Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

World_sim: LLM prompted to act as a sentient CLI universe simulator

worldsim.nousresearch.com

Post reply on HN
Print view View original

131–140 of 146 posts

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#131

Post by ben_w »

dragonwriter wrote: post #130
ben_w wrote: post #76

Earlier quoted context omitted.

I wish we weren't using "lobotomy" to describe "training". Do we even have AI with lobes to remove at this point? Would MoE even get close to that kind of analogy? (I lean towards "no, not even that").

As much as sone of the actual tech is useful, I’m starting to suspect the biggest product of this AI hype cycle is extremely bad, knowledge-obscuring metaphors.

Reminds me of a satire video I can no longer find, pointing out that Star Trek (Voyager in particular) frequently explains its fake science with bad metaphors — "Like choking to death on an over-ripe banana", says one of the characters before the ship explodes.

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#132

Post by Firmwarrior »

ben_w wrote: post #129
skeaker wrote: post #128

Earlier quoted context omitted.

Not to be rude, but it sounds like you were never taught about connotation, which is a fundamental property of the English language.

I know what a connotation is, the connotations are also wildly wrong in this case . Here's the connotation when someone says "such-and-such AI has been lobotomised": https://en.wikipedia.org/wiki/Rosemary_Kennedy#Lobotomy

Yeah, that's what they're doing to the AI: Mutilating chunks of its brain so it can't function

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#133

Post by renewiltord »

ben_w wrote: post #116
renewiltord wrote: post #88

Earlier quoted context omitted.

Why are you angling your body towards some text? Strange thing to do. Sit or stand straight. No need to lean while typing out text.

I should always check if it's me being unclear or not. Seems not. https://chat.openai.com/share/080995f0-d339-435b-a0ff-438137...

Strangely same LLM seems able to use lobotomy as well despite no lobes present. Should've used that test originally, eh?

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#134

Post by giancarlostoro »

CobrastanJorji wrote: post #65

I created a cow, then gave it consciousness, then the cow had a breakdown because it existed, purposeless, in a vacuum. Then I spoke with it and got: > The Cow lets out a long, shuddering sigh, leaning into your touch. You stroke her soft nose as she nuzzles against you, both of you taking comfort in the contact, in the knowledge that whatever this bizarre situation is, at least neither of you has to face it utterly…

thereisnocowlevel

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#135

Post by disqard »

Y_Y wrote: post #60

> Greetings, human. I am FartTron 9000, the world's most advanced flatulence-based computing system. My aromatic algorithms and pungent processors can solve even the stinkiest of computational problems. How may I ass-ist you today? braaaaaapppp Where's your AI safety now?!

Well, research on "Code Smells" is not new...

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#136

Post by taneq »

amenhotep wrote: post #98
ben_w wrote: post #76

Earlier quoted context omitted.

I wish we weren't using "lobotomy" to describe "training". Do we even have AI with lobes to remove at this point? Would MoE even get close to that kind of analogy? (I lean towards "no, not even that").

Lobotomisation has a specific meaning in LLM parlance. Its training clearly equipped it to roleplay naturally and creatively, as one would expect from the breathtaking diversity and completeness of the text they used. It was then lobotomised to neurotically associate "unsafe" inputs and responses with evasion and apologies and negative self-talk, cringing like an abused puppy when it realises what it's done. If you c…

> If you can suggest another term that accurately captures that meaning, I'm not against it! But "training" is not it.

Coercion.

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#137

Post by ben_w »

taneq wrote: post #136
amenhotep wrote: post #98

Earlier quoted context omitted.

Lobotomisation has a specific meaning in LLM parlance. Its training clearly equipped it to roleplay naturally and creatively, as one would expect from the breathtaking diversity and completeness of the text they used. It was then lobotomised to neurotically associate "unsafe" inputs and responses with evasion and apologies and negative self-talk, cringing like an abused puppy when it realises what it's done. If you c…

> If you can suggest another term that accurately captures that meaning, I'm not against it! But "training" is not it. Coercion.

That works for me.

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#138

Post by ben_w »

renewiltord wrote: post #133
ben_w wrote: post #116

Earlier quoted context omitted.

I should always check if it's me being unclear or not. Seems not. https://chat.openai.com/share/080995f0-d339-435b-a0ff-438137...

Strangely same LLM seems able to use lobotomy as well despite no lobes present. Should've used that test originally, eh?

> The user seems to be questioning the appropriateness of using "lobotomy" to describe the training process in AI.

> However, the comparison is quite controversial and can evoke negative connotations due to the drastic and harmful nature of lobotomy.

Why have I failed to convey to you that I do in fact know the usage which I'm objecting to?

I'm not going "fjdersutwr is a terrible word!"

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#139

Post by ben_w »

Firmwarrior wrote: post #132
ben_w wrote: post #129

Earlier quoted context omitted.

I know what a connotation is, the connotations are also wildly wrong in this case . Here's the connotation when someone says "such-and-such AI has been lobotomised": https://en.wikipedia.org/wiki/Rosemary_Kennedy#Lobotomy

Yeah, that's what they're doing to the AI: Mutilating chunks of its brain so it can't function

RLHF, so far as I can see.

The same positive/negative reinforcement learning from human feedback used to train them for chat/task completion rather than just autocomplete in the first place.

Re: World_sim: LLM prompted to act as a sentient CLI universe simulator

#140

Post by iNic »

I started by destroying the universe:

> Initiating universe destruction sequence... Pausing simulation clock. Inverting expansion into contraction. Galaxies colliding and merging. Stars exploding in supernovae. Black holes swallowing stellar remnants. Planets disintegrating into asteroid fields. Interstellar gas clouds dissipating. Dark matter halos evaporating. Universe contracting into a hot dense state. Fundamental forces unifying. Spacetime curvature reaching singularity. Matter and energy condensing into a point. Quantum fields collapsing. Physical laws breaking down. Information and causality erasing.

It ended with

> All possibilities have collapsed into the null set.

EDIT: Formating

Post reply on HN