At some point in the future, artificial intelligence may provide meaningful improvements to data cross-referencing, statistical analysis, tracking of atmospheric and ocean fluid dynamics, and advancing medical science. Some researchers claim they are already benefitting in these ways. None of that means we should accept "generative AI" intefering with our interpersonal communications.
Right now, major corporations that have gained access to people's personal communications by providing email, notetaking, and videoconferencing services, are adding "writing assistants" and transcription services that use generative AI to create text.
Right now, major corporations that have gained access to people's personal communications by providing email, notetaking, and videoconferencing services, are adding "writing assistants" and transcription services that use generative AI to create text.
- In rare cases, users can download an entire large language model to their own device, then run dedicated software to generate text based on that LLM, without sending their prompts or even their generated output to remote servers.
- In the vast majority of cases, however, far more than 99%, the only way to use these services is to send original content (a text prompt, a voice recording or recording of an entire meeting) to an outside service, whose servers run the LLM and generate new content based on what was submitted.
- Many companies providing these services have contracts with Open AI or another provider; the terms of those contracts are often interpreted differently by the different parties, with "zero day retention" sometimes including clauses that allow for storing user inputs and using them for quality assurance.
- In other words, whenever you use a writing or transcription assistant, you are sending sensitive, highly personal information to businesses that have an interest in reusing that information to generate smart-seeming responses to future prompts.
So, there is a massive data-security and intellectual property problem at the heart of this new practice. There is something else, however, which many people might care about even more.
When I receive an email, I weigh the significance of that communication based on its author—the person or organization that created the communication and, apparently, wants me to know something about their thinking, feelings, or work. Personal connections mean more than mass mailings; entities I know and admire mean more than spam emails.
We all tend to prioritize engaging with communications from people we know, trust, and admire. Why would we think those we communicate with deserve less?
When you generate an email using a "writing assistant", you remove yourself as author and substitute an AI chatbot, whose boilerplate best-guess you then pass off as your own writing. Among the many things wrong with this:
- You are deceiving your reader(s), asking them to interact with a chatbot, while pretending you wrote the email.
- You are asking them to spend time reading bot-generated content you did not have the patience, knowledge, or concern to write on your own.
- You are corrupting the relationship between yourself and that person with deception and time-wasting, and with what might be extraneous information unhelpful to everyone involved.
- By asking a bot to do your writing for you, you also remove your own judgment, principles, and values from the output.
- All of this degrades the quality of the relationship between yourself and the reader(s), and with it either your personal or professional connections.
Beyond all of this, you subject yourself to a peculiar, deep form of self-sabotage: You distance yourself from understanding of your own communications. Think of what it would feel like if someone shared something meaningful with you, and you find yourself embarrassed by not having paid full attention to it. Now think of how you look when you have that experience with your own writing and communications.
There is no good reason to use a chatbot to write an email or to create even the outline of a workplace report. People are trusting you to represent yourself truthfully and without deflection or deceit. Passing off chatbot ramblings as writing you want others to spend time absorbing degrades you and your relationships.
If emails take too much of your time, write shorter emails. People will like you for it. If you have doubts about what to write, just be human; that goes a lot further than trying to trick people into reading a chatbot's scribbles.