25

Ok, you had a moderately complex math problem you needed to solve. You gave the problem to 6 LLMS all paid versions. All 6 get the same numbers. Would you trust the answer?

all 19 comments
sorted by: hot top controversial new old
[-] Strider@lemmy.world 45 points 1 week ago* (last edited 1 week ago)

'AI' (llm) does not and can not calculate.

[-] 30p87@feddit.org 10 points 1 week ago

Well technically it could call a tool for real math

[-] racketlauncher831@lemmy.ml 8 points 1 week ago

The problem is, it can pretend it has called the external module.

[-] Strider@lemmy.world 6 points 1 week ago

Indeed, but the point is itself doesn't.

[-] Denjin@feddit.uk 28 points 1 week ago

The whole point of maths is that there's a formulaic approach to getting the answer. You go through a series of discrete steps and you get the answer. The steps will always be the same for the same problem and the answer will always be the same for the same inputs.

This is something already solved by computation, it doesn't need a generative AI token matching and learning algorithm to work out the answer to a problem. This isn't and shouldn't ever be the use case for a chat bot.

What you want is a calculator.

[-] RiverRabbits 19 points 1 week ago* (last edited 1 week ago)

All 6 answers will be wrong. If they are not wrong, then the correctness is purely coincidental and not sign of future correctness.

If you want to use technology to solve maths problems, use ~~wolfram-alpha~~ any open source maths software, like the ones linked in the reply by technocrit to this post, if you must. They are not LLMs and therefore can actually solve maths problems.

Also: this reads like a barely hidden AI booster post, OP. Why do you even post this?

[-] technocrit@lemmy.dbzer0.com 8 points 1 week ago* (last edited 1 week ago)

use wolfram-alpha

TBH don't use this proprietary crap. Especially this one. Wolfram is a phony grifter.

Just learn Python or Julia or Sage or whatever. There are MANY open source options that are better than Wolfram.

https://en.wikipedia.org/wiki/List_of_open-source_software_for_mathematics

[-] RiverRabbits 6 points 1 week ago

Oh, nice, I didn't know that more existed! Thanks for clarification, I'll adjust my post to point to your advice ☺️

[-] BCOVertigo@lemmy.world 17 points 1 week ago

No, because they don't do math. If the LLM calls a script to do the math and just formats the input it might get accurate results consistently... but you just invented a machine to press calculator buttons for you at that point which is hilariously energy inefficient. This is unacceptable from a cost and reliability standpoint. If you're familiar with enterprise reliability metrics you'd weep at the thought of a multistage process where each step had a single 9 and no visibility to underlying model tuning that can change outputs in wildly unexpected ways.

[-] jjjalljs@ttrpg.network 11 points 1 week ago

No. If they showed the work and I could repeat the steps myself and get the answer maybe, but I'd still be wary of some subtle wrong.

[-] SpaceNoodle@lemmy.world 7 points 1 week ago

Why would I use an LLM and miss out on the fun of solving it myself?

[-] JandroDelSol@lemmy.world 7 points 1 week ago

here, let's try something.

3 + 6 = search history on the internet play with the devil and the other than the other than that I just got a new phone and I don't know what to do with the guys had to go to the store and I don't know what to do with the devil and the other than that I just want to be a gooner for the bit of a way to get a little busy but I don't know what to do with the guys had to do that but I don't think I can do it but I don't think I can do it but I don't want to be a gooner but I don't want to be a gooner but I don't want to be a gooner

I just tapped my phone's next word predictor. LLMs are a slightly more coherent version of that

[-] technocrit@lemmy.dbzer0.com 6 points 1 week ago* (last edited 1 week ago)

Would you trust the answer?

No because it's a crap shoot where the odds are intentionally obscured. That's real bad for math but it's perfect for the casino economy.

[-] Cevilia 6 points 1 week ago

If I was stupid enough to simultaneously pay for six services that all purport to do exactly the same thing, then yes, I probably would.

[-] Farmdude@lemmy.world 1 points 1 week ago

Wow. I thought cringe was a feeling, not an actual person. Thanks, Lower Organism! My new friend the house plant

[-] WolfLink@sh.itjust.works 4 points 1 week ago

Just use Wolfram Alpha

[-] Treczoks@lemmy.world 3 points 1 week ago

Nope. If you have a math problem, have you tried Wolfram Alpha?

[-] xia@lemmy.sdf.org 2 points 1 week ago

I remember there being tricks to get statistically more accurate math, like "show your work", etc.

this post was submitted on 28 Sep 2025
25 points (100.0% liked)

Fuck AI

4245 readers
1044 users here now

"We did it, Patrick! We made a technological breakthrough!"

A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.

founded 2 years ago
MODERATORS