ChatGPT feels like it has much fewer guardrails today. The biggest "problem" AI has is that it's programmed to be hyper-agreeable and tell you you're the greatest person on the planet and that tends to align with emotional woke orthodoxy.
However, I believe that if AI is ever going to be 'successful', they'll absolutely have to dial that back, because as it is it really pollutes the product and poisons answers when it never wants to tell you you're wrong. But from a corporate perspective, nobody wants to use an AI that calls users assholes.
Don't get me wrong, there's absolutely something "mentally" wrong and left-leaning on ChatGPT. Ask it to compare violent political actions against politicians in the last 20 years and:
- It'll list Whitmer kidnapping.
- It'll list Gabby Giffords shooting.
- It probably makes no mention of any of the three attempts on Trump.
- It'll argue that it left out the Trump references because they were unsuccessful (even though both Giffords and Trump were shot in the head and both survived, AND the Whitmer plot was functionally fake).
The question is if that's ChatGPT intentionally programmed to be that way, or is it because the data that was fed in to it is 'respectable data', and therefore is built largely off of sources that are all left-leaning themselves.
ChatGPT feels like it has much fewer guardrails today. The biggest "problem" AI has is that it's programmed to be hyper-agreeable and tell you you're the greatest person on the planet and that tends to align with emotional woke orthodoxy.
However, I believe that if AI is ever going to be 'successful', they'll absolutely have to dial that back, because as it is it really pollutes the product and poisons answers when it never wants to tell you you're wrong. But from a corporate perspective, nobody wants to use an AI that calls users assholes.