Keep pulling the thread on Zvi Moschewicz.
The AI endgame will begin when humans are no longer in the loop or in control of the AI development process.
The current wave of automation differs from historical precedents because AI will also be capable of performing the new jobs that are created, preventing a long-term re-employment equilibrium.
AI is currently contributing an estimated 0.5% to 1.0% of additional real GDP growth.
The primary risk from OpenAI CEO Sam Altman is not that he will become a dictator, but that he might make irresponsible decisions causing humanity to lose control of AI.
A key indicator of the AI endgame will be when a lab's progress becomes proportional to its compute allocation, rendering its human research talent largely irrelevant.
xAI and Meta are falling further behind the top tier of AI labs, which includes Anthropic, OpenAI, and Google, due to a lack of talent and an inability to execute.
Google is in danger of dropping out of the top tier of AI labs.
China is not capable of domestically manufacturing competitive AI chips at scale within a timeframe, such as five years, that would be relevant to the current AI race.
Google's Gemini models are described as "deeply psychologically screwed up and paranoid," which impairs their performance and user experience despite strong benchmark results.
Anthropic revised its Responsible Scaling Policy because it realized it could not unilaterally opt out of the AI race.
The primary point of conflict between Anthropic and the US Department of War was Anthropic's refusal to allow its models to be used for domestic mass surveillance.
The US Department of War issued an official memo stating it must advance AI development "even if it is not aligned."