4 ms·
This reminds of Haidt's contrived moral dilemmas that are designed to trip your moral sensors, even though you can't really rationally articulate why you find i
by paytonjjones 4mo ago
This reminds of Haidt's contrived moral dilemmas that are designed to trip your moral sensors, even though you can't really rationally articulate why you find it objectionable.
Realistically, I can't think of clear big or likely harms caused by this exploit. But I really really don't like this latent space existing in my AIs. It just makes me uncomfortable.
And over time I've learned to trust those moral intuitions more than I trust reason alone.
- superb_dev 4mo agoThere’s the obvious harm that some people are just not equipped to see these graphic images, especially with no warning. Like people who have trauma from being in or around the acts being depicted
- paytonjjones 4mo agoOh oh, I do research on this :) https://journals.sagepub.com/doi/10.1177/2167702620921341 https://journals.sagepub.com/doi/10.1177/2167702620921341 (Research aside, it seems unlikely to me that a lot of people would stumble on that prompt accidentally in any case)
- superb_dev 4mo agoFascinating! I’d be very interested in further research on people with trauma/PTSD
- paytonjjones 4mo agoYou might enjoy this, by a colleague of mine. It's a rarer situation, but this could be one harm pathway for those types of images. (In most cases, exposure is a good thing for people with PTSD) https://journals.sagepub.com/doi/10.1177/2167702620917459 https://journals.sagepub.com/doi/10.1177/2167702620917459
- queenkjuul 4mo agoExcept the 100,000 or so who read the initial prompt on Twitter?
- paytonjjones 4mo agoIf they saw it on Twitter then actively went and tried it, that wouldn't be very 'accidentally'
- qingcharles 4mo agoThe prompt has been going around for months. 99.9% of the output it generates is simply weird, in a funny way, not horrific like in the article.
- butlike 4mo ago> We found substantial evidence that trigger warnings countertherapeutically reinforce survivors’ view of their trauma as central to their identity. What I've suspected for a long time
- applfanboysbgon 4mo agoPerhaps those people can refrain from jailbreaking ChatGPT to produce graphic imagery. There is not a single person in the world who will type any of the prompts noted in the article by accident.