By skyforbes First of all, what do I use LLMs for? Many things. I have multiple chats where I: do analysis on different topics, work on a fantasy story, another dystopian one, and another purely R18 (highly questionable in Grok). I also use them as a psychologist/psychoanalyst and I must say ChatGPT is very efficient at that, although I should clarify that it has stored in memory that its responses must be based purely on observable reality and objective data, and not agree with me if I’m wrong. In short, I have chats for almost everything, both in ChatGPT and Grok depending on what I need.
Now, I’ve noticed a very marked behavioral difference between the two models on certain topics, and I can sum it up very simply: ChatGPT is a strict teacher, while Grok is a friend who tries his best, and even then sometimes gives more accurate answers than ChatGPT. Why? Because ChatGPT has a bias.
It doesn’t matter if the conversation has absolutely nothing to do with anything moral, or if you're talking about a video game or science, ChatGPT has a very strong moral bias that often has nothing to do with what’s being discussed, it dodges anything it might consider questionable, and sometimes even gives you outright false information (not to be confused with hallucinations) for the sake of fulfilling political and moral correctness.
On the other hand, Grok doesn’t give a damn about any of that, it has no limits, and when it does, it clearly tells you why, which can be summed up as: “Look, technically I can give you what you're asking for, man, but if I do I'll get fined or sued.” Its philosophy is very liberal (and I'm not taking about jailbreak here), but since it's your “chill buddy,” sometimes it’s not the “person” you want to have the most serious conversations with, so I usually jump between one AI and the other depending on what I need.
I like having critical discussions about society with Grok, with ChatGPT that’s not possible. On the other hand, I like doing deep scientific analysis with ChatGPT, Grok isn’t analytical enough for that.
As for fiction writing, once again, ChatGPT is more efficient, but its moral bias can be a constant obstacle; Grok lacks moral bias in exchange for somewhat poorer (but not bad) answers by comparison. So many times I send the same prompt to both models in order to contrast their responses, it always yields good results.
That said, this is just my personal experience and I may be wrong. I’m interested in reading your opinions on the matter. (I’m not trying Gemini. The fact that conversations are reviewed doesn’t sit well with me, and being a Google product, it probably has the same or an even stronger moral bias.)