Anthropic's Claude chatbot drives 26% of research and development
Anthropic's Claude chatbot is stepping up in a big way: it's now behind 26% of the company's research and development, up from effectively nothing earlier this year.
Claude is also teaming up with staff on 90% of their work, showing just how central it's become to building smarter AI.
Anthropic to grant evaluators employee-like access
Anthropic says it sets aside 6% to 12% of its computing resources for AI safety checks, depending on the project.
To boost accountability, they're planning to bring in third-party evaluators who'll get insider access similar to employees.
CEO Dario Amodei is openly pushing for government rules to manage AI risks, a different take from OpenAI CEO Sam Altman and NVIDIA Corp. CEO Jensen Huang, who say AI companies would safely pace AI on their own, and Meta Platforms Inc. CEO Mark Zuckerberg, who says AI companies should rely on independent evaluators and advisers to ensure that models are safe.
The debate over how best to keep AI safe is definitely heating up.