Anthropic, launched in 2020 by former OpenAI researchers, is known for its Claude-2 model, which can process prompts containing up to 100,000 tokens, making it suitable for tasks like summarizing lengthy legal documents. The company will use Amazon's AWS Trainium and AWS Inferentia chips for model development and AI safety projects, and will also collaborate with Amazon in developing future versions of these chips.
Key takeaways:
- Amazon.com Inc. will invest up to $4 billion in Anthropic, a San Francisco-based startup that develops generative artificial intelligence software.
- The investment will help Anthropic develop a highly advanced foundation model, Claude-Next, which could be too hardware-intensive for even the world’s fastest supercomputers.
- As part of the investment, Amazon will buy a minority stake of unspecified size in Anthropic for $1.25 billion, with the option to eventually invest another $2.75 billion.
- Anthropic will run the majority of its workloads on Amazon Web Services (AWS), using AWS's internally-developed Trainium and Inferentia chips to support its research and improve them through collaboration.