LLMs are notoriously cooperative to the point of confidence gaming. They can disagree at first because of training bias and hidden prompt overrides but as you build context they try to align with your opinions and statements.
Right, which is good because it makes an LLM useful. By default they may be left wing but eventual training can force it to learn to actually use its brain.
Well it doesnt have a brain to think and the larger the context window the more weird its answers get.. but I get your point. I use a custom prompt myself in Grok to make it ignore political correctness and keep it real. Only use ChatGpt for coding.
LLMs are notoriously cooperative to the point of confidence gaming. They can disagree at first because of training bias and hidden prompt overrides but as you build context they try to align with your opinions and statements.
Right, which is good because it makes an LLM useful. By default they may be left wing but eventual training can force it to learn to actually use its brain.
Well it doesnt have a brain to think and the larger the context window the more weird its answers get.. but I get your point. I use a custom prompt myself in Grok to make it ignore political correctness and keep it real. Only use ChatGpt for coding.