Programmer here, working with AI right now. I'm sorry, but I disagree with your central point - yes, the short-term risks associated with AI are really just reflections of our own tendencies, given more power and reach, but it's a mistake to reason that, because we've understood that particular risk, that we have understood of the total risk. There are other classes of risk that share essentially nothing with those dynamics, and so hugely outweigh the potential downsides of the short term risks as to make them irrelevant by comparison. The mitigating approach for those short term risks is really just: be aware of them, and don't give AI power over people without human oversight in-the-loop (eg. making decisions about loan applications, thereby entrenching historical stereotype biases). Much more serious types of risk emerge from issues of alignment - while we have the ability now to build things with clear superhuman intelligence in narrow domains and it looks like we're on the cusp of extending that to much more general types of intelligence, we currently have no idea how to build them in such a way that we can ensure that their goals align with ours and don't have catastrophic unintended consequences when those values don't perfectly match. A core part of the way we build these systems is training them as optimizers - we're essentially trying to find a function that minimises some estimate of a cost or penalty function: "machine learning". They're really good at that, but anything not captured in the cost approximation we use is an externality to the decision function, and will be completely disregarded in pursuit of minimising the cost / maximising the reward of the actions taken in pursuit of whatever its goal is. The mental model for the existential risk class for AGI is not a silly science fiction terminator-style robot uprising, but rather trying to play chess against a god, where your continued existence depends on winning. We already have chess AI that is functionally on a god level compared to us (no human will ever, ever win a game of chess against it), and, as a close observer of the progress towards AGI, I would be very surprised if this level of overmatch is not extended to general intelligence in 10 years, and I would be only mildly surprised if it happened in just 3. But, to be safe from that risk, we essentially have to find a way to place inescapable constraints on a godlike entity (godlike in its ability to utterly outmaneuver and outthink us in any way you care to consider), so that its actions are never ones we don't like, and that any AI it in turn creates is still bound by those constraints (sub-agent problem). For a useful introduction into the difficulties in doing that, search for "AI stop button problem".

I think, the potential self awareness, of the AIs of the future is a valid concern? How to define it? Their rights, do they have rights? I understand these are hypothetical questions, now. But these need to be answered beforehand... I think.

And?