Pretty sure the author means "must not" or "should not", because they absolutely can.
No they can’t. The author’s point is these are computer programs written by humans. It doesn’t matter if they’re LLM backed or not.
If I write a program:
if (rand() % 2) launch_missles();
And wire that function to actually launch a missile, the computer didn’t make a decision. And so it is with people prompting agents.
Well, if I ask "option A or B?" and the computer rolls a boolean, it can still count as a decision but no context was taken in.
No they can’t. The author’s point is these are computer programs written by humans. It doesn’t matter if they’re LLM backed or not.
If I write a program:
if (rand() % 2) launch_missles();
And wire that function to actually launch a missile, the computer didn’t make a decision. And so it is with people prompting agents.