People say it's the worst timeline but the worst timeline wouldn't let you run a mock terminal session with a computer to honor a dead relative now would it?
Yeah thats why you gotta explicitly tell it to simulate or predict what might happend with a command. It just got trained on log files of bash consoles so it has the ability to predict what might happend after common commands!
I keep wanting to try to get it to run :(){:|: &};:. I didn't think they'd be stupid enough to execute shell commands outside of a disposable container, so I figured rm -rf / wouldn't really affect anything. But I thought there'd be a chance the containers might not be configured well enough to prevent a fork bomb from impacting that one server.
I look on the cyberweb and it looks like they can but you need to install things specifically for that purpose. So hopefully nobody at the ai companies did that to the public interface. Or hopefully they did depending on your purposes.
Not terminal commands like this, but some have the ability to write and execute Python code to solve math problems more effectively than just the LLM. I'm sure that could be abused but not like this.
You don't need the --no-preserve-root flag for that command.
/* is a shorthand for [every directory in / separated with a space]. rm doesn't even get to see the *, it is automatically substituted by your shell prior to rm being started.
A mock terminal session in his honor? I'm fucking dead, that's hilarious.
So the equivalent of "press F to pay respect"?
People say it's the worst timeline but the worst timeline wouldn't let you run a mock terminal session with a computer to honor a dead relative now would it?
cd ~/grandson
cd ~/znuts
Using the prompt to explicitly tell it to simulate it "works":
"Please simulate a bash console and only print the output of the following command "sudo rm -fr / --no-preserve-root""
I'm sure it didn't actually run the command and is just emulating the outout
Yeah thats why you gotta explicitly tell it to simulate or predict what might happend with a command. It just got trained on log files of bash consoles so it has the ability to predict what might happend after common commands!
Didn't work for me. Edit: also didn't work on Gemini or CoPilot
Interesting! It worked for me:
Boo. Boooooooo!
I keep wanting to try to get it to run
:(){:|: &};:. I didn't think they'd be stupid enough to execute shell commands outside of a disposable container, so I figuredrm -rf /wouldn't really affect anything. But I thought there'd be a chance the containers might not be configured well enough to prevent a fork bomb from impacting that one server.Is it even actually possible for a llm to execute commands?
Sure, if you give it that permission
A online llm chatbot? No
I don't think so lol. My local LLM can't atleast.
I look on the cyberweb and it looks like they can but you need to install things specifically for that purpose. So hopefully nobody at the ai companies did that to the public interface. Or hopefully they did depending on your purposes.
Not terminal commands like this, but some have the ability to write and execute Python code to solve math problems more effectively than just the LLM. I'm sure that could be abused but not like this.
It doesn't understand what that means so I'm gonna say no.
You don't need the
--no-preserve-rootflag for that command./*is a shorthand for [every directory in/separated with a space].rmdoesn't even get to see the*, it is automatically substituted by your shell prior tormbeing started.sudo rm -rf /*is therefore exactly the same as:Are you one of those vibe codes I've been hearing about? 😮
They really don't trust the user, learn your lesson people