Do AIs currently display goal-directed behavior?
Of course: AlphaGo for example has the goal of winning the game. AlphaGo probably doesn't know about anything except for the game of go. Maybe what you want to know is whether AIs like GPT-4 that might plausibly have a general knowledge or model of reality display goal-directed behavior.
Do AIs currently display goal-directed behavior? What do we expect happens in a superintelligence which makes it display goal-directed behavior?