Don’t they consider Spotlight to be their search engine?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Experimenting With 'Apple GPT' AI Tool, No Launch Planned Yet
- Thread starter MacRumors
- Start date
- Sort by reaction score
I think they are scared of future Microsoft Windows Ai capabilities, and that many will shift to them, ditching MacOS, likewise to Android.
We’ve been able to chat with our devices for many years. Almost no one does (except for cases like while driving). Visual interfaces are way superior to voice/chat ones, that’s why customers don’t use the terminal but the mouse. I think this trend of “chat interaction is going to be a revolution” is more a desire for something new than thoughtful UX analysis.
We have not been able to chat with our devices. We've been able to bark a limited set of orders. That's a completely different experience from conversing with our devices.
Seems like people have already figured out a stopgap for that. Using LLM to translate what a person says into the exact commands Siri will understand. Someone made a hundred or so step Shortcut for it, seems like Apple could do it directly fairly quickly until they can completely rewrite Siri.
Voice capability has its place => a perfect example being while driving. When driving, my focus is on the road and its surroundings so having an excellent voice recognition comes in handy.
Remember this article? Apparently Siri is not fixable, according to Apple’s own engineers. I mean they could probably keep the name intact, but would have to rebuild it from the ground up to be more useful. 🤷🏻♂️
Report Details Turmoil Behind Apple's AI Efforts, 'Siri X,' and Headset Voice Controls
Siri and Apple's use of AI has been severely held back by caution and organizational dysfunction, according to over three dozen former Apple...
www.macrumors.com
A couple of months ago an Apple engineer emailed Accidental Tech Podcast to say that any Siri upgrade takes weeks to compile. The foundation of Siri is not made for expandability. I’ll see if I can find the episode.
EDIT: This is what I’m thinking of.
Ex-Apple engineer says Siri is way too clunky to ever be like ChatGPT
John Burkey was tasked to improve Siri in 2014 and said that its clunky design and code makes it tough to implement new features in a timely manner.
www.businessinsider.com
Last edited:
To a higher or lower degree, we’ve been able to do that. If you mean GPT-like, in case it was so relevant, we would be watching the death of visual UI… but here we are. Visual interfaces are much more direct and discoverable. Imagine selecting, dragging and dropping, tagging, etc. by typing. It’s a nightmare.
I would love to see Craig at one of the Apple events give a 15 minute dissertation on Siri, starting with the apology that they gave for Apple Maps -something like “ I know, some of you have been disappointed with Siri in the past, but this is Siri now!” And give a lot of impressive demonstrations of her new, powers..hyper blasted by machine learning power. Keep the branding , but change the paradigm.
Every interface has it's pluses and minuses. I won't argue that. We have not been able to converse with our devices. We've been able to issue commands in a more natural manner than we could in the 90s, but we're not conversing. We're not able to string together a series of commands, respond to questions from the AI, etc.
A GPT-like interface that allows for a back-and-forth interaction with one's device will be huge. It won't replace touch, or mouse, or even typing (just like none of these methods of interacting with the machine have replaced each other), but it will be a game-changer in how we interact with our devices.
The above shows just how different the public thinks AI is from how it really works. The current technology can not create new things. Yes you can ask it write a program in Python that already exists. Like "write a program to blink an LED" and this works because the model was trained on hunded of blogsthat show beginnig students how to make an LED blink.
But what it can't do is this "Hey GPT, our robot code is designed to keep the center of gravity inside the support polygon. This is good, but to work better we need to keep the zero-moment point inside the polygon. Can you modify the code to use IMU data to do this."
The current tech is not able to do the second example, and it is not "just a matter of time" until it can. It is like waiting for railway locomotive technology to be able to fly. It never will. We need very different technology to fly, not just upgraded trains.
Our current AI tech has ZERO understanding of what it does. The harder coding request absolutely requires understanding and not just a huge memory.
That said, MANY interesting problems can be solved using only "huge memory". Most of us in our day-to-day lives don't need to invent new things
I always wondered how the voice-dictated audio logs on Star Trek weren’t full of “uhh”, “scratch that”, and “oh, I forgot to add”s. Maybe I’m just scatterbrained, but from personal experience, conversational computing is suited mostly for quick simple tasks; anything that requires more than a moment’s worth of focus or thought will likely need a faster and more interactive interface for revising, and proceed in fits and starts. Environmental factors are another issue: talking to “yourself” is a big faux pas in most quiet environments, and noisy busy places are not likely to encourage lengthy focused work. Add to that the current tendencies of LLMs to make things up and—like Siri!—completely biff your prompts, and you get a recipe for frequent frustration.
Before the Vision Pro was revealed, I was fond of spouting off that Apple would wait for a simpler, mobile AR-only experience to be possible, since emulating VR sets like the Oculus seemed foolhardy at best. I’m happy to say I was totally wrong on that one and that Tim and co. did manage to show an interesting, useful, and above all restrained take on AR/VR. I’m confident that the same restraint is at work here: they won’t jump on the LLM train until it doesn’t derail every third mile, but when they do, it will be used quite skillfully and well compared to Microsoft and Google’s current “boast loudly about slathering it everywhere” campaigns. Apple devices already use quite a lot of machine learning behind the scenes, so I highly doubt they are complacently falling behind. With my track record for predictions, I hope I haven’t just jinxed them…
Good. We are all going to be killed by this AI crap, but at least Apple's version won't be selling our privacy...
Siri continues to provide entertainment with its comically errant results at times.
Most of what passes for Actual ( human ) Intelligence, is not intelligent either.
My aging brain can't handle the same name (Ajax) being used for something totally different. Wires are getting crossed!
Yup, totally true. But human beans ... I mean beings... are simplistic. They don't care about specifics. They just like buzzwords.
Maybe the time has come for "copyright" to die? Once something is created, it should be owned by humanity, maybe? Everyone that releases open-source software follows that mantra, and we all benefit from it. The very device you are using is powered by a ton of open-source software components.
Siri isn’t a language model. At best Siri may use it, but then I don’t want Siri to feed my confidential information to Apple GPT.
I'm guessing you're not a successful author, or a successful writer in general. I mean it would be great if folks wrote successful books or music and said "this is now owned by humanity, so it's free to all", but that's not how the world works.