How to get ChatGPT to be more reliable…#greenscreen #artificialintelligence #chatgpt #openai #generativeai

Advanced Prompt Engineering for ChatGPT Part 1. In this series I'm going to go through specific prompts that you can use with ChatGPT to get better output. This first prompt has to do with how do you get ChatGPT to give more reliable answers. Sometimes ChatGPT can come up with wrong answers or factually incorrect information. So here are some ways to sort of fact-check ChatGPT. First off and most simply, I think at the end of your prompts you can simply add answer only using reliable sources and cite those sources. When you do this it will actually provide you the URLs of where it got the information it used from. Another great way to prompt ChatGPT is to force it to give you its reasoning. So here's the prompt you can use for more elaborate questions that require ChatGPT to think in steps. So let's say I'm asking it what is the meaning of life according to Buddhism. After I give it the question I'm saying use the format of Q, repeat question, A, let's think step-by-step, give reasoning, therefore the answer is final answer. So I want ChatGPT to answer in a specific format by kind of filling in these blanks here which will give me a hundred percent clarity in its reasoning and its final answer. So therefore when ChatGPT comes back you can see it's thinking in step-by-step and how it actually gets to its final answer. And finally the last thing you can do on any ChatGPT interaction is to actually ask it is there anything factually incorrect in your answer. And funny enough in the previous prompt it actually did think that it included some inaccuracies. I would say the question I asked was a little abstract but it is interesting to see what it noted as incorrect and what it included previously. So these are three methods to really fact-check the outputs that you're getting from ChatGPT. Part two of advanced prompt engineering is coming soon.