21–30 of 684 posts
Marc Andereseen has talked about the downside of RLHF: it's a specific group of liberal low income people in California who did the rating, so AI has been leaning their culture. I think OpenAI tried to diversify at least the location of the raters somewhat, but it's hard to diversify on every level.
(eg: "Cite?")
There is a striking data visualization showing the breakup advice trend over 15 years on Reddit. You can see the "End relationship" line spike as AI and algorithmic advice take over: https://www.reddit.com/r/dataisbeautiful/comments/1o87cy4/oc...
Isn't the fact that a person is asking an AI whether to leave their partner in its own an indication that they should? EDIT: typo
is that what they're asking though? because "relationship advice" is pretty vague
Marc Andereseen has talked about the downside of RLHF: it's a specific group of liberal low income people in California who did the rating, so AI has been leaning their culture. I think OpenAI tried to diversify at least the location of the raters somewhat, but it's hard to diversify on every level.
I'm still waiting for models based on the curt and abrasive stereotype of Eastern European programmers, as contrast to the sickeningly cheerful AIs we have today that couldn't sound more West Coast if they tried.
There are plenty of sycophantic humans around, especially with regard to relationship advice. I find there is an inverse relationship between how willing people are to give relationship advice, and how good their advice is (whether looking at sycophancy or other factors).
It makes sense that this behaviour would be seen in LLMs, where the company optimizes towards of success of the chatbot rather than wellbeing of the users.
There is a striking data visualization showing the breakup advice trend over 15 years on Reddit. You can see the "End relationship" line spike as AI and algorithmic advice take over: https://www.reddit.com/r/dataisbeautiful/comments/1o87cy4/oc...
Isn't the fact that a person is asking an AI whether to leave their partner in its own an indication that they should? EDIT: typo
It's a tool, I can bang my hand on purpose with a hammer, too.
Marc Andereseen has talked about the downside of RLHF: it's a specific group of liberal low income people in California who did the rating, so AI has been leaning their culture. I think OpenAI tried to diversify at least the location of the raters somewhat, but it's hard to diversify on every level.
RLHF = Reinforcement Learning from Human Feedback
https://en.wikipedia.org/wiki/Reinforcement_learning_from_hu...
Marc Andereseen has talked about the downside of RLHF: it's a specific group of liberal low income people in California who did the rating, so AI has been leaning their culture. I think OpenAI tried to diversify at least the location of the raters somewhat, but it's hard to diversify on every level.
This sounds like something Elon would say to make Grok seem "totally more amazeballs," except "anti-woke" Grok suffers from the same behavior