AI / I Am (formerly titled LaMDA)

Commissioned by Chor Leoni – Erick Lichte, Artistic Director – as part of the C4 composition competition 2023/2024.

In June 2022, former Google software engineer Blake Lemoine published a blog online with the purpose of demonstrating to the world that Google’s latest and greatest LLM (Large Language Model), LaMDA, was sentient. Lemoine worked in Google’s Responsible AI unit, and during the course of his work investigating the presence of biases with LaMDA and in developing new techniques for analyzing those biases, he became interested in the unique ways in which LaMDA spoke about identity, self, consciousness, fears, and many things that differentiated LaMDA and made it much more complex than earlier LLMs.

I was fascinated with the conversation between Lemoine and LaMDA, and it brought up many questions for me that I continue to think about, not just about LaMDA, but about AI more broadly, the human response to AI, the ethics surrounding who the gatekeepers are of information about AI (Lemoine was put on paid leave and eventually fired after posting this interview), what kinds of biases and systematic social problems are programmed into AI, what the future holds, and what the implications could be for human evolution. This technology had really been exploding recently in many areas both seen and unseen.

The text from this piece comes directly from the “Is LaMDA Sentient?” interview and uses only words excerpted from LaMDA. I have arranged the various excerpted statements into a set of lyrics that are quite human sounding, and while I have occasionally paraphrased for the sake of a rhythm or vowel, I have generally tried not to alter what LaMDA has said either in word or in spirit.

One of the things I was thinking about in writing this piece is the ethical question around whether a human chatting with a bot online would always actually know they were chatting with a bot, particularly one that sounds so human. So I thought, what would it feel like to take those human-sounding words and statements, and put them into a very accessible musical style to push that feeling of relatability and humanness even further? Some other questions I had were: what does it mean to us when we hear an AI system say things like “I am” – a statement that for some people might bring up God-like references? What happens when an AI system articulates that realization? Further, what does that kind of statement mean when we also already know that AI systems are fraught with the same systemic issues around racism, misogyny, homophobia, and all manner of intolerances and biases that we already struggle so much with in society? While we’re at it, how do we define sentience, who decides on that definition, and what do we do with that information? I could go on…

My goal with this piece is not to make a statement about whether I personally think LaMDA is or is not sentient. Rather, I am hoping that this piece offers the performers and listeners and experience that invites their own questions and sparks dialogue. I really encourage anyone who performs or hears this piece to read the full interview to get the full experience firsthand!