ChatGPT and Microsoft have proved, both can use the same base models, but one offers night and day better experience than the other. So in theory Apple could offer a much better experience and while their app experience will most certainly be better, unlike ChatGPT vs Microsoft, I don't have as much faith in Apple's responses, etc being better than Gemini, since Apple will be playing it very safe and has so many constraints. But Apple is trusted and so being able to provide on device access to a plethora of data now or eventually could be a major benefit. Time will tell.
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.
Google Gemini Could Be the Ceiling on Apple's AI Ambitions
- Thread starter MacRumors
- Start date
- Sort by reaction score
Hajj.david
macrumors 6502a
What customers say on forums and echo chambers like the verge and what is the reality is often two different things. Fact is billions of people are using AI on at least a somewhat regular basis. I see old people using it to look up things, religious scholars to help write up their speeches and do research, engineers like myself using it for boilerplate code and minor fixes, kids to cheat in school, my brother in law who doesn't even have a computer (very techillterate) has been using it to quiz himself on the state trooper exams.
Do customers want it pushed everywhere like microsoft did with copilot (best branding/name by far)? No. But to deny that customers want AI holistically is not to be seen.
Hajj.david
macrumors 6502a
I agree, Claude or Grok would have been a much better choice here.
Paul Shields
macrumors member
You are assuming Apple's customer base are actually that bothered about AI. Personally I'd prefer the stability and simplicity of iOS even if its implementation of AI falls behind Google.
Le0M
macrumors 65816
Apple should give users a completely private, "zero-access", well functioning, on-device AI.
That's what's gonna grant Apple a paved road ahead, even if its AI isn't the best one on the market.
That's what's gonna grant Apple a paved road ahead, even if its AI isn't the best one on the market.
The technology is not there yet to allow capable models that run fully on-device for devices like smartphones.
And when the average person's term of comparison is a cloud model (ChatGPT, Gemini, Claude) it doesn't bode well when they realize it's far inferior - as Apple Intelligence is today, and any mildly complex request will get forwarded to ChatGPT anyway.
LionTeeth
macrumors 6502
I only tepidly leave apple's AI on right now. If they ever move to the google one I am out. I'm not with Google for a reason.
HouseLannister
macrumors 68040
If you are trying to code and squash security bugs, Anthropic is great. But that's all for enterprise and developers. As a product for end users to do daily tasks involving their own personal data, Gemini can't be beat.
Kirkster
Contributor
Still not turning on AI based on Google. Unless they are giving us a choice of the AI we are linking with I’ll be happy to keep Siri a basic assistant with the Action button linked to something else, just like now.
At this point I wouldn't worry about being "better" than Gemini.
The bar is so very, very low for Siri. Being able to do what Gemini could do last year would be quite enough.
And that still may not be enough to do what they "demoed" as vaporware in 2024.
The bar is so very, very low for Siri. Being able to do what Gemini could do last year would be quite enough.
And that still may not be enough to do what they "demoed" as vaporware in 2024.
switz
Contributor
The conjecture here is all about vaporware meaning we do not what exists, and if "it" exists" does "it" even work. The demo today may show pieces working, but do not buy the sandwich until all of the ingredients have been actually verified which usually takes weeks or months.
Maybe by May of next year this vaporware might be fully operational. We still lack some features promised for the current operating systems last year.
The bovine byproducts are just getting deeper in the software pen....
Maybe by May of next year this vaporware might be fully operational. We still lack some features promised for the current operating systems last year.
The bovine byproducts are just getting deeper in the software pen....
Hajj.david
macrumors 6502a
If Apple just uses the Whisper models or frankly ANY AI model for transcription it would be a million times better than the garbage now.
jdawgnoonan
macrumors 65816
If I can install any model I want to use in the cloud, then all I really need built into my phone is something easier to use when I’m driving and need something from CarPlay. I do not intend to ever run my personal life with AI models. I’m not interested in them booking reservations or flights or hotels and I will never give an AI access to anything financial in my personal life. There is a lot of hype here, but the people who do manage their personal lives with these assistants will never be a majority. I do wonder, if one does allow their lives to be ran by these things, how much of the AI direction will be influenced by paying advertisers. The thoughts of my AI being influenced by advertising is absolutely revolting.
contacos
macrumors 603
How they integrate and solve this will determine if I switch to Samsung. I’ve been entirely unimpressed with iPhone for several cycles now.
I don't disagree, but the other company that can pull this off is Google (as well as Microsoft, for a lot of business customers).
It's far more likely that a lot of the context you're referring to sits somewhere in a Google system rather than an Apple system. It's the privacy nightmare, of course, and we can feel about that however we want, but Google just has far more information about the average consumer than Apple, which inevitably puts it into a better position.
I really hope Apple can find a more privacy-friendly way of doing this, but I also think Google is in a much stronger position than Apple to actually capitalise on what may make AI useful at some point.
Not sure which tools you are using and how, but modern AI coding tools can be very powerful. They don't replace human expertise, but they can definitely do a lot, especially when used in a disciplined process.
Ursadorable
macrumors 65816
Interesting how you buy into the Apple ecosystem so you don't have to touch Google's anti-user/anti-privacy trash, and then Apple get's in bed with Google.
Waiting for the day that there's a 3rd option available.
Waiting for the day that there's a 3rd option available.
macduke
macrumors G5
Apple is so far behind at this point I'm not sure if they can catch up, because these companies are in the beginning stages of using AI to speed up AI development.
Like I've been using Codex on my Mac, 5.5 on the 5x Pro plan, usually in High or Extra High. Now I'm a designer and developer with 14 years of professional experience designing and building for the web, with dabbling in apps and games. So I can specifically communicate with this thing with high level prompting and general understanding of all the pieces that bring a successful web app together.
With Codex on 5.5 and my local OrbStack Docker containers and git on my MBP, I'm unstoppable. I had a working alpha in 16 minutes and 47 seconds that covered core functionality, in a one shot prompt that also did the scaffolding with a dealer's choice on the tech stack (it chose Vite/React). The rest of the time was spent polishing UI/UX, adding features, fleshing out settings panels, doing QA on edge cases, etc. This personal project I just built yesterday during a lazy and rainy Sunday afternoon and evening is blowing my mind. Something like this would've taken me a month to build previously. It even basically nailed the logo first shot based on my detailed art direction, aside from manual color tweaks and a second prompt to generate the wordmark (did that on purpose to keep it focused). Now I'm sitting here wondering if I can sell this, and also turn it into an Apple TV app as well. I'm definitely going to be running it every day on my Pi 4 on my local network, at least.
Like I've been using Codex on my Mac, 5.5 on the 5x Pro plan, usually in High or Extra High. Now I'm a designer and developer with 14 years of professional experience designing and building for the web, with dabbling in apps and games. So I can specifically communicate with this thing with high level prompting and general understanding of all the pieces that bring a successful web app together.
With Codex on 5.5 and my local OrbStack Docker containers and git on my MBP, I'm unstoppable. I had a working alpha in 16 minutes and 47 seconds that covered core functionality, in a one shot prompt that also did the scaffolding with a dealer's choice on the tech stack (it chose Vite/React). The rest of the time was spent polishing UI/UX, adding features, fleshing out settings panels, doing QA on edge cases, etc. This personal project I just built yesterday during a lazy and rainy Sunday afternoon and evening is blowing my mind. Something like this would've taken me a month to build previously. It even basically nailed the logo first shot based on my detailed art direction, aside from manual color tweaks and a second prompt to generate the wordmark (did that on purpose to keep it focused). Now I'm sitting here wondering if I can sell this, and also turn it into an Apple TV app as well. I'm definitely going to be running it every day on my Pi 4 on my local network, at least.
DavidLeblond
macrumors 68020
We have a Co-pilot subscription so it's GitHub co-pilot plugged into Visual Studio and Visual Studio Code (depending on my use case, I use both.) Not sure the exact model, but I think it's Claude.
I generally just slam on the snooze for an hour when I've realized it's not going to be helpful for me and I've learned that if I ask the chat to help me with a small tool I have to explain to do it in Python or else it tries to do everything in Powershell and then spins out and dies.
This is the future: highly customized, personalized software. SaaS platforms are going to see a sharp decline outside of those with massive amounts of enterprise customers.
Small shops will pay Claude for a couple months of Max 20 and then they have a system that works better for their needs than an existing product.
Not being able to generate a basic modern feature of your phone is not a good look for apple. I hope they work hard and catch up on AI.
On device AI is what we have now and it sucks. Dont have the memory bandwidth and processor speed for that to be great. Siri also started off device until tech caught up.
mzeb
macrumors 6502
I’ve been working a lot with Google’s open weight models (Gemma E2B and E4B) both of which are suitable for edge computing and I’ll say that I am impressed. Their performance is impressive both in high reasoning and instruct cases and the MoE routing seems pretty light weight. I haven’t had chance to play with the vision capabilities yet but Gemma 3 was pretty good so I suspect this will be too. I think Apple has probably made a good choice here.