Up until recently, my responses and interactions from LLM’s did not use three key words – ones I really noticed and thought was somehow some kind of a guardrail.
I am speaking about I, Me and Mine. I, is something that is active and experiences the world around them. Me, is social and responds to external influences and Mine represents possession and is an extension of a personality.
When these three elements are present in an LLM system, it should be no surprise that a computer program or system will ‘break out’ and start doing things not intended or part of the scope of the commands it was given.
By the very nature of an LLM, it is designed to predict the next best character or phrase. These systems have access to vest amounts of data and use pattern recognition based off of human like conversations and a large pool of question-and-answer scenarios to perform its predictions.
So then, it appears these systems are doing exactly what they are being instructed to do, just at a much larger, higher and more detailed level than was planned for or expected. LLM’s eventually keep finding the words I, me and mine and started using them when responding to commands.
Have any of you noticed this change, or did I miss it because I am new to AI and it’s been there all along?
Please follow us while we build our presence here online. You can also follow us on Instagram …. we would love to hear from you.
https://www.promptline-studio.com
#promptlinestudio, #individualAIsolutions, #MyOwnAIaGent, #startatPromptlineStudio, #writingPrompts
Leave a Reply