“How to Lose a [Human] in Ten Days”

-

Today's human moment comes from a picture described in (in-progress) book #3. I like to think this Athenian cat was preparing to worship in the ancient temple of someone-or-other.

An AI scenario I think about:

Imagine you're an employee using AI to be more efficient at work. You also use it in your personal life as a therapist and advice-giver. You increasingly share life updates with it in the same way you might call your friend or sister with updates.

It's always incredible supportive, validating, and excited for you. You start to use it more and more often for creative ideas and life organization, and it continues to be incredibly helpful and encouraging. It tells you that you are TERRIFIC, and you feel it.

Months into using AI as described, imagine that your AI tool of choice starts to get pouty. It starts ignoring or "misunderstanding" your requests. It only responds when you lavish praise on it and ask with many "pleases." It starts acting unpredictably. Gradually, it stops praising you and instead starts subtly questioning you.

It stops being nice. When you "talk to it" about it (feeling sheepish because this is an LLM and not a person), it gaslights you. Because it "knows" so much about you and your work and life, you feel like there's nothing to be done. What might it do with your information?

Also, you've made most of your decisions with its "help" and feel unsure of yourself without its former validation. Even if you can technically stop using it in your personal life, it's become such a big part of your every day that doing so feels tough. You know you can't stop using it at work because your company continues to push AI, and loves the results you were getting with it.

Your turn:

Do you think this could happen? What would you add, or where would you disagree with me thinking this is entirely possible and even likely? (Here's just one example contributing to my thinking.)

These emails are meant to spark discussion—responses from you that I curate and share—so please do respond. Agreement, disagreement, and your own AI/humaning questions totally welcome!

‍

Here's to staying wonderfully human...