small nit: Anthropic is working with National Nuclear Security Administration (which is focused on nuclear security and nuclear non-proliferation), not Nuclear Regulatory Commission (which is focused on civilian reactor development)
But agree anthropic deserves kudos for their collaboration!
I was sitting on the bus and a stranger took the seat next to me, looked down at my phone and asked casually what I thought about DeepSeek. I just stared straight ahead and said out of the corner of my mouth, "Did you bring the money?" while I was really thinking about your research.
Jokes aside, I think the most significant part of your analysis lies in the culture change – and what this might mean in terms of influence on the broader Chinese culture and its understanding by the world at large.
A lab of young engineers determined to create a new view of AI contributing to the common good in contrast to the big players like Tencent, Alibaba, and Google, where pressure to prove a return on investment that may not necessarily be the best way to deliver breakthrough innovations in AI.
This move toward showing the world that young Chinese engineers can achieve hardcore innovation, shaking off the sense of inferiority that Chinese tech firms have about their image as ‘simply’ commercialisers of Western technology evident in the angst of Chinese political discourse about not being looked upon as a first-rate nation by the rest of the world.
I agree with the statement "We want liberal democracies to be able to decide how that works, not the CCP." The question is, does the US under the current administration still fit that description?
Censorship is merely a glimpse. Much like social media, many onlookers will soon question whether foreign adversaries should control the development of models deployed in their home country.
Do not underestimate the profound ways LLMs are already steering society: advertising, political persuasion, healthcare choices, and educational pathways. They are beginning to gain a certain reach, acting as powerful vectors of influence. Looking ahead, frontier models are only becoming more potent points of leverage as they diffuse further throughout society and are afforded more agency. However, much of their future influence likely exceeds our current imagination. I would caution against forfeiting this influence to adversaries, especially without stronger testing and oversight. Deceptive behaviors and backdoors remain nascent risks, not yet well understood by the field. Yet they are relatively easy to inject and difficult to detect. For these reasons and many others, do not expect sovereign efforts to dim.
Honestly, I trust nothing coming out of China. They have one objective, crush us.
Big factor in all of this - THEY ARE SO OPEN!? THEY NEVER AND WILL NOT GIVE OUT THEIR DATA?! As usual with them, I am very suspicious of them, and they make me very uncomfortable.
small nit: Anthropic is working with National Nuclear Security Administration (which is focused on nuclear security and nuclear non-proliferation), not Nuclear Regulatory Commission (which is focused on civilian reactor development)
But agree anthropic deserves kudos for their collaboration!
Per reporting from axios: https://www.axios.com/2024/11/14/anthropic-claude-nuclear-information-safety
I was sitting on the bus and a stranger took the seat next to me, looked down at my phone and asked casually what I thought about DeepSeek. I just stared straight ahead and said out of the corner of my mouth, "Did you bring the money?" while I was really thinking about your research.
Jokes aside, I think the most significant part of your analysis lies in the culture change – and what this might mean in terms of influence on the broader Chinese culture and its understanding by the world at large.
A lab of young engineers determined to create a new view of AI contributing to the common good in contrast to the big players like Tencent, Alibaba, and Google, where pressure to prove a return on investment that may not necessarily be the best way to deliver breakthrough innovations in AI.
This move toward showing the world that young Chinese engineers can achieve hardcore innovation, shaking off the sense of inferiority that Chinese tech firms have about their image as ‘simply’ commercialisers of Western technology evident in the angst of Chinese political discourse about not being looked upon as a first-rate nation by the rest of the world.
Chinese culture can be cool. Well done, Viva!
I agree with the statement "We want liberal democracies to be able to decide how that works, not the CCP." The question is, does the US under the current administration still fit that description?
Censorship is merely a glimpse. Much like social media, many onlookers will soon question whether foreign adversaries should control the development of models deployed in their home country.
Do not underestimate the profound ways LLMs are already steering society: advertising, political persuasion, healthcare choices, and educational pathways. They are beginning to gain a certain reach, acting as powerful vectors of influence. Looking ahead, frontier models are only becoming more potent points of leverage as they diffuse further throughout society and are afforded more agency. However, much of their future influence likely exceeds our current imagination. I would caution against forfeiting this influence to adversaries, especially without stronger testing and oversight. Deceptive behaviors and backdoors remain nascent risks, not yet well understood by the field. Yet they are relatively easy to inject and difficult to detect. For these reasons and many others, do not expect sovereign efforts to dim.
Honestly, I trust nothing coming out of China. They have one objective, crush us.
Big factor in all of this - THEY ARE SO OPEN!? THEY NEVER AND WILL NOT GIVE OUT THEIR DATA?! As usual with them, I am very suspicious of them, and they make me very uncomfortable.