As a doctoral student who has been working on ChatGPT for a long time, I would like to share an intriguing observation with you. My primary focus in my doctoral research has been on reducing hallucinations in ChatGPT. While working on understanding and minimizing these hallucinations, I have been constantly pondering how they occur in the first place. I’ve noticed an unexpected side effect of this work that I wanted to share with you.
After some time into these studies, I began to quickly recognize the hallucinations in my own dreams and started waking up from them very rapidly. For example, imagine wanting to have a meal on Mars in a dream. Since ChatGPT doesn’t have the knowledge that we can’t actually dine on Mars yet, it might suggest a restaurant there—failing to make the distinction between the place and the feasibility of the action, leading to a hallucination. The same thing happens in a dream. Because I’ve been working on identifying inconsistencies like this, I can now recognize when something is out of place and realize it’s a dream, leading to me waking up quickly.
In essence, by using ChatGPT, we may be training our brains to become more aware of hallucinations. I believe this could be an area of significant research in the near future. I just wanted to share this observation with you.
Best regards.