I’m not (yet) a ChatGPT user but your article left me flabbergasted.
Sorry to ask you this but my knowledge about this subject is very little. What are custom GPT’s ? Are they models that you train to do something? If they are and as in your article they actually do what they were not supposed to (actually, in a way this is not correct because they do actually what you ask but are not aware that is not good). Does it mean that they don’t understand what’s wrong and right? They just execute what one says? An average intelligent human being would understand and not execute it. How can we speak here about ‘intelligence’?
Maybe I don’t make any sense at all. If it really is so, sorry about that.
So custom GPTs are actually chatbots that you can create in the ChatGPT platform. They are customizable as to what they should do, how they should respond and interact with anyone who chat with it. Imagine you need help in learning something. You create a GPT, train it with certain knowledge base (be it a book or your own thoughts and words). So when each time you have a question about that subject, you just ask it from the GPT chatbot and it will search through the knowledge base you provided and will then answer your question based on that.
The example I provided is about a GPT's rules that it was trained on. For some people these rules or knowledge base will be a secret and they wouldn't want the public to see given its importace. But GPTs are hackable and will expose those rules/knowledge base which shows a dark side to the ChatGPT platform overall.
GPTs are great things for productive work and makes out lives easier as long as you don't provide anything that you wouldn't want the public to know about.
Thanks for the explanation, Sami. I’ve just got wiser on AI matters😊. It remains an amazing tool and as such it has good and bad things. Maybe time for a try out!
I’m not (yet) a ChatGPT user but your article left me flabbergasted.
Sorry to ask you this but my knowledge about this subject is very little. What are custom GPT’s ? Are they models that you train to do something? If they are and as in your article they actually do what they were not supposed to (actually, in a way this is not correct because they do actually what you ask but are not aware that is not good). Does it mean that they don’t understand what’s wrong and right? They just execute what one says? An average intelligent human being would understand and not execute it. How can we speak here about ‘intelligence’?
Maybe I don’t make any sense at all. If it really is so, sorry about that.
Hey Luz. That's totally ok. I'm glad you asked.
So custom GPTs are actually chatbots that you can create in the ChatGPT platform. They are customizable as to what they should do, how they should respond and interact with anyone who chat with it. Imagine you need help in learning something. You create a GPT, train it with certain knowledge base (be it a book or your own thoughts and words). So when each time you have a question about that subject, you just ask it from the GPT chatbot and it will search through the knowledge base you provided and will then answer your question based on that.
The example I provided is about a GPT's rules that it was trained on. For some people these rules or knowledge base will be a secret and they wouldn't want the public to see given its importace. But GPTs are hackable and will expose those rules/knowledge base which shows a dark side to the ChatGPT platform overall.
GPTs are great things for productive work and makes out lives easier as long as you don't provide anything that you wouldn't want the public to know about.
Hope I've answered your question.
Thanks for the explanation, Sami. I’ve just got wiser on AI matters😊. It remains an amazing tool and as such it has good and bad things. Maybe time for a try out!
It is an amazing tool no doubt. But as everything has a bad side, we should be more careful with what we input to training it.