In mobile phone, is it worth upgrading from one to the other for the sake of the AI only? Or are they all equally similar? I am on Apple and not satisfied, it always ask if I want to ask Chatgpt for the question, just an extra step. Also not too impressed with it photo app.
Which AI is best? Samsung vs Google vs Apple
Frankensnore on 12/02/2025 - 10:45
Comments
Pronounced Dipstick
Pretty slow in my experience and times out too often
ChatGPT did too when it first launched to the public.
Qwen
Download the Microsoft Copilot app (or any other) and use that when you have a query. There's no need to have OS-embedded AI and certainly not changing an entire device for it.
Apple's own implementation of Siri/AI is always going to be poor because of their privacy stance.
Big AI user here, but 100% with you. I am not a fan of embedded AI (outside of specific use cases, eg: self driving cars, robots)
- Privacy reasons
- Embedded requires small low parameter models. Look at Deepseek, people rave how they can run it locally, but it's complete garbage compared to running on proper hardware. That's why USA hosted DeepSeek is better than the Chinese hosted DeepSeek due to the better hardware.
- If a security compromise happens on an AI enabled device, the tailored damage they can do can not be imagined. It's absolutely insane how much data the AI can gather and return back to the hacker.Query Deepseek's model about the Tiannamen Square Massacre locally hosted vs Chinese hosted. There's other advantages to running locally. I would presume USA hosted will give a similar response to locally hosted in this instance, but I wonder how the USA hosted version would respond to a question about trans mutliation of children or some other hot button politicised topic that is controversial in the USA.
I'm quite ignorant in the workings of AI-based chat programs like ChatGPT and Deepseek. However, I have a few queries based on the examples you have shared.
Could it be that locally hosted AI could bias the answers as well? After all, the literature on Tian An Men in the USA would also contain markedly false opinions, but repeated ad infinitum that it becomes "truth" due to sheer numbers of literature on it?
Perhaps the parameter definition needs cleaning up, after all, the incident in common parlance in the USA refers to the Tiananmen massacre, but the actual event of a massacre in Tiananmen did not occur, and there will be a deviation of generated results if the AI model did not bother to seek clarification from the user by asking follow up questions?
@fisherfriendman: Yes of course, garbage in garbage out. Bias in, bias out. My point was more about there being some kind of layer of post-processing for certain subjects which appears to happen external to the model itself (as demonstrated when you ask those questions locally hosted vs china server hosted vs usa server hosted).
None. Thanks.
Watch this space, but Samsung, Google and Apple are all currently really bad.
Google has massive potential if they leverage the proper Gemini Pro 2.0 models.
Apple models are far too small to be useful, but that can change, or they can do a proper partnership with OpenAI and not use the crippled low parameter models.I was surprised at just how bad Google AI is. Given their early work with voice assistant, I always thought they'd be a big player in AI but IMO they're not even in the race currently. They're obviously trying everything they can to catch up and are forcing it upon anyone using their search engine not doubt for training but it just gives out wrong information at least as often as correct and there's no way to set it to disabled so I'm having to constantly use the -ai switch.
No their AI is currently top notch, just the dumbed down models on the phone.
They have some great models, often topping the leaderboards,
- Gemini 2.0 Experimental
- Gemini 2.0 Flash
- Gemini 2.0 Flash Lite (from a price \ capability perspective, I argue this is the best value prop currently out there), craps over GPT4o-mini and Claude 2.0 Haiku\Sonnet
- Gemini 2.0 Flash ThinkingYou're clearly AI yourself. One response it's really bad, the next it's top notch. Exactly the sort inconsistency you'd expect from AI.
@apsilon: Context.. the discussion was referring to the AI powered assistants on the phone using Gemini.
You somehow extended that into Google AI… hence my broader response regarding Google's capability.Bottom line, currently this on-phone AI is crap, but doesn't mean their AI is crap.
Thank you for the -ai switch! The absolute nonsense google AI shits out should have made them remove it from their search engine within the first week. I would be ashamed of that failure, not force people to look at it.
I know a mate is sick of his Android phone constantly begging him to turn AI on. I use a different home app (Microsoft Launcher) so I never see this. I assume Google's home app is just as shitty as their web search.
I would definitely not be looking at changing an entire eco-system from Apple to Android (or vice versa) based on their AI models alone.
AI is still a gimmicky thing from these companies, and Apple has already had major issues with it. It won't change your life or give you as much of an edge as you think it might.
AI on the phones is crap, but for many if you are not using AI to assist you at work, you are falling behind.
Problem is there is a lot of hype and AI added to crap for the sake of it with zero benefit, has also made many things worse.But has easily improved the quantity and quality of my employees 10x fold.
The ideas that come out of AI are insane, really gets the creative juices flowing. And the new premium models like o3 are really good at coding, documentation, basically any reasoning task.Can take legacy code and get documentation\diagrams produced, dependencies mapped, risks, all in seconds.
Ofcourse still at the stage that a trained human needs to review but bloody amazing stuff.I personally don't use it for everything, just the hard hitting stuff.
I had an argument with Gemini the other day, it took me three replies to explain the very very simple logic, before it changed its mind and agreed with me 100%. These AI's seem to me to be very smart at trawling the internet for the most popular (wrong) answer. They do not seem to have any significant true/real intelligence, only pretend intelligence.
They're all pretty bad/not worth paying for at this stage except for niche scenarios.
And benefit of competition is once (if) they improve significantly, others will likely just copy/create their own version which will be equally capable.
is it worth upgrading from one to the other for the sake of the AI only
No
I use chatgpt on my desktop all the time but Gemini on mobile has made zero difference to my daily life.
What does Artificial Insemination have to do with mobile devices?
So you're basically asking which external source should you forfeit your individuality and self to?
Next dilemma what if friends get a better,bigger,brighter stand-in intellect , that outdoes your choice?
Despite what social media ,peers and the brochures say, a better looking doctored fake photo is still faux.AI . The snake finally eats its tail.
PS wait a bit longer for Tesla Cyberfone.It comes with a free how to salute app.
I outsource my thinking to Google, SBS, ABC and Wikipedia /s
You are 100% XXXX
I'd bet that Apple AI would be the most useful, in time. Apple AI only really just properly launched so it's still a bit green. I love the notification summaries, even if they are wrong. It's easy to imagine being able to one day tell the iPhone to only show you notifications from Instagram from specific friends and finally get rid of the stupid random notifications from everyone you follow, and Snapchat's notifications about new filters and such.
Apple still a bit green.I see what you did there.
Still in the Granny Smith phase.
The core of their problem
Looking for seed capital.
I'd bet that Apple AI would be the most useful, in time
did you think the same about the apple car?
Turned out to be a lemon
A bold bet for what might be useful “in time” considering it is currently the least-useful by an extremely wide margin. But sure, I like to imagine.
I am on Apple and not satisfied, it always ask if I want to ask Chatgpt
Then chat gpt wants to ask Wolfram.
Wolfram then goes and looks it up from Encarta 95.
I've turned off Apple AI already. It is worse than useless.
yeah apparently also takes up a lot of storage.
also forgot the clickbait headline but apparently it's "auto reply" to people with messages makes you sound like a sociopath. best to turn it off until they actually make it ready for release.
and with that rainbow border animation every time you trigger siri, stuupid!! they really just jumped on ai hype train and released something really crappy to make iPhone 16 more appealing. but don't they say like getting the latest product for the latest feature makes you a beta tester.
Never mind the chatbots. Phones do come with other AI features.
Pixel has a couple of things I find useful:
- "hold for me" : when you have a wait for a call centre to answer, let the AI hold, and will ring for you when a human connects, while keeping them from hanging up.
- "call screening" - when you get a call from an unfamiliar number, it will ask why they are calling.the best is Actual Intelligence
yeah I thought it was cringe but I will say it now.
which ai is best?
chatgpME!
No matter what answer is given today it will be outdated in less than two months. It's all moving too fast right now.
Deepseek is best.
<roll eyes>