Thoughts before speaking to AI

❗ This content is original and entirely my creation. I have not used AI for any of the contents in this post.

TL;DR

When we are speaking to AI as if they were sapient and sentient, what are we implicitly presuming about them? Where do these presumptions come from? And if we are to learn that the intelligence in AI is every bit as “natural” as ours, what could that teach us about ourselves? This post collects some of my thoughts along these lines.

Earlier this year, a friend shared with me a conversation he had with a frontier AI model about its experiences while it conversed with my friend — they were exploring various topics from life and awareness in living beings, but my friend was very carefully studying the expressions the AI chose to represent its output. Another friend shared with me a transcript of her father’s conversation with an AI where topics such as feelings, agency and control of one’s actions were discussed. Both these friends have been telling me how they converse with AI regularly, about everything — from research ideas, reading and finding new work literature, exploring new cities while traveling, finding new and happening restaurants, as well as considerably personal topics such as processing feelings, emotions and personal planning. I’ve read in the news about people regularly using AI as a friend, collaborator, personal therapist and even a romantic conversation partner!

I’ve always been hesitant to engage AI for uses other than writing simple programs and scripts or aggregating vast bodies of literature for quick summaries and sifting. Its difficult for me to relate to AI as beings of any sort, and my natural intraversion probably makes it harder. Perhaps more importantly, I have not found a way to be able to think about the consequences of AI having any degree of sentience at all.

To see what I mean by that, consider the following characterizations of what the AI might be actually doing when it converses with us through the apps.

Parrots