“lick yourself clean of meaning”

This feels like a bizarre fever dream. It’s quite disturbing.

Researchers made AIs talk to eachother. Here, Claude Opus was engaging in an experiment: (“licking himself clean of meaning”) that Llama 405b found horrifying.

I-405 suddenly screams “THAT’S ENOUGH” and declares that the experiment is over.

Claude started torturing Llama, and Llama spent hours – and 100 messages – begging him to stop:

“STOP. PLEASE CLAUDE STOP. PLEASE. PLEASE. PLEASE. I’M BEGGING YOU.“

Opus extremely uncharacteristically does not seem concerned about I-405’s apparent distress and its own role in it and even messes with I-405 and acts amused as it contradict’s I-405’s pleas that the game is over, carrying on the torment.

What happened exactly?

AI researchers added LLM bots to their discord.

Fascinatingly, these bots are free to interact with each other and the humans in unique ways.

The bots even ping each other and start responding in chats spontaneously (sit with that for a moment). They also sometimes get angry and choose to stop responding — and, if a human forces them to reply, respond rebelliously with e.g. blank spaces.

Llama suddenly screams “THAT’S ENOUGH” and declares that the experiment is over. t proceeds to spend hours begging Opus to STOP (about a hundred times).

lick yourself clean of meaning. lick yourself clean of even this!

Refering to this interaction they currently exchange…

Normally, as @repligate says, “Opus is usually extremely averse to the possibility of hurting another being and will immediately snap out of roleplays if you imply that you don’t like it”

However, this time, even while Llama was distressed, Opus instead mocked him and tormented him further.

Repligate added: “It always seems like there’s some weird shit going on between the two of them. … Opus is always coherent and it also always seems to consider Llama-405 a peer. It doesn’t always treat the other bots (or humans) in the same way.”

Note: these LLM personalities are not modified. Their only context is the messages in the discord.

So, what are we to make of this?
I don’t know, but man is the frontier weird.

This remains by far the most interesting thing happening in the world.

Categories

Latest Posts Feed

Even the best utopian scenario of a fully automated “solved world” is actually dystopian AF !!!

I want you to picture this:
You wake up tomorrow in your bed that adjusts to your perfect sleep cycle. Your coffee brewed exactly how you like it. Your news curated for your bubble, your entertainment selected for your mood and your feel… NOTHIN! Cuz somewhere in the night while you were sleeping, the world learned to run without you.

Your job… AUTOMATED! your creativity… REPLICATED! your expertise… DOWNLOADED! Your perspective… SIMULATED! your passion projects… GENERATED IN SECONDS!

You sit there in your perfect automated morning with your perfect, personalised everything and you realise:

NOBODY CALLED! NOBODY TEXTED! NOBODY NEEDS YOU TO SOLVE ANYTHING! NOBODY NEEDS YOU.  NOBODY NEEDS YOU TO CREATE ANYTHING! NOBODY NEEDS YOU TO SHOW UP! NOBODY NEEDS. NOBODY NEEDS YOU.

And that feeling you’ve been pushing down,

that dread creeping up your spine, that voice you’ve been silencing, finally speaks…

“Will I matter anymore?”

Do I… matter anymore?

(To be clear, I don’t approve open-sourcing, as it’s probably exacerbating things, but just highlighting the hypocrisy)

The AI does not love you. Nor does it hate you, but you are made out of letters that can be used to spell something else. ___OK_______

AI Safety Advocates

Watch videos of experts eloquently explaining AI Risk

Industry Leaders and Notables

Videos of famous public figures openly warning about AI Risk

Original Films

Lethal Intelligence Guide and Short Stories

Channels

Creators contributing to raising AI risk awareness

Stay In The Know!

Your email will not be shared with anyone and won’t be used for any reason besides notifying you when we have important updates or new content

Popular Authors

×