However, there are tasks that AI is pretty darn good at, such as translation. My ability to quickly and easily translate German, Russian, French, and Dutch Houdini posters and newspaper articles has unlocked new information about Houdini's performances in these countries. It's a game-changer that has gone a long way in helping me solve the Germany problem, and has greatly benefited my book.
However, things get trickier when it comes to using AI for research. While it's very good at very specific tasks – such as finding a street address for a long-gone circus building in 1901 – it's not good at deep Houdini research. Its knowledge of Houdini is basic and largely inaccurate. Also, as I recently discovered, it will lie. Not "hallucinate." That's a word people use, but I think it's too polite. AI will deliberately lie and even fabricate evidence for that lie. I know because it happened to me.
I spelled out my experience for my patrons below and have also provided the entire chat between Gemini and me. I think you will be as shocked as I was. There's also a good discussion in the comments.
It's a wild new world! Be careful out there.
