this post was submitted on 27 May 2025
70 points (100.0% liked)

chapotraphouse

13921 readers
668 users here now

Banned? DM Wmill to appeal.

No anti-nautilism posts. See: Eco-fascism Primer

Slop posts go in c/slop. Don't post low-hanging fruit here.

founded 4 years ago
MODERATORS
 

let's see Deepseek...

I mean props for getting it right but holy shit why did you give me a novel (the pictured portion is literally the last 3 paragraphs but there are 15 more i didn't screenshot). This is how you know it is the most leftist LLM; it gives you a dissertation instead of a simple answer.

let's see ChatGPT

...Is Google just fucking stupid??

all 17 comments
sorted by: hot top controversial new old
[–] FuckyWucky@hexbear.net 32 points 1 month ago (2 children)

Since search is so widely used, Google probably uses the "dumbest" model to reduce costs and compute usage.

[–] FunkyStuff@hexbear.net 30 points 1 month ago

Could they just not make that garbage show up on every search? Yes. But that would make our dear porky-point very sad.

[–] TrueStalinistPatriot@hexbear.net 12 points 1 month ago

This explains a lot actually

[–] Mindfury@hexbear.net 18 points 1 month ago

i mean, deepseek told you how it tried to figure out the context of the question, so the yapology is at least somewhat, i dunno, logical?

google literally did multiplication then couldn't even do fucking addition while confidently telling you the wrong thing wtf

[–] abc@hexbear.net 16 points 1 month ago

before you ask NO i wasn't asking AI to count to 100 on purpose, I just forgot I had it in my clipboard from work and accidentally pasted it into my url bar & went very-smart when I saw the dumbass AI response

[–] moondog@hexbear.net 12 points 1 month ago

I've tinkered with deepseek locally a bit, a lot of fun. Seeing machines perform thinking processes (like chess.com analyses where you get to see each step of it trying to find the best move) or deepseek reasoning with itself step-by-step is pretty cool I think.

[–] godlessworm@hexbear.net 11 points 1 month ago

google wouldn’t lie. it’s math that must be lying.

[–] FumpyAer@hexbear.net 11 points 1 month ago (1 children)

ChatGPT uses the wolfram alpha font, makes me wonder if they steal its answers.

[–] invalidusernamelol@hexbear.net 2 points 1 month ago

It supports mathjax/MathML/LaTex formatting

[–] Hestia@hexbear.net 9 points 1 month ago* (last edited 1 month ago) (1 children)

AI is literally dumber than a basic 5 dollar calculator

[–] Le_Wokisme@hexbear.net 8 points 1 month ago (1 children)

the 4-function calculator will give you wrong answers if you use it wrong as well.

"using" an llm for this kind of math is good for exposing that they're not intelligent but it's not the right tool for the job if you actually want a maths answer. use wolfman alpha or whatever that is instead if it's actual math.

[–] invalidusernamelol@hexbear.net 2 points 1 month ago

Wolfram Alpha is just an expert system/1st gen "AI". It's closer to how the original search engines worked which is why it's so much better.

[–] TraschcanOfIdeology@hexbear.net 8 points 1 month ago

I'm saving this one for when people tell me "I asked [insert shitty chat bot] x]"

[–] miz@hexbear.net 8 points 1 month ago
[–] DragonBallZinn@hexbear.net 8 points 1 month ago

porky-scared-flipped: “Thinking for yourself? What’s wrong, poor!? Why do you hate muh innovooooshun?”

[–] heatenconsumerist@hexbear.net 4 points 1 month ago