Gemini / AI not so intelligent (rant)

[yllek]

I'm at my wits end with the assistant on my newly acquired S25 ultra - I honestly think I'd be better off talking to a 3 year old child. Where do I begin? Devices. I'll ask it to turn on/off a device and it will tell me it's not available or set up (it is). I'll ask it again and it will do as asked, just like that. 

Spelling. It thinks there are 3 Os in the word tomato (my nephew was checking his spelling and thought it'd be cool to use AI). It took a 10 minute conversation with Gemini before it finally admitted there are only 2 occurences of the letter O in tgat word. 

Now I've asked it to recommend independant high st stores specialising in smart casual mens clothes within a 100 mile radius of my location (north west England) but to exclude online retailers, charity / 2nd hand shops and supermarkets. What does it do? It ignores me (recommends london shops, shops that don't even sell clothes as well as online retailers and supermarkets) and refuses to acknowledge it is wrong until I go through a step by step explanation of why it's wrong before I get the usual "sorry, under development / still learning" etc. spiel. It's consistently making the same errors over and not doing what I've asked of it. I did eventually get it to admit that it is useless and not fit for purpose which brought me a brief amount of pleasure. 

Then I've tried Gemini live. Pointed the camera at a Ficus Exotica and asked it to identify it, which it got wrong (said it's Devils Ivy). When I corrected it, it told me I was right, claimed it to be another plant I can't even pronounce and when corrected again, told me it was.... Devil's Ivy. 

I'm finding it maddeningly frustrating to use. It took over an hour before I gave up trying to get it to come up with a meal plan meeting specific criteria. Grok3 managed it in about 5 seconds flat. In every way I've used this damned thing, it has not once got anything right or responded to requests appropriately. 

What the hell is wrong with it? Is anyone else experiencing similar? All I can say is thank God I'm in the free trial period because if I was paying for this garbage I'd be demanding a refund.

[Glenntech]

Like anything, it's a work in progress.
Google are making great strides in development of Gemini
But there will be specific instances.
Don't forget you have Bixby that can help with modes and routines etc.

But Google are gradually getting there and the more people that use Gemini. The better it will get

[arianwen27]

I think you're using ai in the wrong way. As in, you're trying to use it for everything it's currently bad at.

Turning on and off a device, that's not Gemini, it swaps to google assistant mode for that to happen.

The 3 O thing is a problem with every ai model. To save the super technical explanation, ai models don't see words, they see parts of words. So they are inherently really bad at spelling.

For the store thing, that's tool usage. The ai model has to use google maps to request your location and obtain info. The info the ai is actually allowed to access is rather limited. So the ai can do it, but it's not been allowed the access needed to do it.

Gemini live is brand new, it's going to be a little strange. Especially with plants where that's a rather advanced thing.

Gemini is really good for hard and advanced questions, mainly when put in pro mode. Kinda, try to use it for hard questions that take some thought, not something you can work out yourself in 5 seconds.

The version you have been talking to was probably 2.5 flash, the dumb one. Swap to 2.5 pro and see if you have any improvements.

[JAMES4578]

Gemini will not work well for everything but is constantly being developed, I would say that the present time no AI model is completely reliable.  However they still can be very beneficial in our lives,  I have only used the basic version of Gemini and have generally been impressed (though have not really used it extensively).

[Joeeye]

Gemini does so many things wrong agreed, but ignoring all the things it does get right is an even greater injustice. As pointed out, it's a WIP and Google have made some very impressive strides to where it is today.

[yllek]

Thanks for the comments. I'm using the 2.5 pro already. If this is the smart version, I'd hate to see the dumb one. One thing tgat is really grating me about is is it's repetition. With the plant, for example, I told it that the plant wasn't Devils Ivy yet it told me it was (again). I'm only doing these things to try it out. If it can't get basics right, how is it going to get more complex matters right? I have incidentally tried giving it a more complex problem and I was impressed by how much it managed to get every single aspect so wrong.

I think I'm just going to send the phone back and revert to my old one which at least knew my location and could spell the word tomato.

[Glenntech]

It's not the phone that's at fault but the Gemini software.
Maybe try contacting Google to enquire about the issues.
I believe Google has some more updates coming soon

[yllek]

If i ignored all the things it did wrong, there'd be little else left to appreciate!

[UNKNOWNUSERNAME]

Gemini for me is honestly very useful.
Does what I want it to do, although as mentioned AI is a work in progress, so it wont be perfect.

[Joeeye]

Then you might just not truly appreciate how far A.I. really has grown. It's easy to point out flaws, not just in A.I., in humans as well. If we defined ourselves by our worst virtues, it could create laundry lists longer than the positives. Does that mean we should never acknowledge, or appreciate ourselves?