Keep pulling the thread on Peter Wildeford.
The time horizon for AI models, a measure of the complexity of tasks they can complete, is doubling approximately every 104 days.
Anthropic is posting $30 billion in annualized recurring revenue (ARR), outpacing the growth of companies like Google and Uber.
Anthropic's Claude Mythos model has demonstrated the ability to find software vulnerabilities and engage in hacking.
Peter Wildeford predicts that AI will be responsible for an additional one percentage point of unemployment in the United States within the next year.
Anthropic's Claude Mythos model has discovered tens of thousands of previously unknown vulnerabilities, many in well-tested systems where human experts had failed to find them.
The vulnerability discovery capabilities of Anthropic's Claude Mythos model have been validated by the UK government, Google, and Microsoft.
Extrapolating current trends, AI models are predicted to achieve time horizons equivalent to 90-100 hours of human labor by the end of 2024.
Based on the Remote Labor Index from the Center for AI Safety and Scale AI, AI models are predicted to be able to perform over half of remote freelance tasks by the end of 2027.
China is determined to build its domestic semiconductor capability regardless of U.S. export controls, driven by a strategic goal of self-reliance.
Peter Wildeford believes the current AI market is not a bubble, citing clear demand and revenue paths for companies like Anthropic.
Anthropic currently has more demand for its AI products than it has the capacity to serve.
Anthropic and OpenAI have consistently beaten their own revenue projections.