Anthropic

Anthropic is an artificial intelligence safety startup based in San Francisco that develops beneficial AI systems designed to be helpful, harmless, and honest. Founded in 2021, Anthropic aims to shape the future of AI in a way that benefits society through both technical research and applied engineering.

Anthropic takes a principles-first AI safety approach focused on reliability, interpretability and constitutional AI. Constitutional AI sets human-compatible constraints on system behavior, rooted in norms like honesty, pretend play, and self-supervision. This contrasts with intelligence-first approaches fixated on optimizing arbitrary goals or metrics.

Anthropic's first product is Claude, a conversational AI assistant designed for natural language understanding and common sense reasoning. Built on a Constitutional AI architecture, Claude aims for truthfulness, thoughtfulness, and transparency. The underlying technology combines self-supervision, natural language feedback, and human conversation data to produce safe dialog. They have and developed advanced generative models like CLARA.

Co-founded by AI researchers Dario Amodei and Daniela Amodei, Anthropic brings together leading experts in AI safety, ethics, linguistics and related fields. The company has assembled an advisory board with responsible AI champions like Stuart Russell, Herb Lin and Tom Dietterich. Backed by top Silicon Valley investors, Anthropic collaborates with academics, researchers, regulators, and civil society to steer AI in ways that uplift humanity. As AI capabilities advance, Anthropic aims to demonstrate solutions that align emerging technology with human values.

On September 25, 2023, Anthropic announced a collaboration with Amazon Web Services (AWS). Amazon will invest up to $4 billion in Anthropic and become a minority stakeholder. Anthropic will use AWS Trainium and Inferentia chips for its AI models and collaborate on developing the next generation of these chips. Amazon engineers will access Anthropic's models through the Bedrock platform to enhance natural language capabilities. The agreement also focuses on expanding access to Anthropic's AI systems for AWS customers and implementing safety best practices.