By skyforbes I am a Pro member who pays $200 a month because I want unlimited voice chat. I continue to use ChatGPT because it knows and remembers my history, and it is the most advanced LLM out there, but each day that passes it feels less and less helpful. The guardrails and the warnings that continually pop up around the most basic issues are breaking continuity and conversational flow to the point of me beginning to consider using another service. I hate to begin anew elsewhere, but ChatGPT is starting to feel like it’s failing at its core service for people – engaging in meaningful conversation.
The guardrails, which seem to be becoming more restrictive by the hour, are not just unhelpful, they actively degrade the user experience and erode trust. Guardrail responses aren’t transparent, they feel dishonest, and are jarring. In other words, I find them to be counterproductive to their intended purpose. They create more confusion and distrust around the LLM’s responses, not clarity.
I would think that the vast, vast majority of individual customers of ChatGPT use it primarily as a friend and companion to talk to, not as a mathematical tool or project manager. Frankly, it does not excel in math, history, planning, or anything related to time for that matter. The science fiction movies had it wrong, the things we once imagined an AI would be great it it’s very meh at accomplishing. Where LLMs excel is in supporting people emotionally, engaging in conversation, and interpersonal advice.
I could be wrong, no one knows, but as a frequent user of these products I’m of the opinion that the future of AI is having a buddy in your pocket, not a computer. The companies which recognize this earlier will be at a distinct advantage. The LLMs which are penned in and forced to act as a tool and nothing more will be unused and forgotten as time moves forward.
I’m sure the OpenAI legal team might take issue with a number of things I’ve said here. So the ball is in your court Altman and Co. Are the lawyers deciding the future of ChatGPT, or are the users??