#ai # [[Epistemic status]] #shower-thought #to-digest # Related - [[Computing/From prompt engineering AI to prompt engineering humans]] - [[Computing/Prompt]] - [[Computing/Intelligence/Machine Learning/Reverse engineering language models]] - [[Computing/The future of GitHub Copilot]] # TODO > [!TODO] TODO # Semantic prompt search The [[Philosophy/Rationality/Hack/Goal|goal]] of prompting a [[language model]] or a human is to produce some computation that will help to reach an outcome. When you want to know if Bob likes apples you will prompt some semantically similar strings in the [[Conversation|conversation]] and then when enough context is given you can ask the fatal question. Ok another example. you want to know John deep secret, you will start discussing the weather, then politics, then some more emotional topic, then goes the [[Big Talks]], you put enough context for this. So yeah, it’s the same for computers, you need to run some semantic search on the [[Information|information]] necessary to reach the desired outcome and insert it in the prompt of the [[language model]]