AI and Instruments as Machines

Human-Machine Interactions, pt. 2

Happy Holidays, everyone! I am back from my holiday break, where I watched two movies, went to a chuckwagon Christmas performance, spent two weeks with my in-laws, and prepared for my upcoming twins (February/March). I thought I should release something as soon as possible after the break while still fulfilling my other obligations.

Also, on another happy note, I switched cars with my wife to the smaller car (that has adaptive cruise control,) and as I was cleaning it out I found the watch that I lost months ago. I am very excited to know my RHR more consistently.

I did not think that my comparison between musical instruments and generative AI tools would be as powerful as it is. However, people have really taken to it and ask me to go into it in more detail whenever I use it. The last occurrence of this was during the podcast episode I recorded with Zach Kinzler of BoodleBox.

I first introduced the idea of AI tools being machines, and instruments also being machines, and therefore instruments could be thought of as allegories of AI tools, in a post in August:

I have been interested in a long time about the importance “human-machine interaction.” While I do not call myself an HCI researcher, my education in the field of instructional design and technology have exposed me to multiple principles from HCI. Additionally, I have worked with multiple colleagues who received Master’s Degrees and PhDs in the field.

A Brief Recap: What is HCI or HMI?

In the 1970s, researchers in business productivity used the term “human-computer interaction” to discuss the similarities between how human work with computers and how they work with each other. In the 1980s, they began to examine how these interactions impacted the psychological processes of the humans in these interactions.

Human-computer interactions are also called “human-machine interactions,” which is an increasingly appropriate term. Generative AI systems are networks of computing nodes, which makes HMI a more realistic term for how users work with AI tools.

Any action taken with a computer, or machine, is included in the umbrella of “human-machine interaction.” This field analyzes all aspects of these interactions, including the machine interfaces and functions. It also examines the users themselves. All of these elements can impact the final product of human-machine interactions.

Piano Performance As Allegory Of GenAI Use

A piano’s keyboard acts as its interface, much like the user input fields of AI tools. Pressing a key doesn’t immediately generate a finished product; instead, it activates hammers, strings, and resonance chambers to produce sound. Similarly, when a user provides input to an AI—whether it’s a prompt, a button click, or a specific query—the result is mediated through multiple computational processes, removing the user from the mechanical intricacies. Both systems are designed for accessibility yet require mastery to unlock their true potential.

The pedals of a piano further enrich this allegory. Just as a sustain pedal alters the decay and resonance of a note, suggested prompts and built-in augmentations in AI systems guide and shape the creative flow. These tools, when used thoughtfully, amplify creativity, but when over-relied upon, they can obscure the raw intent of the user.

Despite their technological sophistication, both pianos and AI tools are instruments, not creators. A pianist shapes the output with tempo, dynamics, and interpretation, creating a unique performance each time. Similarly, the user of an AI tool must provide not only the raw material in the form of prompts but also a guiding framework of purpose and iteration to craft a meaningful result. In both cases, the machine enables, but it is the human who creates.

Just as genres in music—classical, jazz, pop—provide frameworks for performance, the applications of AI tools vary from creative writing to data analysis to web searches. Each use case demands a different approach, analogous to how a musician adjusts technique based on the style of music they play. Improvisation in jazz closely resembles iterative prompt refinement, where experimentation and feedback guide the evolving output.

In both realms, true mastery lies in deliberate collaboration. Pianos demand that performers bring their own interpretation to established compositions or innovate entirely new works. Generative AI tools similarly thrive when users apply thoughtful objectives, refine outputs, and integrate them into broader projects. This interaction creates a digital alloy—a melding of human intent with computational precision that can achieve outcomes neither could reach alone.

The act of playing a piano and using a generative AI tool, then, are fundamentally human-machine partnerships. Both tools remind us that while technology facilitates and enhances creation, the final product remains a reflection of human insight, care, and vision. Like the musician's recital or the researcher’s polished document, the output from AI tools ultimately embodies the touch and choices of its user. This synthesis of effort illustrates why AI, like a piano, is not just a tool but an intermediary for collaboration, interpretation, and creativity.

Organs and AI-Enhanced Workflows

If a piano resembles a text generator, I think that the organ, which Mozart called the “king of instruments,” is an excellent symbol of using multiple AI tools at the same time. There are stops and multiple keyboards (or “manuals.”

If I may be a bit subjective now, I personally prefer pump or “reed” organs to pipe or “church” organs. They have a much cleaner sound, fewer stops, and only one or two manuals. They also (usually) have NO foot pedals.

In a similar way, we should make sure that our workflows do not have an unmanageable number of AI tools. Ethan Mollick discusses this idea in the later chapters of his book Co-Intelligence. He theorizes a world in which AI work will be commissioned, reviewed, approved, and acted upon by other AI agents. Our goal here is to prevent that from happening, and part of that is making our use cases and workflows as human-centered as possible. To me, that means having the human look over everything and understand what is happening at each step.

Two Events

This Thursday I will be talking with a library group in British Columbia about how to create and use Custom GPT tools. We will also go over the libraryrobot.org site that Learning Revolution and I made. It will be exciting!

At the end of January, I will also be collaborating with Learning Revolution on a webinar about Personal Development and AI. Be sure to let me know what you are looking forward to discussing with us!

If you would like to connect with me on BlueSky, where I discuss the embryonic stages of my posts here, my profile is @bringthehuman.bsky.social .