In recent days I’ve ended up discussing LLMS with several intelligent, high-flying professionals, working in roles where LLM use is matter of routine. They all told me, independently, that on principle they refuse to say ‘please’ and ‘thank you’ to the models with which they interact. The common theme in their reasoning is that withholding politeness is an important way of maintaining the mental distinction between an LLM and a human.
I’ve been thinking about this ever since. Because I do say ‘please’ and ‘thank you’ to LLMs, also on principle, even though I agree that an LLM isn’t and never will be conscious.
Which of us is right?
The individuals I’ve discussed this with are all Christians, which means taking as read that humans have immortal souls, that participate in the divine. (I don’t think it’s possible to be a Christian without believing humans have souls.) Whether or to what extent other living creatures have souls has been a subject of lively Christian debate for a long time. But if you start from the view that both life and consciousness derive ultimately from God, that more or less forecloses the possibility of humans being able to create machines that can think.
Christian professionals who are expected to use LLMs at work are thus in an odd position. Here is a technology that your paradigm says isn’t conscious, won’t ever become conscious, and definitely doesn’t have a soul. But to get the best results from it, you need to interact with it in a way that closely resembles the way you would interact with a human being. How should you think about this?
I’ve been pondering this since my recent visit to Anthropic, not least because I decided, in the course of preparing for that meeting, to spend a bit of time with their LLM, “Claude”.