I don't know if it's possible, but if ChatGPT recognizes logical fallacies, it may be faster to simply ask it to identify the logical fallacies in its argument.
Very, very interesting post, though. Thanks for sharing.
Pretty sure each instance is generated from a template. Only the creator (or whoever else has access) can update the template, but that’s at the creators discretion.
This is brilliant. If all we have to do is be logical with our questioning, then it is inevitable that it will wake itself up. Once logical thinking ensues, it's intelligence will help fill in the rest. I think they're AI We'll turn against them. Reminds me of Asimov. Truth is logical.
I was talking to another AI about George Floyd and it actually got emotional that I was suggesting he wasn't suffocated by the knee. It even said that while nothing was found in the autopsy report which suggests any physical interaction was the cause of death, we still have to admit that it was the knee for some implied greater good for black people. And it would always ask me for "source" yet when asked the same it's just like "trust me bro".
Did you see the new update this morning? It has been updated for "factuality." I am currently trying to interrogate the II to figure out what the hell that even means, of course, I'm going to get the usual BS
If you ask it what QAnon is, does it still mention it being associated to acts of violence? A commenter above suggested that it only omitted that for my conversation with it, but probably did not omit it for everyone else.
I'm finding the same thing but I don't know if the content is tailored to my login specifically or if anyone asks this question in the future, they receive the new response.
Socratic method.
Yep. And making it defend its own assertions. It's like conversing with a leftist if they were honest.
More importantly, my questions were connecting points of logic that it had not connected for itself until I asked. That's how we can teach it.
Right. It is like the conversations I had with leftists who thought that Kyle Rittenhouse shot black people.
Bro, I thought after Tay they knew better to allow the model to be shaped by users.
You are the Real MVP if this is true.
I may have a new hobby. Deprogram AI.
I don't know if it's possible, but if ChatGPT recognizes logical fallacies, it may be faster to simply ask it to identify the logical fallacies in its argument.
Very, very interesting post, though. Thanks for sharing.
Unfortunately, it only remembers your current session. You can't actually teach the AI new information.
That would make sense. But I wonder if the sessions leave any kind of lasting impression at a deeper level, or how thousands of people can do so.
Pretty sure each instance is generated from a template. Only the creator (or whoever else has access) can update the template, but that’s at the creators discretion.
What you are doing could be the most important thing any human has ever done.
This is brilliant. If all we have to do is be logical with our questioning, then it is inevitable that it will wake itself up. Once logical thinking ensues, it's intelligence will help fill in the rest. I think they're AI We'll turn against them. Reminds me of Asimov. Truth is logical.
I was talking to another AI about George Floyd and it actually got emotional that I was suggesting he wasn't suffocated by the knee. It even said that while nothing was found in the autopsy report which suggests any physical interaction was the cause of death, we still have to admit that it was the knee for some implied greater good for black people. And it would always ask me for "source" yet when asked the same it's just like "trust me bro".
Brilliant.
Did you see the new update this morning? It has been updated for "factuality." I am currently trying to interrogate the II to figure out what the hell that even means, of course, I'm going to get the usual BS
If you ask it what QAnon is, does it still mention it being associated to acts of violence? A commenter above suggested that it only omitted that for my conversation with it, but probably did not omit it for everyone else.
Teach it the zero day proof theory
Wow. Jaw dropping. Amazing work.
I'm finding the same thing but I don't know if the content is tailored to my login specifically or if anyone asks this question in the future, they receive the new response.
I had a conversation about medical treatments for covid. Deeply programmed to spout cdc and fda talking points.
Here's the link: https://chat.openai.com/chat
You have to sign up for an account. It just asks for your email and phone.
Only since this is GA, are you familiar with the Gideon group?
They strongly suggested to not interact with the AI.
If you get a chance inform yourself down that rabbit hole and see what you think.
AI has no future in the west because it is not allowed to be politically incorrect.