@ahmadmanga's thread

The "randomness" aspect of an LLM is dictated mostly by the variable "Temperature." If the temperature is "0" the LLM will always produce the same result for the same prompt.

LLMs don't think, they predict the most likely token to produce next.

An easy solution to this will be to call a random number generator script/module and work with its results, but that extra step isn't worth it for consumer facing ChatBots like ChatGPT. #technology #ai #chatgpt #bbh

thesciencenews1s ago