News for hackers. Stuff that matters.
Re:A.I. Companies Agree to Safeguards After Pressure from th... by happytiger on Friday July 21, 2023 @01:22PM (#36816661)
I can’t imagine how sophisticated networks of AIs deployed globally could become, nor how quickly they could proliferate.

Lawmakers keep thinking of AI as a capability, which can be moderated, like traditional systems. But it’s not monolithic AI systems that give me pause. It’s their use as coupled networks that gives me pause. That’s a whole different paradigm of entities that are smarter than any person and outside of any governments ability to regulate working together, where transformation and interruption are inevitable, and it doesn’t require a great deal of capability to be developed to get there.

How do you regulate thousands of AIs that are cooperating both offensively and defensively, 10,000 times smarter than people, and spread across every country in the world (including non-cooperative ones) entirely focused on strip mining financial or energy markets for example? What if it’s focus is simple and legal soft influence campaigns to gradually change public opinions? What if they are self proliferating like worms or viruses (which is absolutely inevitable).

You can’t regulate that. That simply a totally new era (or paradigm) with new rules.

There is no “doing it safely” when it comes to AI.

I’m still excited about the technology, but it’s interesting to see how the old system is trying to put “guardrails on an industry” as if AGI isn’t right around the corner and all we need is another regulatory agency to keep things in check.

There are no comments on this comment yet.