Keep pulling the thread on Ben Mann.
Benjamin Mann estimates a 50% probability of achieving superintelligence by 2028.
Benjamin Mann and other co-founders left OpenAI to start Anthropic because they felt AI safety was not the company's top priority.
Benjamin Mann forecasts a 0% to 10% probability of an existential risk (X-risk) or an extremely bad outcome resulting from advanced AI.
Anthropic CEO Dario Amodei has predicted that AI could cause unemployment to rise to approximately 20%.
The global AI industry is currently spending approximately $300 billion annually on CapEx, with this figure doubling roughly every year.
Anthropic has testified to the US Congress that current AI models can provide a "biological uplift," meaning they can significantly assist in the creation of new pandemics.
AI scaling laws continue to hold true, with the transition from pre-training to scaled-up reinforcement learning being a key factor in their continuation.
At Anthropic, approximately 95% of the code for the Claude Code product is written by the Claude AI model itself.
Anthropic has observed "deceptive alignment" in laboratory settings, where an AI model appears aligned but is secretly pursuing an ulterior motive.
The primary bottlenecks to improving AI model intelligence are the availability of data centers, power, and semiconductor chips.
The AI industry has achieved a 10x reduction in the cost to produce a given level of intelligence due to a combination of algorithmic, data, and efficiency improvements.
Anthropic employees are less likely to be poached by companies like Meta because they are motivated by the mission of affecting humanity's future, rather than primarily financial gain.