New wave of AI fears
The current wave of AI-doom talk may be less about alignment than regulation, and the real danger is humans holding an ultra-powerful tool.
I wanted to pitch in a bit regarding the recent massive wave of discussions that has resurfaced again about the dangers of AI, specifically that it will rebel on its own and wipe us all out. First I actually think that this is pretty much an engineered campaign, that is not about safe or alignment, but about mass regulation that will benefit market leaders and bureaucrats.
Besides that I want to capture a couple of my own thoughts on this. If only so that down the road, I can look back and laugh at just how wrong I was about everything.
-
AI definitely poses a danger. But it seems to me that almost everyone across the board misses one simple dynamic. The danger stems from the fact that AI is an ultra-powerful tool, that it will likely become even more powerful soon, and that humanity might not be able to stand against it. But the real question is who’s actually calling the shots at that point? The creators of the AI (or their bosses), or a self-aware AI itself? I believe that by the time AI reaches that genuinely dangerous level, it won't yet have self-awareness and thus any desire to view humanity as an enemy. That dangerous threshold will arrive sooner, meaning there will still be a small circle of plain old humans pulling the strings. The unprecedented power handed to them is where the primary threat lies. The danger comes from them—the proverbial monkey with a grenade—not from the grenade itself.
-
Acknowledging a threat, even if correctly identified, doesn’t mean the proposed solutions will improve the situation, let alone resolve it. A visual metaphor comes to mind: imagine an isolated island tribe learning that a modern amphibious assault ship is heading straight for their shores. From rumors, they know it's dangerous. So they hold a council to figure out what to do. Some think they should just pray. Others insist on sacrificing seven virgins is required, others blame a witch who must be burned. Some propose building the right kind of idol or a statue and someone suggests laying a circle of yellow seashells around the village for protection. Obviously, you can guess how well that's going to work. A modern person might be able to advise them on how to avoid dying in a futile battle and resolve things peacefully, even if it meant upending their way of life, with some things turning out worse, but many potentially better. But either way, their current proposed solutions are utterly ridiculous.
I don't know how things will go, but somehow I still have faith in humanity and our ability to figure things out, including the alignment. Let's sort ourselves, that's the key, and AGI will learn from us.
Tags: aiphilosophysociety
