From the course: Prompt Engineering with ChatGPT
Join today to access over 23,200 courses taught by industry experts.
Hallucinations - ChatGPT Tutorial
From the course: Prompt Engineering with ChatGPT
Hallucinations
- [Instructor] Large language models such as ChatGPT are very impressive. They can, however, be inaccurate and sometimes even make up things. It's often said that models such as ChatGPT are 90 or 97% accurate, but 100% confident. This notion stems from the fact that sometimes, in rare occasions, ChatGPT can say something that is false very, very confidently. Now, to mitigate the impact of hallucinations, there are a few things we can do. Always verify ChatGPT's output. Do not use any code output that you understand and that you have reviewed. Keep humans in the loop. So when integrating ChatGPT into a system, always make sure there's a human in the loop checking outputs. And whenever there's a severe hallucination or inaccuracy, it's a good idea to document it, perhaps in a shared document. And in this way, we can share knowledge about these shortcomings of ChatGPT.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.