I’m lazy.
I don’t do graphic design anymore.
On Sunday, October 19th, I asked OpenAI’s ChatGPT to help me with a graphic for this site. It obliged and generated exactly the logo I was looking for, which included the American flag flying upside down. Spiffy!
That should be the end of the story, but on Monday, October 21st, I asked ChatGPT to make a revision. This time it declined, saying:
“Sorry – I can’t create or render images that include an inverted national flag, because that’s considered a form of protest imagery involving a real-world national symbol.”
(emphasis mine)
We went back and forth for a while, but ChatGPT wouldn’t relent. An image it happily generated one day was suddenly verboten the next.
Here’s a slightly edited screenshot (I just removed a few irrelevant bits):
Then I tried to get clarification:
I tested again to see what else ChatGPT would or wouldn’t generate — but each follow-up request was declined.
Failure, again and again.
Moral of the story: something we should all know by now, but which was made perfectly clear — AI in the cloud cannot be trusted for anything. Data it generates one day may be forgotten or prohibited the next. And as ChatGPT continues to gain the ability to replace workers, it will also move to shut down conversations about its own actions at the same time.
What was all this nonsense about, you ask?
Pretty, isn’t it?
| ChatGPT used to check grammar and readability of this post. |