First of all, Musk sounds like an alarmist... only if you read him as forecasting a Terminator (movie) event, literally. Of course, he isn't. Zuckerberg however, probably interprets Musk's warnings too literally, and is therefore laughing at the wrong image.
Musk's main concern is that robots have already reached a point where handlers cannot accurately predict their actions. Clearly this is dangerous if bumbling robots accidentally come across weapons and act like bumbling robots. Also, every passing day raises the level of uncertainty about robot actions. So far, nothing I have described is short of hard fact.
The question is not whether we should regulate robots. The question is how, and after that, the question is when? We know the answer to neither question.
Musk is simply on track to say we ought to invest time in trying to figure out these two answers.
(FWIW: I too used to think Musk was joking. Then it seemed we should take his gaffs a little more seriously.)
Convergence
In the long run, synthetic minds are as unpredictable as human minds. Since it is the unpredictability of humans which causes the entire spectrum of political-economic behaviours to emerge... in the long run, synthetic minds will be regulated in a very similar fashion to natural born humans.
If an AI does something undesirable, it will generally not be possible to track down why it did it - negotiation with the AI may be attempted, but if the AI proves to be sociopathic, then it will face capital punishment - possibly death, or solitary confinement. At the end of the day, machines are people too. People in general are just complex information systems.
(In the context of a conversation on how the use of machines changes the way organic people interact with each other:)
I mean, you can ask the question "at what point does it become the case that a synthetic person has been raped by a natural born human person?" And then the same penalties apply.
/
Micro-economically, the individual events of organic-synthetic interaction can be addressed at one level of regulation. This is pathology and medicine at the level of individuals.
The formation of such regulation can take into account the macro-economic effect... that's possibly addressed by another level of regulation. This is public health, and civil society design etc.
But the two levels/orders/resolutions of concern, and the enactment of policies at each level, will inform activities at the other order. We write public health law with concern for the well-being of individuals, and we address individuals with therapies with consideration for the gross effects on the public that their individual psyches will engender.
This is already the case with organic-organic person relations.
My point is simply that you should not address organic-synthetic relations any differently, because in the long-run they will converge to where organic-organic person relations are currently.
/
I think in the long run the synthetics are going to be more complex, and therefore arguably more sentient than individual unmodified organics. LOL.
/
