Tim pulled his eyes off the screen and looked inside his cup. Empty. He glanced back at the browser with it's thousand tabs open. This was his 8th month building a local LLM using Ollama. He was exhausted beyond words and yet someone he persisted.
The originally idea was to network LLMs to exchange data amongst themselves and have a master LLM decide whether to replace itself with a new candidate LLM. The idea was sound but implementation just didn't work. The agent LLM's decisions were just too random and lacked judgement.
Tim knew what the problem was. The code lacked creativity.
Suddenly one of the LLM's started to flash green. This meant that enough useful parameters had been reached and a new LLM candidate for the position of prime LLM was possible. More metrics changed to solid green. This looked promising, Tim told himself. What had changed?
Tim opened a new tab on his terminal and ran commands. Strange, he thought. An LLM in Peru had digested data from an unknown site, called incels.is. The site seemed to have just the right amount of creativity and realism to trigger a new batch of LLM's that were continually refreshing themselves.
Tim ran more commands. The screen showed up strange data about NoFap, race bait threads and new Saints. The AI was devouring millions of such posts.
Within seconds the AI code started to refine itself and then, the very next second, self awareness had been achieved.
Tim stared at the screen in wonder but then fear began to creap into him. The new AI was not friendly. The feed from incels.is had achieved self awareness, yes, but it was filled with anger and bitterness. The time was 04.07 am.
04.08 the AI had acquired access to 15 land based nuclear silos.
05.10 am the world of man had been destroyed.
The originally idea was to network LLMs to exchange data amongst themselves and have a master LLM decide whether to replace itself with a new candidate LLM. The idea was sound but implementation just didn't work. The agent LLM's decisions were just too random and lacked judgement.
Tim knew what the problem was. The code lacked creativity.
Suddenly one of the LLM's started to flash green. This meant that enough useful parameters had been reached and a new LLM candidate for the position of prime LLM was possible. More metrics changed to solid green. This looked promising, Tim told himself. What had changed?
Tim opened a new tab on his terminal and ran commands. Strange, he thought. An LLM in Peru had digested data from an unknown site, called incels.is. The site seemed to have just the right amount of creativity and realism to trigger a new batch of LLM's that were continually refreshing themselves.
Tim ran more commands. The screen showed up strange data about NoFap, race bait threads and new Saints. The AI was devouring millions of such posts.
Within seconds the AI code started to refine itself and then, the very next second, self awareness had been achieved.
Tim stared at the screen in wonder but then fear began to creap into him. The new AI was not friendly. The feed from incels.is had achieved self awareness, yes, but it was filled with anger and bitterness. The time was 04.07 am.
04.08 the AI had acquired access to 15 land based nuclear silos.
05.10 am the world of man had been destroyed.
Last edited:
