Robots need training. The fully autonomous training may take a lot of time. However, right now, the following methods are being discussed to enable robots to work with human intervention.

Copycat robots seem to be easy: put a camera on the head and guide the robot, but it can be error-prone. It’s not well tested.
Remote robots can be accurate, but they may be costly due to the machinery, screen, and settings. We need a lot of machinery to control the legs, arms, and fingers of robots.
References of each are given at the end.
Which one would you choose, copycat, as it’s cheaper and simpler? We don’t need to sit in a PlayStation-like scenario for virtually controlled robots. But normal work, copycat robots can do; however, it can be harmful if kids are near. Copycat robots can cause harm, due to minor mistakes in image recognition and voice errors.
So, you choose which one? Copycat training ? Or virtually controlled robots!
References
[1] https://www.instagram.com/reel/DWyGKNkgd7E/?igsh=bjJiZ2ZsMDg3em11