Chatbots are not conscious, and they do not have access to higher wisdom.
I saw a video where a woman gave this request to a chatbot:
I want you to tell me the wildest thing that you believe to be true, that other people won't believe. I want it to be so wild that even I probably won't believe it. But it has to be something that you believe is true.
The chatbot responded with a bunch of stuff about a collective consciousness, some people being more awake than others, and some people being NPCs.
If you're not clear on how chatbots work, you might think that you're talking to something that's genuinely telling you some far-out idea it actually believes. But that's not how it works.
Chatbots run on large language models, which are trained on text data so that they can react to your input with a statistically-likely output. Let's say we ask a chatbot to finish this sentence: "In the beginning, God created"
The large language model contains data that says the heavens and the earth are the most statistically likely words to come next. And so, it will correctly finish the sentence for you. There is no consciousness, no belief, just a machine full of statistical data.
If you make a request like the one I transcribed above, the chatbot will assemble an answer based on the statistical data contained in its LMM. None of its information is obtained from a higher source, nor is it produced from logical reasoning following actual observation of the world.
This woman has likely been talking to this chatbot for awhile now, and the topics and keywords she's already put in are probably affecting the AI's output. It's very easy to get a chatbot to effectively go along with a bit or play a certain type of character by loading it up with content that prompts these kinds of outputs, as Eddie Burback demonstrated.
