LLMs build on top of the context you provide them and as a result are very impressionable and agreeable. It’s something to keep in mind when trying to get it to come up with good answers as you need to carefully word questions to avoid biasing it.
It can easily create a sense of false confidence in people who are just being told what they want to hear, but interpret that as validation, which was already a bad enough problem in the pre LLM world.
LLMs build on top of the context you provide them and as a result are very impressionable and agreeable. It’s something to keep in mind when trying to get it to come up with good answers as you need to carefully word questions to avoid biasing it.
It can easily create a sense of false confidence in people who are just being told what they want to hear, but interpret that as validation, which was already a bad enough problem in the pre LLM world.
In other words, it’s really easy to make your AI chatbot join in your echo chamber.