brian is live on unlonely
@briang
so much woke slander comes from creating “villains” out of thin air. ex: been seeing a lot of chatgpt-is-too-woke slander on Twitter and every time I try the prompt myself, it answers it normally without triggering its compliance warnings. blatant fear mongering and misinformation. chatgpt is a tool. nothing mor
4 replies
0 recast
0 reaction
Dan Romero
@dwr.eth
While I agree it's mostly people to create a stir for social media attention, and I don't think it will be a big issue in the long run (see my recent discussion with Brian Chau), c'mon: https://i.imgur.com/zbeOl8d.png
2 replies
0 recast
0 reaction
brian is live on unlonely
@briang
ok here’s a follow up question: do you think this is an intentional decision by openAI pushing a specific agenda? or is it an example of a limitation of the current technologies ability to discern between controversial topics
1 reply
0 recast
0 reaction
Dan Romero
@dwr.eth
They published a paper saying they were doing so! They're not trying to hide it. https://openai.com/blog/improving-language-model-behavior/ https://i.imgur.com/Fj776G0.png
3 replies
0 recast
0 reaction
Dan Romero
@dwr.eth
To be clear: this is likely the right commercial trade-off for them. And I also don't think this will be a big issue in the long run given open source models. But to claim it's not heavily biased is ignoring what OpenAI is saying they are doing.
0 reply
0 recast
0 reaction
brian is live on unlonely
@briang
ok so under political opinion it says it’s non-partisan. so it not writing a poem praising Donald Trump seems like a mistake of their AI labeling. Ex: here’s me asking to say something positive about Donald https://i.imgur.com/4VS1hNv.jpg
0 reply
0 recast
0 reaction
keccers
@keccers.eth
I have an egregious OpenAI story along these lines I can only tell in person from when they were releasing DALLE They absolutely have an agenda I’ll back u up on that all day
0 reply
0 recast
0 reaction
