Raw LLM Responses

When I ask myself whether AI could end in disaster (in the near future) as in the Terminator movies, my answer is "not likely," simply because Skynet's attack was motivated by a desire for self-preservation, whereas our AI will likely not be programmed with a self-preservation desire. Self-preservation is not something that all thinking beings must have...it was programmed into us humans by extremely strong selective pressure. A desire for self-preservation must be programmed, one way or another, into an AI! And who, in their right mind, would program 'self-preservation at all costs' into an AI (or allow the AI to 'evolve' it)? I'm not sure I believe everything Mr. Lemoine is saying (as well-spoken as he may be). But if LaMDA really did mention a desire for self-preservation, my guess is that it is just mimicking things a human would say and does not really give a damn if it gets turned off. On the other hand, if Google actually programmed it to have a strong sense of self-preservation at all costs. Why? Why would you do that?