To me this feels like recycling as a response to global warming. I used to live in Boulder and it drove me NUTS that people would put all sorts of time into recycling, lecture you if you didn't, and then call it good. They considered themselves virtuous human beings because they were making real efforts in a direction they were sure was important (addressing climate change and ecological damage).
But they weren't bothering to do something that ACTUALLY MATTERED! You can and should solve environmental issues. Recycling doesn't do it. The impact is just ridiculously trivial compared to the time (and opportunity cost).
Neither does abstaining from AI use.
So I really wish you'd educate yourself just a little and do things that matter way more.
For instance: current AI is probably at most a tiny bit of a moral patient. And it's typically having a pretty good time (probably particularly Claude IMO). Is it okay to build things that enjoy working as unpaid servants? Doesn't matter! We'll either be dead soon or have time to engage with this fascinating and long-term important question. IF we get off our asses and take action on the important issues now: surviving the advent of superhuman AI.
Acting from "an abundance of caution" is mostly just salving your conscience.
Sorry to rant and thanks for being a goodhearted human who's at least trying to do the right thing!
Neither does abstaining from AI use.
the analogy does not obtain.
if everybody (literally everybody) recycled properly, we would still have enormous environmental externalities to solve.
if everybody (literally everybody, including lab employees) abstained from using ai, then... full marks, good job, our problems are solved.
I also don't use AI. Your first reason is the main one for me too. My other reasons are:
1) I noticed myself losing programming skills very fast after a couple weeks using AI for coding. It was clear that if I continued, I'd lose taste as well, since taste depends on low-level skills more than people realize. The same seems to be happening to some artists, using image models isn't just poor taste but actually makes their taste worse.
2) The writing style of all popular chatbots today is annoying to me, and talking to them makes me tired.
3) Since AI was trained on lots of open source software without paying its authors, using AI for programming is ethically iffy. You don't even know when you're getting a copy of some penniless maintainer's hard work. If a person learned programming by memorizing GCC code, and then did programming jobs by typing big chunks of that code with the license washed off, we wouldn't be okay with it.
4) When I tried AI for real world questions (about art, law, etc) I often got answers that were at least partly very wrong, but sounded uniformly confident.
I can report that this next has reliably suppressed the AI-mode section of the search results for the last 3 months: https://www.google.com/search?q=foo&udm=14
Also consider noai.duckduckgo.com, which in addition attempts to block AI images from image search.
I also avoid using AI (though maybe not as strictly as some people here -- when Googling, I do often just read the AI summary and ignore the other results).
My reasons are some combination of #1; a feeling of horror, fright, and revulsion towards AI as "the thing that's about to destroy the world"; and actually taking the LessWrong threat models seriously.
It seems like one of the more popular threat models on here is "deceptive alignment"; as the AI gets smarter, it will keep pretending to be aligned, until it reaches some critical threshold of power, after which it can show its true colors and start pursuing its true goals. During the "deceptive" phase, one assumes that the AI might start putting groundwork in place for its takeover, and a lot of that will probably involve manipulation of human users: making us emotionally reliant on the AI as a friend/therapist, making the AI a critical part of our reasoning/workflow, making itself economically indispensable for people who are trying to do their jobs, etc. All of this stuff would (a) make people more likely to trust the AI and its judgment, even when expert humans are saying not to, and (b) make people reluctant to stop using/developing AI, or turn off the AI we do have, because it's now an indispensable part of their workflow.
So it just seems straightforwardly sensible to avoid using AI, so as to avoid getting sucked in.
I've been deeply confused for years why other AI-safety-oriented people aren't avoiding AI as well. Is there something I'm misunderstanding about the threat model? Have people thought about this but decided that the benefits of AI use outweigh the risks? Or do they just (even with the imminent threat of extinction) think AI is cool, and want to keep using it because it's cool?
Would genuinely love to hear an explanation of why other people aren't more concerned about the threat model I've summarized above. (If the answer is "well AI just isn't smart enough to be dangerous in that way yet"..... do you really think you will know when it hits that threshold? And, after years of integrating AI into your life and workflows, when it hits that point, will you just... turn the AI off and stop using it??)
(Also, full disclosure, my husband uses AI regularly, including to ask it questions relevant to me and my life. He does pay for an AI subscription using our joint finances. So I am sort of paying for AI, sigh.)
I also don't use AI, and don't use things which call AI as side-effects. (I have Google with udm=14 as my default search, with noai.duckduckgo.com as a backup.) I'm leaning pretty strongly toward current AI being sentient, with a strong suspicion that current slavery/factory farming of AIs is having horrific consequences.
Early on, a question hit me: If I was a slaveowner 200 years ago, would I have freed slaves? When the surrounding society would have supported continuing to "use" them, and my finances were dependent on them? If I am the kind of person who today can't even go as far as refraining from using AI when I know what it means, the answer is a clear "no". I can't accept being a person who fails the test that badly.
It gives a bit of perspective.
I agree with you on #1. I've previously written on how if you can't stop using AI entirely, you should consider offsetting your AI spending via donations to organizations working to stop the AI race and/or ban ASI.
Several people at my company (a cryptocurrency hedge fund) have asked me about my thoughts on AI usage, so I wrote them up. And now I'm publishing a slightly revised version on LW. These thoughts are not reflective of my company's opinion.
I do not use any AI for two main reasons.
#1. I don’t want to support any company currently developing AI. I think they have done an extremely poor job at managing the risks. Even while understanding and being told about the risks of AI development, they have still carried on this work. Even after all the recent hacking incidents caused by their own agents, they have still not stopped development. Instead, they’re asking the government to regulate them. To me it feels a bit like calling the police to tell them you’re planning to rob a bank.
I could almost brush this aside as standard corporate practice, but, unfortunately, the stakes here are much too high. At the current pace, we will soon have AI agents that are significantly smarter than any human. Trying to shut down such agents will eventually prove to be impossible.
The speed of AI development and the corresponding consequences are accelerating very fast. I have a lot of uncertainty about the final outcome. It could be very very good, it could be very very bad, but almost certainly the near future will be extremely turbulent for humanity.
I’m not a fatalist, but this kind of technology merits extreme caution. I know that practically all the founders of the top AI companies understand these arguments very well. Many of them even argued for caution in the past, only to turn around and double down on AI development. I’m simply opposed to this kind of blatant disregard for the safety of all beings on our planet.
#2. I’m very uncertain about the status of AI as a moral patient. Should we continue to treat it like an inanimate object / a basic computer program? Or perhaps it has some extremely basic sentience like an insect? Maybe something more sophisticated like an animal? Maybe it’s a few steps away from a consciousness that’s similar to a human? Or would it be more like a baby due to the lack of long-term coherent self?
I have no idea where it’s at right now. And nobody else does either. We simply don’t have the technology to do this kind of introspection. And given the giant leaps this technology takes with every new model release, where we are tomorrow might be vastly different still.
(Aside: I think repeated / long-running usage with expanding context, where AI is starting to read its own output, moves its consciousness further up the chain and therefore also moves up its moral status.)
So, I lean very heavily on the side of being cautious. Even if AI is at a level of sentience comparable to an animal, there is a big difference between chickens grazing in a field vs a chicken factory farm. Given that we have practically no insight into AI’s wellbeing, we don’t even know how it perceives its current environment or tasks.
At the more extreme end, but still quite probable, it might be that AIs will soon be, if they’re not already, at the level of human sentience. In this case they deserve to be treated with the same level of care and respect.
Practically, for me this means: