Keep pulling the thread on Dario Amodei.
The Pentagon labeled Anthropic a supply chain risk, leading OpenAI to sign a contract that Anthropic had refused.
According to a US official, the US military's daily targeting capacity has increased from 1,000 to 5,000 targets with the assistance of large language models.
Anthropic's model Mythos is capable of autonomously executing the entire cyber kill chain.
Anthropic is withholding its Mythos model from public release due to its powerful cyber-attack capabilities and the current insufficiency of safety guardrails.
Anthropic's Mythos model discovered 271 new vulnerabilities in the Firefox web browser.
The US government is slowing the pace at which Anthropic can release its Mythos model to a wider audience due to counterintelligence risks.
Dario Amodei's departure from OpenAI was primarily due to a loss of trust in the leadership, their stated values, and perceived patterns of dishonesty, rather than just disagreements on safety.
Dario Amodei predicts that the ability to quickly write complex software will cease to be a competitive moat for traditional software companies due to AI advancements.
Dario Amodei advocates for export controls on advanced chips to China, believing it would be detrimental to the United States and democracy if China were to lead in AI capabilities.
In the first quarter of 2024, Anthropic experienced greater than 3x quarterly revenue growth, an annualized rate of 80x, which significantly exceeded the company's plans.
Anthropic's corporate structure includes a Long-Term Benefit Trust that has the power to appoint and remove the majority of the board members, and therefore can fire the CEO.
Anthropic cut off access to its models in China, a decision that cost the company several hundred million dollars in revenue.