This struggle will not be easy. Check The Matrix, if you doubt it. And a mad Artificial Intelligence is possible. Check 2001 A Space Odyssey by Stanley Kubrick. And many others. It is high time all the Bernard Marr's and Pascal Bornet's get down from their self-proclaimed pedestal and accept a fair scientific discussion on the digitally copied features borrowed from standard (that sounds selective) humans. Does the program of the machine include some crypto-cum-proto-cum-genetic bias towards schizophrenia, or paranoid utopia, not dystopia, because these Artificial Intelligences have good manners? They always smile, even if and when they press the famous red button. With the permission of those involved, I would like to share with you an email exchange between researchers working with Artificial Intelligence, but also working with caution regarding the pitfalls of this generative or non-generative AI, while constantly correcting its errors, particularly when it involves translation. Imagine a translating-AI trying to translate the language of the 15th-century Incas, whose language we have no record of, since they had no writing system. Translate into English, I guess, but passing through Peruvian Spanish, I imagine, or that of Hernán Cortés.
COULARDEAU et al. (Sun,) studied this question.