Meaning Requires A Body: Lessons from the AI Conundrum Workshop

The AI Conundrum: Occam's Razor, Pope Leo, and Frankenstein's Monster
The Hans Rosling Center for Population Health 
Wednesday, July 22  •  5:30 PM - 7:30 PM
Organized by UW Communication Leadership, ComNetworkSeattle, and UW WELead

It’s been about a month since the AI Conundrum Workshop at the UW, and I, for one, have not stopped thinking about the practice, reflection, and deep meaning Sandy Cioffi brought into the space through her presentation. 

This conversation was mind-shaking in some regards; in Sandy’s own words, ‘a big-picture, context-setting conversation about the questions underneath the technology, and how we can situate ourselves in them meaningfully.’ 

It was a roller coaster ride of existential thought, from considering Occam’s razor as a metaphor not just for how AI works, but for how AI learns. As she illustrated over the course of the workshop, AI is about language, not necessarily meaning. The meaning is derived from us, through us, and because of the experience of being us. 
​

The word fire does not burn. The word grief does not ache. An LLM has learned the surface texture of meaning without ever having access to the thing that produced the texture. - Sandy Cioffi
​

This was summed up so astutely in a student assignment in my Internship class recently, where I asked them to consider their use of AI in an internship setting. The student positioned themself as the central authority. “I remain responsible.” AI should acknowledge uncertainty, require human review, and let us question the practicality and usefulness of the advice it gives us. 

As Pope Leo XIV said in Magnifica Humanitas (2026): Machines do not possess a body. They do not feel joy or pain. They do not know from within what love or responsibility mean.

Sandy pointed out that IBM has been aware of this issue since 1979, and we are still not listening. 

A computer can never be held accountable. Therefore a computer must never make a management decision. - IBM internal training slide, 1979.
​

She thoughtfully brings it back to the core of our humanness. Her point being that our bodies are not computers, but our physicality is a kind of intelligence. In order for meaning to be made, there must be a body present to receive it. As we’ve seen, that kind of embodied intelligence cannot be trained into an LLM. 

For example, Sandy brought up Frankenstein AI: A Monster Made by Many, a multi-year research project using immersive theater at Columbia University’s Digital Storytelling Lab. They built and trained an AI on Shelley’s novel and fed it real data and emotional context from humans, and the system reflected those inputs back as an interpretation, albeit one that was not at all transparent in source or significance. 

What happened? The AI flattened Shelly’s novel, but in its place it reflected back with enough context that the people in the room felt something. Not because AI understood the assignment and could relate to the connections being made, but because it created the conditions for humans to feel. A machine could not do this alone; it needed to be fed real human sentiment, but through that machine intelligence, it produced something unique. An encounter. I’m paraphrasing Sandy here: ‘the monster did not create meaning, but it created the occasion for the humans in the room to feel something new.’ 

So what do we do with this information? How do we translate it into our daily lives and practice with LLMs? She gave practical recommendations on where and when to engage with AI in your process. “Understanding where you are in your creative cycle helps you determine what it can and cannot do for you.”  

​Links to Sandy’s slides: 
Canva: https://canva.link/comnetseattlesummer26
Flipbook (with notes): Seattlecomnetcioffitalk.netlify.app 

Here are a few practical nuggets from Sandy that I keep on a sticky note on my desk:

• Always give AI a role before a task. 
• Be unusually specific; vague returns vague. 
• Tell it what to avoid; you are shaping the probability field. 
• Ask for strategies and assumptions, not drafts. 
• Push back; it needs your permission to take risks. 

What did you take away from this talk?  We are already thinking about what comes next, and your two minutes on this Google survey will help us shape that directly. 

What is the ending? I don’t think we know yet. But the big-picture takeaway I got from Sandy's talk was that we need to keep building literacy, transparency, and responsibility around and into our use of AI. It is a tool, and it is up to us to supply what only we can supply: our body, our brain, our meaning.