Remember that ChatGPT's output is the starting point from which to refine, test, and iterate. It's not the finish line.

Check your Work

You should never take the output as it is and blindly publish, post, share, or otherwise use it with customers (or anyone, for that matter).

Aside from formatting and layout, within the content itself look for factual errors, make sure it fits your tone, formality, etc., and please check for any inherent bias that might bubble up from the sewer that is the vast majority of human speech it was trained on.

ChatGPT has a very robust protection layer to keep it from being a way to quickly generate hate speech at scale. But even with the best protection layer, sometimes things slip through.

ChatGPT is technology built by humans and trained on human language. There's gonna be some bias in there somewhere.

An Example of Inherent Bias

I asked ChatGPT to add some relevant emojis into some text I wrote - without changing the source text - so I could use it as a caption on Instagram and fit in with the cool kids.

Since it wasn't changing the source text, I just skimmed it to see where it put the emojis and where I said, "You should consider ChatGPT your assistant." it put an emoji of a lady.

I didn't ask it to put an emoji next to "assistant." It decided that assistant needed an emoji and the female emoji was the most appropriate one. I'm glad I caught that.

Just be Aware that Bias Exists

I'm guessing since gender bias is likely one of the most pervasive of all the biases, present in human speech going back to the beginning, that eradicating this entirely from AI output will be a challenge.

So just be aware of it.

Was this a big deal? To me it was. I want the content I produce and share to represent me and that one emoji where it was sent a message that's not in-line with my world view.

So,... check the output, even on the simplest little tasks.

I'm glad I caught it.