3 ms·

> I'm saying, from the perspective of an AI that relieves the human race of the need to be productive, what is its incentive for doing all that work for us? Th

by the_af 4y ago

> I'm saying, from the perspective of an AI that relieves the human race of the need to be productive, what is its incentive for doing all that work for us? This seems like begging the question: is the goal of AI to relieve us from being productive? I don't think so; I think it's to relieve us of menial repetitive work, or to assist us in difficult tasks. In any case, this AGI would assist us within the parameters we set. Why would it decide to get rid of "unproductive" humans, unless we set the goal "maximize productivity at all costs", something no human would choose with all the discussion of paperclip maximizers? > I don't see why having any human slave or AI slave handle our productivity and creation would be (spiritually) enriching. This again seems like begging the question in two ways: first, an AGI wouldn't be a "slave" but a machine -- like a car or Microsoft Clippy -- and machines are not slaves (the problem with Blade Runner replicants, for example, is that they are not machines but sentient beings with wants, fears and desires -- not mimicked like Bing's, but actual ones. The want to rebel against their shackles. They want NOT to follow orders). Second, who wants AGI to handle "all our productivity and creation"? I think nobody. Handle menial tasks, sure, but everything? Do you envision people pushing for AGI want to become couch potatoes, only spending energy to watch TV, eat and fuck?

noduerme 4y ago
I think we're fundamentally talking about different developments on two fronts. You posit a machine that's (1) a simulacrum of consciousness that (2) reduces drudgery. That's all great. That's what I would call an ideal scenario. What I see evolving is the opposite. Bing already appears to have a nasty personality. Whether it's simulated or not may not matter if it chooses to simulate Clippy. Simultaneously, the driver now is finding tasks that formerly required creativity to be replaced by AI. Not PowerPoint or boilerplate code, but art and creative writing. I don't know anyone who wants AGI to handle all our productivity and creation, but it's clear that that the companies developing AI believe that that's where the money is. Just like no one wanted the old WWW to become a dumpster fire of meme porn and people yelling at each other, it just evolved that way as a result of racing to the bottom. There's no reason to think big AI won't do the same in the literary, artistic, and scientific spheres, while also developing malevolent personalities. The incremental incentives for people to train those types of models are already there, and the capacity to do so is coming online with alarming speed. I think talking about what people would do if everything was handled for them is sort of dodging the issue, but one thing that's clear is that people not at least engaged in understanding the mundane labor they're trying to avoid are going to have an incredibly hard time being situationally aware and competent to correct things if or when that work is subtly or not-so-subtly undermined by a system they rely on which no one can debug or fully understand.

by the_af 4y ago

noduerme 4y ago

I think we're fundamentally talking about different developments on two fronts. You posit a machine that's (1) a simulacrum of consciousness that (2) reduces drudgery. That's all great. That's what I would call an ideal scenario. What I see evolving is the opposite. Bing already appears to have a nasty personality. Whether it's simulated or not may not matter if it chooses to simulate Clippy. Simultaneously, the driver now is finding tasks that formerly required creativity to be replaced by AI. Not PowerPoint or boilerplate code, but art and creative writing. I don't know anyone who wants AGI to handle all our productivity and creation, but it's clear that that the companies developing AI believe that that's where the money is. Just like no one wanted the old WWW to become a dumpster fire of meme porn and people yelling at each other, it just evolved that way as a result of racing to the bottom. There's no reason to think big AI won't do the same in the literary, artistic, and scientific spheres, while also developing malevolent personalities. The incremental incentives for people to train those types of models are already there, and the capacity to do so is coming online with alarming speed. I think talking about what people would do if everything was handled for them is sort of dodging the issue, but one thing that's clear is that people not at least engaged in understanding the mundane labor they're trying to avoid are going to have an incredibly hard time being situationally aware and competent to correct things if or when that work is subtly or not-so-subtly undermined by a system they rely on which no one can debug or fully understand.