Comment by bm3719 | Hacker News Reader

Comment by bm3719 | Hacker News Reader

There's two main schools of thought regarding the downsides of AI: One is apocalyptic, focusing on existential threats of the kind seen in scifi literature, and is largely influenced by the exact warnings given therein. The other, which I'll call "socio-realism", often looks into the actual past at the introduction of automation and its displacing effects.

The "AI safety" movement is largely an answer to only one side: the apocalyptic one. Doomers like Yudkowsky and his ilk are popular over there. Note how his most recent book reads like scifi, including with its implausible "solutions"? These people are living in a fantasy land. This isn't to say p(doom) = 0, but we've lived with it being non-zero since the introduction of nuclear weapons. Very little has changed there, I'd say.

For the socio-realism side, well, this model aligns well with exactly the kinds of real-world misery often posted on this site, as intellectual workers' livelihoods and telos are negated by automation. Sellers of "AI safety" don't care about them. There's no calls from that community to carve out space for us to retain a life of meaning within society when our ability to contribute is nullified. The fantasy of AI safety doesn't have room for that: they're too busy warding off imaginary demons; hence the cult-like behavior.

> There's two main schools of thought regarding the downsides of AI: One is apocalyptic, focusing on existential threats of the kind seen in scifi literature, and is largely influenced by the exact warnings given therein. The other, which I'll call "socio-realism", often looks into the actual past at the introduction of automation and its displacing effects.

Yes! Also the "apocalyptic, focusing on existential threats of the kind seen in scifi literature" serves as a useful strawman fo AI advocates to deflect attention away from the socio-realism critique.

It's happened a few times where I say I don't really care about in the apocalyptic sci-fi scenarios [1] but I'm more concern about the "socio-realism," but some AI proponent keeps trying to bend the discussion back to the unrealism of self-replicating von Neumann machines or something.

[1] The most interesting thing about the apocalyptic sci-fi scenarios is that they show the fundamental irresponsibility and recklessness of many very important AI leaders. Those people often seriously entertain those doomsday scenarios while simultaneously racing as fast as they possibly can to build the technologies that they say would enable them.

  > The most interesting thing about the apocalyptic sci-fi scenarios is that they show the fundamental irresponsibility and recklessness of many very important AI leaders.

To me at least, it looks like a very cheap and transparent way of bragging and elevating one's self-importance. If we take a practical approach, then AI is just another tool that makes some jobs easier and more automated, which sounds about as sexy as inventing the mass-production of interchangable screws. An astonishing improvement for people who used to mill everything by hand, and absolutely critical technology for the modern world, but barely anyone has even heard of the people who made it possible.

Another aspect is the world that people like the Altmans and Musks come from. The wealthier someone is, the more detached they tend to be from the manual labor that runs the world. When your goods simply appear at your door after a few taps on an app, it's easy to forget that someone made them with their hands (especially true for high-end products), someone packed them, someone drove them to a warehouse, someone photographed and edited them, someone wrote the listing, someone packed them again, and someone eventually brought them to your door.

The entire physical world gets compressed into an app. From that perspective, it can start to seem as if the app is the world itself, and that if someone can hack the app, they can destroy the world.

I'm pretty sure the CEOs of Tesla/SpaceX, OpenAI, and any company depending on data center buildout are painfully aware of physical supply chains and manual labor scarcity right now.

> [1] The most interesting thing about the apocalyptic sci-fi scenarios is that they show the fundamental irresponsibility and recklessness of many very important AI leaders. Those people often seriously entertain those doomsday scenarios while simultaneously racing as fast as they possibly can to build the technologies that they say would enable them.

It's the fascist playbook to champion the end of times while simultaneously proclaiming to being the chosen one to save humanity. It's not different than with these LLM companies.

Naomi Klein's and Astra Taylor's book go over just this: "End Times Fascism."

Use common sense! You yourself are either an LLM or one half of a self–replicating von Neumann machine. You cannot possibly claim that von Neumann machines are ”unreal“.

If a giant, Chicxulub-sized asteroid is headed toward Earth, it makes sense to focus on the existential threat, instead of worrying about whether it will cause inflation or have disparate impact.

You can reject the idea that the asteroid will hit the Earth, but you need to actually make that argument, not just call the people warning about the asteroid an apocalyptic sex cult.

Reply on news.ycombinator.com