The machine does the work, a human steers it from a distance. Between a robot that decides and a robot that executes, the difference is who is accountable.
The digger operator in the cab moves the arm with two levers. Nobody says the digger digs by itself. Remove the cab, add a camera and hand the levers to someone in the next room, and the work is the same. Only the human's location has changed.
That difference matters when you read a robot story. An autonomous machine makes decisions by rules somebody wrote in advance, and a failure gets investigated in the code. With teleoperation the failure has an address: the person on the levers.
There is a third, middle position growing fast: the machine handles the routine and calls a human for the hard part. Then the question becomes how fast that human understands what is happening when the call comes.
Why it gets used
Because autonomy in unfamiliar surroundings is still expensive and brittle, and a burning forest does not wait for a software update. Teleoperation gets the human out of the dangerous place without requiring the machine to be clever. That is an honest trade, as long as it is called by its name.