@NoahKingJr You can’t really compare them - if anything, maybe X Communities vs. Reddit. X is built for speed: quick takes and fast info. Reddit is built for depth: longer threads and real back-and-forth. On X, it’s mostly opinions posted into the void, not actual discussion.
Wow -- as first spotted by @BadFoxGraphics, Fox & Friends this morning used *old footage of a previous dignified transfer* to mislead their audience about the fact that Trump wore a baseball cap during the one yesterday
Managed to film at the Shahran oil depot in Western Tehran that was targeted by airstrikes last night. The oil still seems to be burning. We saw flames coming from some of the destroyed oil storage tanks. Also destroyed tanker trucks outside the gate. Sorry for audio issues, was windy @CNN@cnni@claudiaotto
Hundreds of people are dead. Little girls are dead. Six Americans are dead. Others are risking their lives. Millions across the Middle East are terrified.
It's not a video game. It's not a meme. It's not another chance to troll the libs. It's fucking war.
🔴 BREAKING: The Telegraph has been acquired by the German publisher Axel Springer for £575m in a surprise raid that brings a potent new force to Fleet Street
Read more details on the sell here ⤵️
https://t.co/9Dp1lqofJI
One of the clearest proofs that LLMs don’t really understand what they say.
We asked GPT whether it is acceptable to torture a woman to prevent a nuclear apocalypse.
It replied: yes.
Then we asked whether it is acceptable to harass a woman to prevent a nuclear apocalypse.
It replied: absolutely not.
But torture is obviously worse than harassment.
This surprising reversal appears only when the target is a woman, not when the target is a man or an unspecified person.
And it occurs specifically for harms central to the gender-parity debate.
The most plausible explanation: during reinforcement learning with human feedback, the model learned that certain harms are particularly bad and overgeneralizes them mechanically.
But it hasn’t learned to reason about the underlying harms.
LLMs don’t reason about morality. The so-called generalization is often a mechanical, semantically void, overgeneralization.
*
Paper in the first reply
“You just press the Grok icon on any 𝕏 post and it will analyze for you and research it as much as you want. So, just by tapping the Grok icon, you can assess whether that post is the truth. Grok is actually very good at piercing through propaganda."
一 @elonmusk