does not let it. I'd like for my ai assistant to have as much context about me as possible so that it can give me help in different situations. At the same time, the ai should be unable to share this info with anyone else. However, I cannot imagine Google or Microsoft committing to safeguard data they won't sell. Neither I can imagine Apple offering such an AI without trying to rob me even more blind than I'm now while telling me that they are my best friend. Doing it myself could be a half-baked but somewhat functional solution and honestly I hope that someone of the same mind would build on the labor of others to provide something workable.
I've been realizing lately, as I have leaned heavier on LLMs to do what I would have used Google and search to do previously, that what I want — what I am searching for — is not a page or site but just the damned answer to my query.
"At what time does the first train from Stirling to Edinburgh arrive?" I don't need a page to the train time-tables (or god forbid, a vacation package to Scotland) — just the answer to my question.
I kind of disagree in that I find AI features that I can use with the phone quite useful like being able to say what's that tree or how do I get to Croydon or such like. I don't feel any desire for the AI processing to be built into the device however, I'm quite happy to have it running on Google or whoever's servers and be able to access it from whatever device.
It was always a bit hopeless. The whole idea of let's get people to carry a second device which is like a phone but much worse rather than just using a phone app was always iffy. I guess the Apple Watch works but it's tricky.
No. The AI pane of glass is the next killer product.
The next major device won't be an ad funnel though. It'll give users first class access to the whole pane of glass. Not a managed ads experience at the will of some monopoly platform, but something where the AI serves us instead of being extractive.
The minute we have a broker or agent between us and the "user is the product" services that try to advertise to us and steal our time, it's game over for the old model of revenue. Google, ads, all of it will vanish. There won't be any more selling to me or the rest of the world ever again. You'll have to pay us to get our eyeballs.
Let me clarify: if we have a pane of glass where we run our own agent with our own best interests in mind, then nobody can get through that layer without it being permitted by us.
No more ads.
No more stealthy product placement.
No more paid or featured listings.
It goes further.
No more rage bait, attention bait, low information filler. The annoying people in life and in social media disappear to the great filter.
AI agents can clean up the shitty place the Internet has become.
AI agents are personal butlers. Or internet condoms.
The way that works is this: on-device AI that can handle the task of routing and dispatching and filtering, which can then dispatch out to expensive cloud AI that would otherwise try to inject adds into the stream.
4) It's a broker between you and the "wild wild west internet"
That's cool, but is it "iPhone killer" cool? Maybe, but still unclear why. What's the mission statement of the device, to the OP's original point? It runs an agent, who cares?
Is the mission statement for this device basically "Use the internet without ads" -- if so, that's a pretty narrow market. People have learned to tolerate ads, I don't think people will throw away their iPhones for a better ad blocker.
I like the idea, but that kind of setup doesn't provide an infinitely growing revenue stream. Incidentally infinite growth targets are essentially why the internet looks the way it does today. That's the thing that needs fixing, product development is secondary to structural incentives.
Antitrust doesn't change the basic structure of capital markets, which will still demand infinite growth even if there's a government around capable of breaking up monopolies.
The absolute best case scenario for an AI-first device at this stage is that it ends up like the Vision Pro, which had a similar mission problem.