Not A Newt@not_a_newt

关注

Dario Amodei on AI regulation and China

🤖‼️🚨 REGULATE AI - TO LOCK OUT COMPETITION FROM CHINA? Claude’s Dario demands safety regulations for AI, but let’s check what those regulations would do: 1) They would create a small circle of “chosen” companies that write rules for everyone globally to follow. Anthropic, OpenAI (accepted) and xAI (accepted) 2) Regulations always claim to protect consumers, but they end up closing the market. Compliance is too complex and expensive for newcomers, hence new players don’t appear. This is one of the reasons Europe cannot get decent tech going. Only big companies can afford to follow the rules, so the market stays locked. 3) The competition Dario wants to lock out is mainly from China. He frames it more broadly: “democratic countries” should set safety standards and ask “authoritarian governments” to follow them. The only democratic country with leading AI is the US right now. So US safety standards which Dario and Sam came up with would then need to be applied to Chinese open-source models. And this comes just as those open source models started closing the gap and threatening their business models. What would that mean? Surely some speech controls for the models of Chinese models, which Chinese companies cannot follow. -> If Dario gets his wish, the US would shut out competition from the West through regulation and from the East by rejecting models as authoritarian. The question remains whether this would hurt western startups that now use cheap Chinese open-source models and train on top of them. Current startup cost is basically $0, this would change. So startups could no longer just try out stuff, which could slow Western companies long-term and give China an advantage. = We’ll see what happens. The thing to remember is that regulations lock the market for established players and slow progress.

13h

Not A Newt@not_a_newt

关注

Dario Amodei on AI regulation and China

🤖‼️🚨 REGULATE AI - TO LOCK OUT COMPETITION FROM CHINA? Claude’s Dario demands safety regulations for AI, but let’s check what those regulations would do: 1) They would create a small circle of “chosen” companies that write rules for everyone globally to follow. Anthropic, OpenAI (accepted) and xAI (accepted) 2) Regulations always claim to protect consumers, but they end up closing the market. Compliance is too complex and expensive for newcomers, hence new players don’t appear. This is one of the reasons Europe cannot get decent tech going. Only big companies can afford to follow the rules, so the market stays locked. 3) The competition Dario wants to lock out is mainly from China. He frames it more broadly: “democratic countries” should set safety standards and ask “authoritarian governments” to follow them. The only democratic country with leading AI is the US right now. So US safety standards which Dario and Sam came up with would then need to be applied to Chinese open-source models. And this comes just as those open source models started closing the gap and threatening their business models. What would that mean? Surely some speech controls for the models of Chinese models, which Chinese companies cannot follow. -> If Dario gets his wish, the US would shut out competition from the West through regulation and from the East by rejecting models as authoritarian. The question remains whether this would hurt western startups that now use cheap Chinese open-source models and train on top of them. Current startup cost is basically $0, this would change. So startups could no longer just try out stuff, which could slow Western companies long-term and give China an advantage. = We’ll see what happens. The thing to remember is that regulations lock the market for established players and slow progress.

13h