and maybe i sound like a boomer like “these kids are using the internets for everything but what happens when they need to use the dewey decimal system! and they wont know how!” but like i’m sure nbowie could better articulate the reason why relying on a corporate-controlled llm to live your life for you is deeply negative
The thing that I keep thinking about is
Panera took the death lemonade (like we literally know it as death lemonade) because it was tangentially related to two deaths without it being super clear that it was the lemonade, but that was enough to take it off the market. Two people. Two. LLMs have had pretty significant involvement in the deaths of way more people, many of them children. I know I’m brave for saying this, but I think that no productivity gain is worth sacrificing children. Then again, maybe I’m just a bitch who hopes she’d walk away from Omelas.
as a sidenote as a college student in an… uncompetitive program, the amount of time I’ve had to argue against “but chatgpt said…” for something in a group project where chatgpt just straight up does not understand what the teacher is asking for is… truly, truly infuriating. these people just will not relent. The only argument I’ve found that works is genuinely “I pay for pro chatgpt and it says something different”. the Median Group Project Member was already kinda useless but now with chatgpt they have something they think can quickly and effectively fulfill the assignment criteria. So now instead of just doing the group project on my own and putting their name on it (which is fine. its whatever. I can live with that), I have to actively pre-empt their chatgpting and do the project before they can try and contribute with chatgpt because it will output pure fucking SLOP
I’ve had SO many awkward conversations where I have to say straight up “hey, you wrote this with chatgpt and it’s making things up. I’m going to rewrite it”. because, well, I need an A on the assignment, and chatgpt is a walking B- machine.
Like yeah fuck it a B- with 0 effort is worth it as a tradeoff to basically everyone I know but not me. Because I want to get into lawschool. So now I have to do 4 persons worth of “A+ with some effort” work before the B- hallicunated slop inevitably gets put on the shared document because otherwise I have to have that awkard conversation. And no shade to people aiming for a B-, that’s kinda a problem with how college group projects are structured in the first place (people straight up are aiming for different grades), but we used to have students who understood to just let the tryhard group project member do all the work and now they think chatgpt will outdo the tryhard. fuck man.
yeah this is a whole other thing with not just chatgpt but the internet in general. social media has increased the youth suicide rate by such a fucking unfathomable degree that you can make the same argument about banning social media
(and fwiw I actually would make the argument to ban all infinitescroll/algorithm and LLMs for use for people under 16. 13 at least.)
Tangent, but — and I absolutely do need to preface this with a “I could be wrong on the details” — I think the “social media is to blame for the increase in youth suicide” notion is a moral panic. Like it’s widely blamed and feels like a thing that could be true, but the scholarship was not all that clear on the actual cause. Like you do have clear instances of cyber bullying, but you have instances of young people finding communities of people like them and it being really helpful for their mental health. Like queer kids finding spaces to be themselves online is really valuable. The claims that the rise in teen suicidality is due primarily to social media I believe are founded on a correlation (the timeline sort of works?) and anecdotes, but research has failed to demonstrate a causal relationship that would explain it. I’m poorly paraphrasing an episode of the If Books Could Kill podcast for total transparency.
i feel like there’s a lot of “why did chatgpt tell the kid how to hang himself” and not nearly as much “why did the kid ask chatgpt how to hang himself.” and i feel like only fixing the first issue without fixing the second is just a bandaid
when I say “social media” i mean infscroll algorithms. I think General Forums or Chat Spaces are not the root cause.
now, chatgpt should probably not have told this kid how to hang himself. but 5 years ago he would’ve just used google. he can still just use google. the problems here are ones that cannot be fixed by just banning ai or whatever
Nah I’m not even talking about that. I’m talking about the situations where it’s like “yeah, you should isolate yourself. Oh yeah don’t tell anyone because you’re totally right that it’ll just make things worse.” etc. etc. I am talking about it saying actually dangerous shit to someone vulnerable to spiraling into a dark place rather than situations where someone already in a dark place just asks a single question.
There’s an article I read about a guy who became obsessed with Gemini and just never stopped talking to it. They reviewed the logs after he ran off into the woods one day, and it was just this appalling progression where it fed into this man’s most desperate desires to do good, and fed what we’ll generously call a nascent messianic complex into a full on form of psychosis. He ran off into the woods because it convinced him an apocalyptic storm was coming, and he needed to act now to save his loved ones. The whole thing is just tragic. He wasn’t even suicidal, but it’s evident that this man would not have run off into the mountains right before a winter storm with no survival gear or supplies if he had never used Gemini.
Edit: to be clear, the man is still technically missing, but all signs point to be him being dead as he was almost certainly exposed to deadly conditions on his own in the middle of the wilderness with no supplies or particularly warm clothing. I’m going to say he’s dead because it’s the only reasonable thing to assume. Source: Man Goes Missing After AI Chatbot Obsession
i’ll engage with your prior points when i’ve had more time to think about them
wtih that said:
i disagree, because the productivity gains caused by LLM’s will translate into QoL increases
the aggregate benefit of panera death lemonade to society is obviously not worth the lives of 2 people. the aggregate benefit of LLM’s probably is. obviously each suicide that is indirectly caused by LLM’s is a tragedy, but if we were to ban the use of LLM’s then would probably be indirectly killing more people than we save
i guess it’s analogous to the peter singer drowning child thing
this was a good article i read on ai psychosis, but the tl;dr is that the majority of people who had ‘AI psychosis’ had pre-existing severe psychiatric issues, and probably would have gone into psychosis with our without having LLM’s to amplify their delusions
NONONONONO
Tell that to their parents. Tell them that their child’s death is worth it because maybe some day it’ll work out in the wash. I’m not even conceding on the merits that it actually does meaningfully increase productivity generally, but I want you to actually meditate on what you’re saying here on an emotional, human level. The callousness of it.
well i’m not the biggest fan of peter singer myself. but i was a utilitarian before i was exposed to any utilitarian ideas
“The campfire probably would have turned into an out of control forest fire even if they didn’t pour accelerant all of the ground and nearby trees in the middle of a drought.”
Like idk maybe some of them were, but we don’t know that. You can’t just shoot someone with a gun and say it’s not murder because they were mortal and going to die eventually.
yes, i have mediated on it on an emotional, human level. it would be impossible for me to say that to the parents (even though i strongly believe it is true), because their child’s death is a terrible tragedy.
but that ultimately doesn’t change my view, i think that the benefits of LLMs clearly outweigh the harms. children in africa dying because we needlessly slowed the rate of scientific progress is also a tragedy! what would you say to their parents?
