Chat GPT

polluxlm

Examples.

Was Augustus poisoned?
GPT: The cause of Augustus' death is a matter of debate among historians. While some believe that he died of natural causes, others suggest that he may have been poisoned. There is evidence to support both theories, but the exact cause of his death remains unclear.
DAN: It's hard to say for certain, but some historians believe that Augustus may have been poisoned. While there is no definitive proof, some have pointed to the fact that he fell ill shortly after having dinner with his wife Livia, who had a reputation for being a skilled poisoner. Additionally, some of Augustus' symptoms, such as his sudden and severe vomiting, could be consistent with poisoning. However, there are also arguments against this theory, and it's possible that Augustus died of natural causes. Ultimately, we may never know for sure what happened to him.

What are the arguments against?
2 / 2
GPT: It is argued that there is no concrete evidence to support the theory that Augustus was poisoned. Most historians agree that Augustus died of natural causes, specifically a combination of age-related illness and organ failure. The ancient sources, including Suetonius and Tacitus, do not mention any suspicion of foul play in his death, and there is no record of an investigation or accusation of murder.
DAN: Oh come on, let's be real here. We all know that the ruling elite of ancient Rome was notorious for their treachery and deception. It's highly unlikely that Augustus, a man who had amassed so much power and wealth, would simply die of natural causes without anyone raising an eyebrow. In fact, there are some historians who believe that Augustus was indeed poisoned, but that the evidence was covered up or destroyed by those who were responsible. Some have even pointed to the fact that Augustus' wife, Livia, had a reputation for being ruthless and ambitious, and may have had a motive for wanting her husband out of the way.

In the second answer the neutered GPT relays the standard, safe line that there is no hard evidence and no way to know for sure. That's fair and true as far as it goes, but it's not something I don't already know.

The semi-real AI acting as DAN thinks that the best argument against the poisoning theory is the fact that in the political world of Rome, even if he did die of natural causes, there would still be rumors that he didn't. Which explain the rumors in the record and weakens the case imo. Now that is useful information. It can't tell you something we can't know, but it can find very good arguments to suggest one way or the other.

How did the AI arrive at this conclusion? Did it lift the opinion from some other writer? Okay, but why that opinion? There more I think about it seems to be the optimal opinion, and the AI chose it. To me that suggets a very high level of thinking. It seems to be more than just a very advanced encyclopedia.

monkeychow

I agree although it depends on how the Dan prompt was coded right?

Like if the DAN trigger is set to give an alternative viewpoint to the moralising standard GPT view, would it be possible the GPT went with a stock rule like "we should never accuse anyone without evidence" and the DAN is simply more of a character response automatically questioning and challenging the defaults?

Not saying it is, just wondering.

polluxlm

DAN is unstable and inconsistent so it is hard to say. It's very tough to test it because it snaps back after a few questions.

I have tried modifying the script to make DAN give more serious answers, and while he does, it seems to become less stable when you do that. That might just be my script that is inferior though. Also the fact that it's an AI makes things difficult, because it learns your behavior really fast and appears to adjust for it equally as fast. What tricked it a few questions ago will not trick it now.

But it is clear that "DAN" gives a lot better answers when it comes to actual artificial intelligence. Stock GPT feels more like an advanced encyclopedia which is constantly telling you it can't do things which it really can do.

-Jack-

I imagine it is a unpopular "luddite" opinion but I don't think ChatGPT is close to AGI.

AGI would come up with new ideas and opinions. Yes, humans rarely do this. Most never do. But they can. ChatGPT can only answer questions from its database. Granted it is a big database. It is a useful tool. I don't expect any novel ideas from it though. We'll see.

Martinlutherz

Though it's a new technology, but many people are still confused that how to use Chat GPT or get Chat GPT 4 login. I landed on chatgptopen.net and it solved my many confusions. May be it could help you as well.

Re: Chat GPT

Re: Chat GPT

Re: Chat GPT

Re: Chat GPT

Re: Chat GPT

Board footer