ChatGPT confidently makes up false information. Ask specific questions instead of vague ones, keep each prompt focused on one concept, and include photos so the model has something real to reference.
One of the biggest problems with ChatGPT and other AIs is that they regularly hallucinate, which means they often create convincing yet completely wrong information. These are known as hallucinations, and the problem is the confidence that they deliver it. With that often it's hard to tell, but here's a way to get around that. Actually, how about three ways? First, one is asking direct questions. Don't beat around the bush. If you manage to be straight to the point and take out all the room for speculation or misinterpretation, well, then it won't happen. My second tip is similar. Avoid mixing different concepts. So if you start giving it different ways of thinking about something or different frameworks, often it can confuse these because there's logical fallacies that you might not even see, but the model will get confused. But the ultimate way to avoid hallucinations is this. Provided pictures. A picture speaks more than a thousand words, and if you provide it with a photo of your phone, there is no room for misinterpretation there most of the time. So if you manage to prompt and provide it with a visual of the situation, well, in most cases, that's completely going to eliminate hallucinations.
