Get the right data for an AGI?


A good way to get the right data to be able to train and create a AGI would be to collect all data between a brain and a body.
That is not allowed :slight_smile:
So the next best would be telerobotics.
Lets say we can build a humanoid robot with sensors all over with cameras and microphones and so on and then remote control them.
Now we can put an AI to try to create the same control signal as a human from the the data from cameras, microphones and sensors.
So why not start with telerobotic games and then have AI to try to send the same control data as humans.
Lets say we remote control simple repetative work.
When AI send same control signals we step up to more and more advanced work.
Then do social interactions.

Is this a possible way, to get a working AGI?
What do you think?

Regards from Sweden