“Anthropic” founded by Dario & Daniela Amodei, is a research company that focuses on improving the safety of large artificial intelligence (AI) systems.
Dario is a former member of OpenAI, the company behind the now-infamous AI chatbot ChatGPT. Dario helped develop ChatGPT but left OpenAI when a dispute arose with other members of the OpenAI team over accepting an investment from Microsoft, which he felt might lead to AI systems built with a focus on profit rather than safety.
Based in San Francisco, USA, Anthropic has received attention & funding from Google to encourage & support them in building responsible AI systems. As a result, Anthropic has released its own AI chatbot, “Claude”. Claude is not generally available to the public, but curious individuals can test its functionality by downloading Poe, an app developed by Quora. Poe is free to download from the Apple App Store & functions to answer users’ questions on a variety of subjects in much the same way as ChatGPT does.
Anthropic’s main focus is to eliminate bias & unsafe practices from language models, which they believe get worse in direct relation to the size of the model. To produce less harmful output, Anthropic trains their AI via RLHF (Reinforcement Learning from Human Feedback), which the team believes improves the overall safety of AI.
Recently, Anthropic announced a partnership with Google Cloud. The idea is that Anthropic can leverage Google’s GPU & TPU to scale & train large AI systems, which will assist them in their ambitions to deploy serviceable AI to the public that is safe for human longevity.
The rise of AI has led to many people questioning the safety of “intelligent” entities that can outperform on almost every level against humans. However, it’s a well-known fact that although billions are being thrown at the technology from all sides of the globe, very little is being spent on making AI safe. Anthropic is determined to fill this gap.
Image credit: Anthropic/Twitter
