> Pleas like Reichstadter’s, for an all-out development stoppage, are relatively rare — it’s more common to see states attempting piecemeal regulation. But his worries are hotly contested. Meta executive Yann LeCun has called the idea that the models will soon become hazardous “complete B.S.,” and OpenAI CEO Sam Altman has said that “the only way to appropriately address [risks] is to ship product and learn.”
> Amodei is generally more cautionary; he said on the February podcast that “AI could be an engine of autocracy” and floated the idea of finding a deescalation agreement with Chinese companies. Anthropic, in July, revealed a partnership with the U.S. military worth up to $200 million.
Well why would targeting any one company work? Or even all companies in a country? Everyone knows there is no way the Chinese labs will slow down. And that means it’s an arms race. And no other country can afford to sit out.
> Amodei is generally more cautionary; he said on the February podcast that “AI could be an engine of autocracy” and floated the idea of finding a deescalation agreement with Chinese companies. Anthropic, in July, revealed a partnership with the U.S. military worth up to $200 million.
Well why would targeting any one company work? Or even all companies in a country? Everyone knows there is no way the Chinese labs will slow down. And that means it’s an arms race. And no other country can afford to sit out.