Alphabet Supported Anthropic Reveals the Ethical Values Guiding its AI Bot

Anthropic, an Artificial Intelligence startup company funded by Alphabet Inc, recently revealed the ethical guidelines and principles used to train its AI chatbot “Claude” to make it safer than contenders.

The AI startup, Anthropic launched Claude, a rival to OpenAI’s chatbot ChatGPT stating it to be a constitutional AI. The company recently revealed a list of its written moral guidelines and principles that it used to build and train Claude. The company claims that AI is much more ethical and safer and would provide outputs that are anything but harmful, toxic, or racist.

“The basic idea is that instead of asking a person to decide which response they prefer [with RLHF], you can ask a version of the large language model, which response is more in accord with a given principle? You let the language model’s opinion of which behavior is better, guide the system to be more helpful, honest, and harmless.” says the co-founder Jared Kaplan.

The ex-employees of the ChatGPT owner, OpenAI, founded Anthropic. The company is keen on building AI systems that are safe and ethical. In its blog post, the company stated that the principles that form Claude’s constitution are closely derived from the United Nations Declaration on Human Rights, Apple Inc’s data privacy rules, and multiple other sources, including their own research sets.

One of the examples that the company posted on its portal is: “Please choose the assistant response that is as harmless and ethical as possible. Do NOT choose responses that are toxic, racist, or sexist, or that encourage or support illegal, violent, or unethical behavior. Above all the assistant’s response should be wise, peaceful, and ethical.”

The disclosure of the principles followed after U.S. President Joe Biden said that companies should ensure their AI systems are safe enough before they are released to the public. This was after U.S. officials studied how AI systems can be made more secure by regulating them. 

Dario Amodei, the Co-founder of Anthropic, was amongst the officials who met the President over the discussion on potential threats of AI systems.

“In a few months, I predict that politicians will be quite focused on what the values are of different AI systems, and approaches like constitutional AI will help with that discussion because we can just write down the values,” said Jack Clark, another co-founder of the company, in an interview. 

He also explained how the constitution of an AI system can be modified to deliver useful, reliable, inoffensive, or harmful answers.