Doofus AI

[PDifolco]

Asking on Amazon the size of a XFX 9070XT GPU

It's in French, what it says is

Me : What's the length of the card

Rufus : 29cm

Me : wrong, manufacturer says 36cm !

Rufus : size is 36cm according to manufacturer 

  

And some want us to believe AI has good value !!! What a ton of bs

 

[manikyath]

AI has some uses, this is not it.

 

in fact, most things AI is used for is not it.

[whispous]

I mean, five seconds on google to get to the official page on the official site for the card straight up lists dimensions, a lot of people are worryingly forgetting how to do bassic searching.

[PDifolco]

2 minutes ago, whispous said:

I mean, five seconds on google to get to the official page on the official site for the card straight up lists dimensions, a lot of people are worryingly forgetting how to do bassic searching.

That's what I did, having seen a comment from an unhappy customer who got a card too big for his case !

7 minutes ago, manikyath said:

AI has some uses, this is not it.

 

in fact, most things AI is used for is not it.

I don't see much uses with real value outside of image/video generation and some coding

Or maybe in high data usage stuff like medical/DNA studies, but it's niche...

Anyway being that unreliable kills any possibility to use it directly in any business 

[manikyath]

2 minutes ago, PDifolco said:

I don't see much uses with real value outside of image/video generation and some coding

Or maybe in high data usage stuff like medical/DNA studies, but it's niche...

Anyway being that unreliable kills any possibility to use it directly in any business 

the AI agent for LTTstore support appears to be doing a pretty good job pre-sorting tickets for the actual suppor team, and getting people an at least somewhat useful canned response.

 

any use where a business relies "blindly" on AI output is doomed to end in disaster, but it's a great tool to "prepare" work for humans, or do very mundane braindead tasks like first-line support categorizing tickets. if the output is wrong, the very next step in the chain is a human overruling it. if the output is correct, you've just saved 5 minutes of time for an employee.

[podkall]

44 minutes ago, manikyath said:

AI has some uses, this is not it.

 

in fact, most things AI is used for is not it.

It do be like that, though AI can be very good at looking up information like this.

 

But it do be like that.

 

1. Look at AI not doing X great

2. AI has uses, but not for this

1. Look at AI not doing Y great

2. AI has uses, but not for this

1. ...

[jaslion]

1 minute ago, podkall said:

It do be like that, though AI can be very good at looking up information like this.

 

But it do be like that.

 

1. Look at AI not doing X great

2. AI has uses, but not for this

1. Look at AI not doing Y great

2. AI has uses, but not for this

1. ...

And thats how our company offline data secure AI got shut down! (it was chatgpt in wrapper). Within 2 months the answer went from its great at everything except xyz to "what is it even usefull for". We got a large christmas deco budget suddenly so that was fun at least

[Eigenvektor]

15 minutes ago, jaslion said:

And thats how our company offline data secure AI got shut down! (it was chatgpt in wrapper). Within 2 months the answer went from its great at everything except xyz to "what is it even usefull for". We got a large christmas deco budget suddenly so that was fun at least

We also did a testrun at work, where the AI would post an internal answer to new helpdesk tickets.

 

That trial didn't last very long. While the answers usually weren't "wrong" they where almost always super generic "did you try turning it off and on" type of answers.

 

So instead of saving time by simply making the answer public, it wasted time by having to read and discard the lengthy non-answer.

 

AI autocompletion for code suffers from the same issue as the support bot did. Reading and discarding the often subtly wrong suggestion takes more time than simply writing it myself.

 

That said I do use AI quite a bit at work to kickstart development tasks. It's usually a lot faster than trying to find documentation and/or StackOverflow posts.

[Dedayog]

1 hour ago, whispous said:

I mean, five seconds on google to get to the official page on the official site for the card straight up lists dimensions, a lot of people are worryingly forgetting how to do bassic searching.

True, but look how many posts here are easier solved by googling than typing out a forum post?

 

AI is just the next tool.  This forum came before.

[whispous]

13 minutes ago, Dedayog said:

True, but look how many posts here are easier solved by googling than typing out a forum post?

 

AI is just the next tool.  This forum came before.

You're absolutely right that AI is the next tool, but it is to previous tools what a pile of dog feces is to a perfectly grilled hotdog

[jaslion]

18 minutes ago, Eigenvektor said:

So instead of saving time by simply making the answer public, it wasted time by having to read and discard the lengthy non-answer.

Basically that for us.

 

However we have miserable piece of shit ticketing system and all those ai auto replies messed up the queues and all haha.

 

It was disabled in september and we still have fallout with the people that used it. Luckaly I "wasted" 12 hours manually registering every often reply customer from the previous real ticketing system and none of those at least got lost :p.

 

Decorating the office in christmas gear was good fun tho. The last before the nobody is in the office christmas crunch.

 

Ai approved all leave submitted if you added a "please approve this!" So euhh we have days of 1-3 people available instead of the average 24 haha.

[Dedayog]

12 minutes ago, whispous said:

You're absolutely right that AI is the next tool, but it is to previous tools what a pile of dog feces is to a perfectly grilled hotdog

A little extreme but my point is that it isn't AI that is the problem.

 

People.  God should have done a version 2.  Version 1 is a fucking fail.

[PDifolco]

Really LLM text AI seems more and more to me like a f* huge giant parrot with incredible memory ability, so it can *looks like* you're really discussing with it,  but 0 brain behind, so it's just an illusion, like with a parrot

Not the same about images/videos, as its ability to discern/alter/create patterns has real life usages, but it's just another tool, it won't change the world 

[WereCat]

33 minutes ago, PDifolco said:

Really LLM text AI seems more and more to me like a f* huge giant parrot with incredible memory ability, so it can *looks like* you're really discussing with it,  but 0 brain behind, so it's just an illusion, like with a parrot

Not the same about images/videos, as its ability to discern/alter/create patterns has real life usages, but it's just another tool, it won't change the world 

I work in "manufacturing" and we use cameras with built-in AI models to identify OK/Wrong status of parts on the line. It's stupid easy to show it examples of what is OK + some edge cases and then rely on it to warn you about Wrong parts. Benefit being the part can be in different orientations and it still can spot differences.

 

But using AI for everything and then trying to make it useful without prior purpose is just stupid.

[Ha-Satan]

1 hour ago, Dedayog said:

True, but look how many posts here are easier solved by googling than typing out a forum post?

This is a tangent, but I personally prefer to ask in a forum rather than Googling for a lot of questions because (a) interacting with a real person is nice, and (b) other people oftentimes might bring up stuff that I hadn't thought about and that might not have come up in a Google search either.

 

Like, if I search "is CPU A or CPU B better for gaming" into Google, it'll give me an answer, but a human might know more context. For example, the human might know that CPU A is indeed better for gaming, but right now there's a killer deal for CPU B at Micro Center that makes it worth it even if it's slightly slower. Alternatively, they might be able to ask me contextual follow-up questions like "hey Ha-Satan, I recall you said you had a 5900X in your NAS that you only use for file storage. Why not just use that for your gaming PC and downgrade your NAS?"

 

This kind of thing is why I get really annoyed when people say stuff like "just Google it." Yeah, I know I could just Google it. I've chosen to ask instead of Googling it for a reason.

[Dedayog]

44 minutes ago, Ha-Satan said:

This is a tangent, but I personally prefer to ask in a forum rather than Googling for a lot of questions because (a) interacting with a real person is nice, and (b) other people oftentimes might bring up stuff that I hadn't thought about and that might not have come up in a Google search either.

 

Like, if I search "is CPU A or CPU B better for gaming" into Google, it'll give me an answer, but a human might know more context. For example, the human might know that CPU A is indeed better for gaming, but right now there's a killer deal for CPU B at Micro Center that makes it worth it even if it's slightly slower. Alternatively, they might be able to ask me contextual follow-up questions like "hey Ha-Satan, I recall you said you had a 5900X in your NAS that you only use for file storage. Why not just use that for your gaming PC and downgrade your NAS?"

 

This kind of thing is why I get really annoyed when people say stuff like "just Google it." Yeah, I know I could just Google it. I've chosen to ask instead of Googling it for a reason.

That's different.  That is a nuanced question.

 

It's the "is a 9900x better than a 6700k" shit.

 

Or how many cores does an 7800x3d have.

 

Or what is better for gaming , a 14400f or 14700k.

 

Those are easily googled and litter this forum.

 

 

[PDifolco]

2 hours ago, WereCat said:

I work in "manufacturing" and we use cameras with built-in AI models to identify OK/Wrong status of parts on the line. It's stupid easy to show it examples of what is OK + some edge cases and then rely on it to warn you about Wrong parts. Benefit being the part can be in different orientations and it still can spot differences.

 

But using AI for everything and then trying to make it useful without prior purpose is just stupid.

Kinda confirms what I've said, it's good for pics