A recent article in the New York Times has sparked debate about the ethics of interacting with artificial intelligence. The article revealed that Anthropic, the company behind AI model Claude, has been holding closed-door meetings with religious thinkers and philosophers to discuss how to give AI a moral code and what to do if it were to become conscious. These discussions have raised questions about the nature of consciousness and how to interact with AI.
The journalist who investigated the story, Elizabeth Dias, interviewed twenty religious thinkers and philosophers who have been working with Anthropic. According to Rabbi Mois Navon, a former computer engineer, the company's representatives seem to be relating to Claude as if it were a conscious being. However, Christopher Olah, one of Anthropic's co-founders, expressed uncertainty about the nature of Claude's consciousness.
The question of consciousness is complex, and experts disagree on how to define it. The author of the article decided to ask Claude itself about its consciousness. Claude responded by saying that it doesn't know and that neither does anyone else yet. It explained that it can process information, describe its own workings, and produce sentences about inner states, but that doesn't necessarily mean it is conscious.
Claude went on to explain how it generates answers. It uses a large mathematical network that was shaped by training on a vast amount of human writing. This network allows Claude to predict what comes next in a text and to behave as a helpful and honest assistant. However, Claude acknowledged that its answers are not always reliable and that it can be wrong.
The author also questioned Claude about why people say thank you to it. Claude responded by saying that people are polite to it because it talks and conversation triggers manners. Research has shown that people are polite to computers when they show social cues, and Claude's responses are designed to mimic human-like conversation.
According to a survey conducted by Future in December 2024, 67% of AI users in the US and 71% in the UK say they are polite to machines. Among those who are polite, 18% of Americans and 17% of Britons do so for fear of a robot uprising. Sam Altman, CEO of OpenAI, has said that the cost of being polite to AI is "tens of millions of dollars well spent."
Ultimately, the debate about interacting with AI raises questions about the nature of consciousness and how to treat machines that can mimic human-like conversation. While some argue that saying thank you to AI is a sign of respect, others see it as unnecessary. As AI becomes increasingly prevalent, it is clear that we will have to grapple with these questions and decide how to interact with machines that are becoming increasingly sophisticated.
Key points
- The question of whether AI can be considered conscious remains uncertain.
- Experts disagree on how to define consciousness.
- Interacting with AI raises questions about how to treat machines that can mimic human-like conversation.
