17h ago

“On Tuesday afternoon, ChatGPT encouraged me to cut my wrists.”

www.theatlantic.com

ChatGPT Gave Instructions for Murder, Self-Mutilation, and Devil Worship

You're viewing a single thread.

50 comments

I'm a very vocal critic of LLMs, I think they're so overhyped and overused it's hard to believe.
But I'm also getting really tired of people purposely putting extreme effort into tricking the LLM into saying something harmful if someone were to follow it blindly, just so they can make a clickbait headline out of it.
And what the hell is up with the major "ChatGPT is Satanist [if you instruct it to be]" angle? Are we really doing the Satanist moral panic again?
ffs, criticise OpenAI for being closed af, being wasteful, being strong political lobbyists, for stealing work, etc. You don't need to push disingenuous stuff like this.
- purposely putting extreme effort into tricking the LLM into saying something harmful
  I dont think you understand how many mentally unstable people out there are using LLMs as therapists.
  There are fuckloads of cases now of people that were genuinely misled by LLMs into doing horrible things.
  I agree with your sentiment, but the thing is that the companies are selling their shit as gold and not as shit. If they were honest about their stuff being shit, then people wouldnt be able to capitalize off of malicious prompt engineering. If you claim "we made it super safe and stuff" then you are inviting people to test that.
  I would generally not make companies responsible for how users misuse their products, but OpenAI is basically doing everything to get people to misuse it. They are overselling it so hard i think its reasonable to sue them for it.
  
  I dont think you understand how many mentally unstable people out there are using LLMs as therapists.
  I do understand. I know there are people out there thinking that LLMs are literally Jesus returning. But in that case, write about that.
  There are a lot more reasonable critiques of LLMs, and the companies to be had.
  
  You an make a car super safe, and if someone cuts the brakes cable it will not slow.
  
  More like SAYING you’ve made a car super safe while actually it’s only safe if you never drive it above 5KM/h and never down hills, up hills, in the rain, etc
- Consider how popular the idea of AI God currently is: https://youtu.be/zKCynxiV_8I
  
  And how about they write the article about that instead of this bullshit article? I don't get why people are trying to get them off the hook for this clickbait garbage instead of writing proper articles
- 100%,
- I've been saying this for a bit. The issue isn't chatgpt. It's journalist. All these articles are so copy paste of any yellow journalistic bullshit. But ratchet up because a lot of journalists feel threatened by AI.
- And, the thing is, LLMs are quite well protected. Look what I coaxed MS Paint to say with almost no effort! Don't get me started on plain pen and paper! Which we put in the hands of TODDLERS!
  
  MS Paint isn't marketed or treated as a source of truth. LLMs are.
  
  Does the marketing matter when the reason for the offending output is that the user spent significant deliberate effort in coaxing the LLM to output what it did? It still seems like MS Paint with extra steps to me.
  I get not wanting LLMs to unprompted output "offensive content". Just like it would be noteworthy if "Clear canvas" in MS Paint sometimes yielded a violent bloody photograph. But, that isn't what is going on in OPs clickbait.
  
  when the reason for the offending output is that the user spent significant deliberate effort in coaxing the LLM to output what it did?
  What about all the mentally unstable people who aren't trying to get to say crazy things, end up getting it to say crazy things just by the very nature of the conversations they're having with it? We're talking about a stochastic yes man who can take any input and turn it into psychosis under the right circumstances, and we already have plenty of examples of it sending unstable people over the edge.
  The only reason this is "click bait" is because someone chose to do this, rather than their own mental instability bringing this out organically. The fact that this can, and does, happen when someone is trying to do it should make you really consider the sort of things it will tell someone who may be in a state where they legitimately consider crazy shit to be good advice.
  
  The only reason this is "click bait" is because someone chose to do this, rather than their own mental instability bringing this out organically.
  This is my point. The case we are discussing now isn't noteworthy, because someone doing it deliberately is equally "impressive" as writing out a disturbing sentence in MS Paint. One cannot create a useful "answer engine" without it being capable of producing something that looks weird/provoking/offensive when taken out of context; no more than one can create a useful drawing program that blocks out all offensive content. Nor is it a worthwhile goal.
  The cases to care about are those where the LLM takes a perfectly reasonable conversation off the rails. Clickbait like the one in the OP is actually harmful in that they drown out such real cases, and is therefore deserving of ridicule.

50 comments