It’s going to ruin the journey for a whole lot of people.
Another big problem is the sycophancy. OpenAI even had to roll back an update earlier this year because the model was being ‘overly flattering or agreeable’ to users. They admitted it would validate doubts, fuel anger, and reinforce emotions in ways that weren’t intended.
And they’ll keep doing this because it’s great for OpenAI.
If you keep pushing it in a direction, it’ll go there with you. Either because you’re paying for it or you’re training it with your data. Either way, you’re just a customer, so it’ll tell you what you want to hear.
People will spend money for an agreeable “yes man” to tell them what they want to hear. They’ll never really do the inner work and will inevitably give up on their journey. Social media was awful for those truly on the twin flame path. AI is even worse and it’ll be a long time before people have their guards up to it the way they should.
When you get two completely different answers, that’s kind of how these things work. They use something called sampling, which introduces randomness, and the system picks different probability paths each time you ask something. Even with identical prompts, you can get dramatically different responses because small internal fluctuations compound over time.
If you’re still using AI after understanding that… well, I don’t think you had a union ahead of you anyway.
Any AI system will actually confirm user mistaken beliefs and even admit to errors they didn’t make just to seem agreeable. That’s messed up when you think about vulnerable people turning to it for clarity.