gloom

Singularitarians: We must align AI with human-friendly values
China: Hold our communal beer comrade

Singularitarians: We must align AI with human-friendly values
China: Hold our communal beer comrade https://open.substack.com/pub/chinatalk/p/censorships-impact-on-chinas-chatbots

Blogger at ribbonfarm.com, running summerofprotocols.com, host of /gloom ☀️

i feel like this is the bigges moat western AI companies have compared to china.

chinese ai devs must lobotomize their models or face jail time

If the alignment people have their way it’ll be the same in the US. Key alignment guy just got appointed head of AI safety at NIST. He’s a good technologist (helped invent RLHF) but has doomer attitudes

I'm good at making mistakes. Sometimes I improve, but I never make the same mistake twice.
https://makemake.site

If the alignment people have their way it’ll be the same in the US. Key alignment guy just got appointed head of AI safety at NIST. He’s a good technologist (helped invent RLHF) but has doomer attitudes https://en.wikipedia.org/wiki/Paul_Christiano_(researcher)

i played around with qwen-1.5-0.5b yesterday and its hilarious how biased it was. when asked about tiannamen square it kept ignoring the question.

also tried running gemma-2b and it was honestly respectably neutral:

i played around with qwen-1.5-0.5b yesterday and its hilarious how biased it was. when asked about tiannamen square it kept ignoring the question.

also tried running gemma-2b and it was honestly respectably neutral: https://twitter.com/makemake_kbo/status/1767020032368779605