Comment by Kim_Bruning

Right, normally "it's merely outputting what it's trained to be the most likely text to follow what came before."

Now -while it's doing that- if you poke at certain vectors, its predictions will veer off course in interesting ways.

This shows that the activation vectors exist, and that their modification provides a causal contribution to the output.

Sure there's interesting consequences of that. But if we say that's the take-home message, that's good enough for me for one day!