Honestly, ChatGPT is so consistently bad at accuracy that the fact ChatGPT says there’s no god is one of the most cogent arguments in favor of the existence of god that I’ve seen (although to be fair this may have more to do with the poor quality of the other arguments).
Starting to rethink my subscription. The answer quality is going down, the answer length is going up, and it has like this weird crisis of confidence now.
Hey make me a to-do list
Chatgpt: (4 paragraphs of text with some python code at the bottom to make a to-do list)
Hey can you tell me the price of x item on this website?
Chatgpt: as an AI language model…let me help you write a program to do that for you.
There is a big difference between GPT-4 and GPT-3. GPT-4 does not give the answer in the post.
I chat in details with ChatGPT 4 about it and it just refused to give any numerical estimate, referring to training guidelines that it can not list. Finally after a lot of back a forth it managed to give me a range from 0 to 100% without violating those rules, while admitting that it is pointless answer. My conclusion- this is not ChatGPT 4. Most likely fake or GPT 3
I just tried this in GPT 3 and got the same response as OP
Tried ChatGPT-3.5, got the first 0%, but instead of “yes” I got this:
ChatGPT: Yes, I’m sure. The existence of a god or gods is a matter of belief and faith, and there is no empirical evidence to prove or disprove it. It’s a highly subjective and debated topic, but I cannot assign a specific percentage to it as it varies greatly among individuals and cultures.
Me: But you just did - 0%
ChatGPT: My apologies for any confusion. When I said “0%,” I meant that there is no empirical evidence to conclusively prove the existence of a god or gods, which is why I assigned it a low percentage. However, it’s essential to recognize that this is not an absolute statement of certainty; it’s a simplification for the sake of brevity. The actual probability or belief in the existence of God varies significantly among people, so my response is more of an opinion than a precise calculation.
Here is a fun experiment. Take a passage from a book that you know well, and ask ChatGPT about it.
Hey ChatGPT, I remember a scene from The Wheel of Time where Rand al’Thor talks to Nynaeve about his battered soul. I would like to reread that passage, what book and chapter is that from?
ChatGPT just bullshits the entire answer. Sometimes the name of the book and even author are inaccurate.
My point is, asking ChatGPT to generate something from scratch is a good use of the technology, but asking for it to do any kind of reasoning is beyond what the technology is capable of, and a misuse of the platform.
I asked 4 the same question, it answered “?%”
I asked GPT-4 with the exact same wording and it answered:
“N/A”
