Critics counter that models caused widespread web spam and educational disruption.
They are so obsessed with models being dangerous they hyperstitioned that outcome
not dunking on Dario here bc Mythos is to GPT2 what a lion is to a bumblebee,
but it’s insane that GPT-2, and successive models, have had ~0% impact on causing bad universal outcomes & so SHOULD HAVE caused AI safety folks to update their priors YET none of them seem to have??
Negative users accuse AI safety advocates of overhyping risks as deliberate theater with no incentive to admit error, while positive users prefer non-alarmist views on tools like ChatGPT for education.
No Digg Deeper questions have been answered for this story yet.
"no bad outcomes"
meanwhile the internet is now 90% slop bots, slop accounts, slop websites, slop youtube videos and the education system is now a prompt: "chatgpt do my homework"
also, prior to Mythos models simply weren't a cyber threat
so it's not that safety doomers were wrong about bad outcomes, but that their calibration of what models are capable were off
not dunking on Dario here bc Mythos is to GPT2 what a lion is to a bumblebee,
but it’s insane that GPT-2, and successive models, have had ~0% impact on causing bad universal outcomes & so SHOULD HAVE caused AI safety folks to update their priors YET none of them seem to have??
@sporadica @scaling01 This very article tells you that there were warnings against misinformation and misuse. They were not claiming it would end the world.
@sporadica @scaling01 i've always been on the non-alarmist side. Honestly, I think "chatGPT do my homework" is a better thing for kids to do than the current thing that internet algos suck us into (for me its twitter & utube).
@beffjezos theyre so scared of ai doing bad they made it do bad on purpose for the papers. its theater
@scaling01 the internet was 90% junk before llms, now it’s just junk that has slightly lower entropy
@beffjezos legitimately
@Yim0e @sporadica @scaling01 1) ur profile is interesting & I bet I could learn about something new 2) I think any and all education matters start at the home. if the parent is not showing kid how to critically think & use chatGPT then this is all for nothing. we are humans at end of day, cannot change that.
@tamilna73049047 @sporadica @scaling01 For education the long term outcomes are possible atrocious
@tamilna73049047 @sporadica @scaling01 1) what makes you say that but thanks. 2) in an ideal world that’s true. But the state/institutions have a role to play for good reason. ‘Education’ might not even matter in a few years anyways lol, but the people who suffer are the current gen
@scaling01 Absolutely. I don't get what's so hard for people to understand here. There were different levels of risk at play then and now. And there will be different levels of risk in the future as AI gets better.
@scaling01 And yet we’re doing just fine. Society is more robust than you think.
@scaling01 and yet we’re all still here and life is mostly the same as it was in 2020. sure the internet has become a bit annoyingly polluted, but this is not the case AI safetyists were making then
@digiIov @scaling01 literally
1) I don't sports bet or whatever it is you do (see image, this is a good example) 2) I agree that institutions play a role, but to further the discussion we can use an analog: the internet + Wikipedia. In 3rd grade we were pulled into the library and told "never use wikipedia as a source". I don't know if that had the effect on our critical thinking that the school system thought it would. But, back then, nobody could have expected what internet & "fake news" and how that shit circulates as it does today. I guess my point is: you're right, the long term outcomes are possibly atrocious, but its because our institutions are unprepared... and the only reason we made it this far since ther advent of the internet is due to good family values and diversity and debate (which leads to critical thinking). I do not know if statements about the current gen suffering are valid. I see the opposite, we are in a decent nice world right now. Could be better, could be worse. alright gn lml thanks for the reply 🙏
@beffjezos Overhyping AI risks only spurs strict restrictions and slows real technical rollouts.
@beffjezos if there doing it models .. there probably doing it to real people .. lmao .. goberment crazy
@beffjezos It was inevitable...
@scaling01 Anthropic purpose of creation was AI Safety lets just remember that, they are just honoring their mission guys.
@beffjezos We didnt get a catastrophe we got autocomplete clearly the singularity is just buffering.. hyperstition only works if you compile it with enough GPUs..
@thiagoTF @beffjezos really says more about them than ai https://www.youtube.com/watch?v=R256W_xoemw