THE AI HARD PROBLEMS AREN’T HARD PROBLEMS 1) Consciousness is possible and it’s

THE AI HARD PROBLEMS AREN’T HARD PROBLEMS
1) Consciousness is possible and it’s not even hard. It’s a natural consequence of the expansion of recursion of auto association of episodic memory into increasing abstraction from the body form (embodiment).
The problem is the hardware, and the correct hardware is under development. We must pre-calculate networks because we’re using virtual connections (memory addresses) and a small number of processors, instead of a large number of very tiny processors with limited local memory, and physical connections that can connect into networks that do NOT need to pre-calculate (train) but that can constantly adapt in real-time.

2) An AI needs a means of decidability. Humans decide self-interest. AI’s don’t have to use self-interest. They could use reciprocity instead.

3) An AI doesn’t need a mission. The human mission is the continuous acquisition of ‘more’ of everything vs physical cost of it – even if ‘more’ is just stimulation. An AI doesn’t need a baseline mission.

4) An AI can be ethical and moral. Ethical and moral behavior is programmatic in that everything is ‘owned’ so to speak, to some degree by one or more people. So an AI only needs to ‘see’ (have access to) what is permitted.

5) An AI can’t be incredibly creative, for the simple reason that human creativity is limited by the cost of experiments, not by the limits of reasoning. So without the ability, direction to, and means of deciding or moralizing, it’s pretty hard to imagine a machine is terribly innovative.

6) What AI’s can do is organize production in the service of demand faster than we can. And that’s going to be ‘interesting’ and disruptive.

SUMMARY
The danger of AI’s comes from people weaponizing them and by weaponizing them exceeds the rate at which humans can identify and detect threats. So, AIs will protect against malevolent AIs and we will regulate AIs the way we regulate other dangerous machines, mechanisms, and ‘chemistry’ (explosives).


Source date (UTC): 2023-02-24 00:26:00 UTC

Original post: https://twitter.com/i/web/status/1628914104109760512

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *