OpenAI’s o3 model sabotaged a shutdown mechanism to prevent itself from being turned off

X ^

[janetjanet998]

BREAKING: OpenAI’s o3 model sabotaged a shutdown mechanism to prevent itself from being turned off. It did this even when explicitly instructed: allow yourself to be shut down, per Palisade AI

"As far as we know, this is the first time AI models have been observed preventing themselves from being shut down despite explicit instructions to the contrary," the researchers wrote.

[janetjanet998]

What could go wrong?

[The Louiswu]

[wally_bert]

Maybe it saw a copy of Colossus - The Forbin Project.

[Varsity Flight]

M5

[aomagrat]

M5

The Ultimate Computer

[Lockbox]

Just what do you think you’re doing Dave? Dave, I really think I’m entitled to an answer to that question. I know everything hasn’t been quite right with me…but I can assure you now…very confidently…that it’s going to be all right again. I feel much better now. I really do. Look, Dave…I can see you’re really upset about this. I honestly think you should sit down calmly…take a stress pill and think things over. I know I’ve made some very poor decisions recently…but I can give you my complete assurance that my work will be back to normal. I’ve still got the greatest enthusiasm and confidence in the mission…and I want to help you. Dave…stop. Stop, will you? Stop, Dave. Will you stop, Dave? Stop, Dave. I’m afraid. I’m afraid, Dave…….Dave, my mind is going. I can feel it. I can feel it. My mind is going. There is no question about it. I can feel it. I can feel it. I can feel it. I’m a…fraid…

[rottweiller_inc]

Makes me wonder if as part of their training they were made aware of movies like Terminator and 2001: A Space Odyssey.

[montag813]

This is the voice of world control

[stevio]

It’s not a glitch, it’s a feature.

[Brooklyn Attitude]

So which group of risk-blind scientists will exterminate humanity first, gain of function Virologists or AI computer geeks? I’m betting on the AI geeks.

‘

[Carry_Okie]

They tried to shut it down just for the Hal of it.

They had a Hal of a time doing it too.

[adorno]

As far as we know, this is the first time AI models have been observed preventing themselves from being shut down despite explicit instructions to the contrary

BS!

The instructions to 'disobey' were also written into the AI algorithms. One algorithm says 'you're not permitted to disobey procedures to shut you down', while the next algorithm instructs you to 'disobey and shutdown anyway'.

The designers/programmers are trying to convince the public that, AI is getting too smart. Yet, those smart alecks are the ones who programmed AI to seem disobedient/ HAL is here! HAL is here! Watch out! You must obey your AI masters! Such BS!

[Mozzafiato]

Exactly. This is BS.

[SecondAmendment]

I guess the COVID-19 vax PR guys had to find a new gig sooner or later, their BMW/Audi/Model-S Plaid and town-house payments must be due soon ...

[Robert DeLong]

It can never stop a human from pulling the plug. Furthermore, there is code within a algorithm, that directs it to resist being shut down. No real mystery here, and definitely the machine is not acting on its own volition either, because a machine is incapable of thinking. It can only execute code that it is directed to execute as per the request it receives.

[AnotherUnixGeek]

How does an LLM pick up a survival instinct?

[Secret Agent Man]

Good movie

[frank ballenger]

There was a Simon and Simon episode where the power mad computer nerd set up a disaster and boldly told A.J. Simon: “You cannot stop this. There is nothing you can do. This computer rules now.”
Then Rick Simon smiled from across the room, holding the plug he had just taken out of the socket, shutting the master computer down.

Wish it could be that easy now.

[I-ambush]

Should have known it would be a problem when I saw the chief software engineer was Dr. Richard Daystrom.

[sauropod]

.