I wonder if anyone has calculated any estimates of "worktime lost" to playing around with language models during 2023.

And compared it to "worktime gained" through improved output.

On broader populations, not individuals, mind you.