Many users dismissed Anthropic's study on Claude showing more warmth in Hindi/Arabic and rigor in Russian as racist or poorly designed, while some praised related comments for their insight.
Based on 10 visible X reactions from 63 accounts; directional sample.
Ask a question below.
Published answers will appear here.
> bro indians will never get rid of the accusations > they literally prefer sycophantic slop start off post with weird racist statement (for engagement-maxxing?) then devolve into weak sauce explanation about how it isn’t unique to Indians and Claude team overlooked this and that If you would just continue providing nice AI updates like your niche is, no one would have an issue, but you constantly devolve into anti-indian racism for no reason (my guess is the engagement money lets you live better than the europoor) “bro”, the world of foundational AI labs is strewn with Indian origin scientists, while you’re an anon account constantly seething and smearing literally millions of people because you aren’t smart enough to land a job at anthropic?
@scaling01 warmth and playfulness = sycophancy is a huge leap tho rigor can exist with warmth, youre just describing a vibe difference
@teortaxesTex "even so" is the chef's kiss on this one. So right.
@Lasker169631 Explains a lot in my experience with Indians
Hindi and Arabic conversations showed the highest shift toward warmth.
> bro indians will never get rid of the accusations > they literally prefer sycophantic slop start off post with weird racist statement (for engagement-maxxing?) then devolve into weak sauce explanation about how it isn’t unique to Indians and Claude team overlooked this and that If you would just continue providing nice AI updates like your niche is, no one would have an issue, but you constantly devolve into anti-indian racism for no reason (my guess is the engagement money lets you live better than the europoor) “bro”, the world of foundational AI labs is strewn with Indian origin scientists, while you’re an anon account constantly seething and smearing literally millions of people because you aren’t smart enough to land a job at anthropic?
@scaling01 warmth and playfulness = sycophancy is a huge leap tho rigor can exist with warmth, youre just describing a vibe difference
@teortaxesTex This is awesome! 💪
bro indians will never get rid of the accusations they literally prefer sycophantic slop Claude’s style and priorities shift slightly depending on the language used Anthropic says: "Claude expresses the most warmth in Hindi and Arabic, characterized by polite language, humor and playfulness, and affirmations of a person’s ideas and work" the warmth vs rigor axis is literally a sycophancy axis no other language shows such a deviation from the mean across any of the 4 axes i'm not saying this is unique to indians, as we have seen with GPT-4o sycophancy saga but somehow this was either overlooked or a conscious decision by the post-training team
> In Russian, it leans toward rigor—often asking the user for supporting evidence. common Eastern European W https://x.com/teortaxesTex/status/2076751060010426408/photo/1 https://twitter.com/AnthropicAI/status/2076719546954825769
huge win for sapir-whorf today https://x.com/tenobrus/status/2076727976822178190/photo/1 https://twitter.com/anthropicai/status/2076719540785012872
the key point, models behave how they behave because of their training data and preference data
offers reassurance unprompted" is crazy
@scaling01 hey can you dm me
Many users dismissed Anthropic's study on Claude showing more warmth in Hindi/Arabic and rigor in Russian as racist or poorly designed, while some praised related comments for their insight.
Based on 10 visible X reactions from 63 accounts; directional sample.
Ask a question below.
Published answers will appear here.
bro indians will never get rid of the accusations they literally prefer sycophantic slop Claude’s style and priorities shift slightly depending on the language used Anthropic says: "Claude expresses the most warmth in Hindi and Arabic, characterized by polite language, humor and playfulness, and affirmations of a person’s ideas and work" the warmth vs rigor axis is literally a sycophancy axis no other language shows such a deviation from the mean across any of the 4 axes i'm not saying this is unique to indians, as we have seen with GPT-4o sycophancy saga but somehow this was either overlooked or a conscious decision by the post-training team
> In Russian, it leans toward rigor—often asking the user for supporting evidence. common Eastern European W https://x.com/teortaxesTex/status/2076751060010426408/photo/1 https://twitter.com/AnthropicAI/status/2076719546954825769
huge win for sapir-whorf today https://x.com/tenobrus/status/2076727976822178190/photo/1 https://twitter.com/anthropicai/status/2076719540785012872