You Shouldn't Let ChatGPT Write Your Emails

Is it morally wrong to send emails that are mostly written by AI?

In this episode, I argue that the answer is often yes.

Rather than debating whether large language models are "good" or "bad" in general, I focus on a much narrower claim: sending emails that are primarily generated by an LLM is frequently a moral mistake.

I develop two arguments:

• It can involve a kind of deception by presenting AI-generated words as your own communication.
• It can be selfish and anti-relational by offloading the effort of communication while expecting others to engage with the result.

Along the way I discuss changing social norms around AI, whether widespread AI use changes the ethics, and why authenticity still matters even in an age of increasingly capable language models.

I also explain where I think AI use is perfectly appropriate—for editing, brainstorming, improving clarity, generating metadata, standardizing formats, and other cases where it genuinely serves the reader rather than replacing the human author.