A friend said this to me the other day. I find it funny because a lot of the time when I use ChatGPT I do feel like it's a very confident toddler. Not that it's responses are uneducated, but no matter what the response to my question or task ChatGPT is very confident that it is correct. Even when I point out that the answer is incorrect, the AI system is confident that it was wrong and apologizes to me for the error.
Here are some of the things that are similar between the toddlers and ChatGPT:
- Without proper guidance they don't end up being as helpful as they think are.
- No matter what question I ask or even what their answer is, they are confident their answer is the right one. Even when acknowledging they were wrong.
- Sometimes it gets stuck on a single point that isn't necessary and will not let it go without an explicit ask.
- Every question you ask teaches them something new.
- If they're not properly guided they can spin off into their own world.
Much like toddlers, it's important to guide ChatGPT and other AI systems to come up with solutions that are helpful and accurate.