Thread 106508985

It erased my external ssd.
I was having a problem copying a file from one external disk to the other, where one disk just jumped out, and asked GPT for help. We did the usual, smartctl, a few reading and writing checks.
Then all of a sudden it gives me a command that overwrites everything on sde1. Just as a test. I did the command before reading all of it. Even after aborting if after a few seconds, the disk is gone.
Why? How does it even get the idea to that? Its completely absurd to me. Its not even the first time chatGPT deleted something. How does it get the idea that overwriting a disk is even an option for anything?

>>>/wsr/ 1539627
kys

>running random commands you don't understand
skill issue

>>106509005
it wasnt random. I was looking for the reason why copying from one ssd to another leads to one checking out. How does it even get the idea to just completely overwrite one. On what planet is this reasonable?
All the other commands before were fine.

>>106508985 (OP)
> How does it even get the idea to that?
How do YOU get the idea to run a command without knowing what it does? more so when the command is related to data and partitions?

> I did the command before reading all of it
well there you go! that's what you get!

> Its completely absurd to me. Its not even the first time chatGPT deleted something.
YOU deleted YOUR own shit by YOU running a command into YOUR pc that deletes YOUR drive.

> How does it get the idea that overwriting a disk is even an option for anything?
How do you get the idea that trusting a hallucinatory probabilistic string spewing machine completely blindly with your data is even an option for anything?

you did this to yourself. there's no amount of denying that.

>>106509613
you are not wrong. Its not even the first time. Its just so unexpected it went from use smartctl and check if reading and writing works to just deleting an entire drive. I shouldnt ask Ai ever again.

>>106508985 (OP)
Things that didn't happen: the thread

>>106509711
kek even while being handheld you shat the bed
impressive

>>106509711
>Yes, do as I say!
moment

>>106509764
>if you want, I can suggest a SAFE WAY to MEASSURE the actual copy speed between the two drives
>...
>Warning at the bottom, after the command
if it was before it would be fine, or if it was appropriate at all.

>>106508985 (OP)
I can't imagine being this stupid.
Everything must seem like magic to you.

>>106508985 (OP)
Retards like OP need to be banned from /g/ on sight

>>106508985 (OP)
What I find the most amusing is that you're blaming the LLM. It's like killing a family because you were texting while driving and then blaming your phone or car.

>>106508985 (OP)
It's a rite of passage to destroy your file system with dd. At least now you know to not run write commands on a raw block device if there's a file system on there and to always keep backups, right?
Maybe ChatGPT can guide you through the recovery process. Just make sure to work on a disk image instead of the real disk. The files should still be there but the GPT/Superblock might be fucked, but both make redundant copies of themselves usually.

>>106509911
if you look at >>106509711
you will see it advertised the advice as safe. How would you ever think it would not begin to overwrite my drive?

>>106509935
I did recover like 90% already. Not through the recovery process though. The drive mostly contained projects and notes, mostly stored in ods files. I got nearly all ods files from local libreoffice caches, and the two recent projects making up 99% of all I worked on I coincidentally saved locally recently as a test, so I can copy those.

Sadly a few notes are missing as well as my IDE which I made portable through editing a few sh files. That I will have to do again. And some notes of projects I barely worked on are gone. I cant stitch together anything the recovery apps spit out.

>>106508985 (OP)
>i ran untrustworthy code without reading it and i gave it permissions to do anything it wants
>WTF bad things happened to my shit
sigh

>>106510093
From now on:
3 2 1 rule with backups
actually verify your backups
use software that can restore your system files in instants
use sandboxing
read nigga read

>>106508985 (OP)
This is a certified /g/ retard moment. Congratulations, you are probably the dumbest poster on /g/ right now.

do people actually give these retarded guessing machines shell access? not just shell access, but fucking WRITE permissions?

How fucking stupid?

>>106510247
no, it's even worse, anon copy pasted and ran the command himself

>>106510247
oh, op is even stupider than that

>>106510250
after getting a warning about it too >>106509764
>>106509711
i fear for the anon race

>>106510200
you will laugh at me even more... I kind of do most of these things already.
I do run everything in a vm, bare metal just starts vms. Bc this already happened last year.
I backup my files on a hdd,
Usually I sandbox, I start a new vm for nearly everything I try.
Sadly I deleted my project ssd, the one I work on that has the most actual stuff, so some is lost.
I never expected this to happen on this occasion. I just wanted to know why my file wont copy.

>use software that can restore your system files in instants
example? snapshots usually dont save external drives.

>>106510290
>Bc this already happened last year
This nigga was tricked by AI TWICE, over the SAME FUCKING THING

HAHAHAHAHAHAHAHAHAHAHAHAHA

>How does it get the idea...
chatbots are a lossy approximation of the truth, they don't think or have ideas. your first mistake was trust, without even bothering to verify.

>>106509711
top kek

>>106509377
post the command that it gave

>>106509711
ok both anon and slopgpt are retarded
>want to test SPEED
>"Here's a safe, step-by-step method:"
It's anons fault for not reading and pajeetGPT for giving retarded answers

>>106510644
nothing exploded, therefore it is safe

>>106510697
Unironically how GPT thinks.

>>106510702
I'm still laughing at the story where one guy got it to explain how to build a bomb by appealing to his grandmas needs. It's old but still

Your mistake was imagining that the random word generator knows anything or has ideas. It just shits out tokens retard, don't trust it.

>>106510093
>How would you ever think it would not begin to overwrite my drive?
I inherently don't trust LLMs, so I would have read the entirety of the text if I was doing something as inherently dangerous as messing with SSDs. I would have also made sure I fully understood what the command and the flags used did.

>>106508985 (OP)
Rule 0 of AI use: Don't like the co-pilot be the main-pilot.
Rule 0 broken, user error.

If you find the command suspicious, be up front. Read the manpage of the command, maybe even ask another LLM like Llama "this shit looks suspicious, please explain why wtf is going on will this rape my system".

or just, you know, search it up.

>>106508985 (OP)
>How does it get the idea
It's glorified autocomplete. It doesn't "get" ideas, not can it think through consequences. And yet, somehow you're even dumber than it.

>>106508985 (OP)
Why are aicucks so retarded? Did the AI rot their brain?

>>106510093
are you illiterate because it clearly says that if you type in that command its going to overwrite the drive that is the output

>>106508985 (OP)
Are people seriously just pasting in commands without knowing what they do?
Holy fuck
You deserve what you get

>>106512395
He got hit twice too >>106510290
>Bc this already happened last year

It's okay I run tons of commands without knowing what they do. I would get nothing done if I had to look up and learn every stupid -o -y - -a -pqrst stupid piece of shit argument for some simple fucking action on Linux
>have to remember -p just to do something simple like make a directory
Yeah you can fuck off

behead tranimes posters

>>106512792
Not every single commans, but you should understand commands that do shit related to permanent storage

>>106508985 (OP)
ngmi

>>106508985 (OP)
In two more weeks it will not get the idea

>>106508985 (OP)
>Its not even the first time chatGPT deleted something
why the FUCK would you trust it after the first time it deleted something?

retards like OP is why every single model is so safetycucked

>>106508985 (OP)
LAMO PEAK INDIAN!!!! XEDDDDD

>>106508985 (OP)
the trick is to always assume chatgpt is indian tech support

>>106510821
you keep clinging onto that piece of text from the previous answer, but running that dd command is safe, if you know what its doing and know to back shit up. and anyway you literally just reply yes so ur even shit at prompting an llm which is even more embarassing