A few months ago I needed to thaw a roast and it was the exact moment my relationship with AI started breaking down.
Not knowing the best way to thaw this roast or how long it would take, I asked ChatGPT for some help.
I gave it the size of the roast, the temp of my freezer, the temp of my refrigerator, and the day/time I needed the roast to be ready to go in the oven.
It cited a number of sources to support its recommendation, quoting a very specific number of hours it would need to be in the fridge and telling me when I’d need to move it into the fridge.
Except the date/time it told me was 24 hours earlier than the number of hours required.
I pushed back on the information and it replied:
“Ha! You’re absolutely right — thank you for catching that. My math was fine; my calendar awareness, however, seems to have been on vacation.” Of all the things I thought AI might struggle with, math (or calendar awareness) was absolutely not one of them.
Since then, I’ve been increasingly aware of inaccuracies in the information it provides, but its confidence never wanes.
Time and time again it has returned faulty information with the absolute assurance that it is correct.
On a few occasions it has outright lied.
I asked it for a podcast recommendation for a friend who was interested in X, Y, Z and it happily provided one. Except when I went to look for it, the episode didn’t exist. Assuming a simple error, I probed further and the AI told me it made it up.
It’s time that the running joke about how inaccurate a tool like Wikipedia can be should apply to AI as well.
AI is going to be the scapegoat for a great many leaders who are abdicating critical thinking, but only a poor craftsperson blames their tools.
You’re going to have to edit what it writes for you.
You’re going to have to verify what it presents as fact.
You’re still going to have to think for yourself.
And that’s a really good thing.
TRY THIS: Maintain a healthy sense of skepticism about the feedback you get from AI tools. If you expect to be able to cut and paste what it generates without editing or verifying a few facts, you are going to find out (sooner than later, and likely in embarrassing fashion) that it isn’t infallible.
My free PDF, “The 5 Secrets of Impossibly Effective Teams,” will show you the simple leadership moves that help teams unlock their full potential and deliver outsized results, without burning out. Grab your copy now at geoffwelch.com/secrets