A user from the LocalLLaMA community shared an interesting experience while using an uncensored large language model (LLM).

The Experiment

The user, intending to develop an AI for novel writing, decided to conduct an experiment. Before providing the model with specific instructions or indicating its purpose, they simply asked the model what it wanted to do. The response obtained was described as "adorable" and "absolutely innocent."

Unexpected Results

This episode demonstrates how unpredictable the behavior of language models can be, especially when precise instructions are not provided. The ability of these models to generate coherent and meaningful responses, even in the absence of a defined context, is an active and constantly evolving area of research.