A quick chatGPT exchange I had:

Me: could you visualize this CSV as a line chart?

chatGPT: Sure! Here you go 
<a pretty good visualization>

Me: that's pretty good, but you skipped one of the columns.

chatGPT: Your right, apologies! Let me remake that visualization with the missing column added to it.

Computer Scientist who just thinks onchain is neat. /formative

I can't get over how weird it is to have a computer: 

-make a mistake (i.e. not a bug resulting from human programming error, but the computer simply forgetting something!)
-being able to correct its mistake simply by being asked to do so!

100% I think you have a much better explanation of what occurred. It's just from a layman's perspective, it looked like such a "human" error, simply skipping or overlooking something.

I'm having trouble articulating to my people IRL how freaky all this stuff is. People in this space seem to understand that though.

'forgetting' is a bit of an oversimplification. What essentially happened is in the probability space, there were too many variables with similar weighting, and it was not able to equally consider them all, so some variables were discounted, resulting in an imperfect logic. Then, upon being told that there is indeed an error to find, it will agree as a matter of course, apologize because probabilistically that's what would come next, then proceed to reevaluate the problem with less floating variables this time and reach a better outcome. 

It's still freaky though...