Virtually yours

Talking with other attendees at Wired magazine’s annual Big Interview event last fall, between events and at lunch, I was struck by the handful of folks who were most excited to tell me about their AI companions, therapists, and advisors in startlingly intimate detail. They consistently told me how meaningful their relationship was with a virtual, software-generated facsimile of a person, how it had improved their outlook on life. Their sincere appreciation and affection were evident in the tone and tenor of their comments: in their words and nonverbally, through their tone of voice, facial expressions, postures, and gestures, conveying the kind of admiration and adoration reserved for the closest human affiliations and alliances.

Like you, I’d read about this phenomenon, but this being the central topic to almost all of the more than a dozen and a half conversations I had with people of different ages, races, and genders from around the US alerted me to how widespread this was, at least among this technologically oriented slice of the population.

Whatever you think of humans forming what they consider meaningful relationships with a virtual entity, the phenomenon is only growing as AI becomes incorporated into more and more applications and programs. The convenience and efficiency of AI are convincing more and more of us to adopt its use.

I’m in that group, for sure. Though I have no one-on-one “relationship” with any virtual pseudo-being, the program I deploy regularly to check grammar and spelling incorporates artificial intelligence programming. Though I don’t use it to write my blog — I tried a couple of times to no avail, but that’s another story — I do use another of these apps to edit and polish my posts.

What's concerning is how we think about, we feel toward, and we refer to these disembodied phenomena. For instance, writing about a project he’s working on that incorporates the use of and depends entirely on the functionality of an AI program just the other day, a close friend kept referring to the software as ‘he.’

I can’t help but wonder how unconsciously and pervasively anthropomorphizing AI influences our understanding of technology and our use of and relationship with it.

The simple pronoun — 'he' — points to a deeper linguistic and conceptual problem. When we refer to artificial intelligence programs and the kind of virtual creatures they create, most of us say ‘virtual’ as if it were another classification for considering the types of living organisms that have somehow expanded beyond animals, plants, fungi, and protista. We talk and think about these epiphenomena as if they were something real and tangible, as if they truly exist.

But they don’t exist. They can’t. They’re virtual. That's the opposite of real, remember?

Certainly, even though it can abstractly mimic the functionality of abilities, software isn’t a living organism. Yet, linguistically, we refer to ‘it’ as if it were alive; we attribute human characteristics to it, much like we do to our pets or, at least for some, plants. How do we pull this weird feat of modern-day sci-fi magic off? What kind of twisted special effect is this?

(Whether you have been wondering how we got here or not, thank you for staying with me so far.)

For the longest time, the thing happening, the action that's coming about, couldn't be separated from the object or organism bringing it about. Your dog begs for food. The camera takes pictures. An abacus calculates.

The person-facsimile that software generates is non-existent; it is not an actual thing — you can’t put it in an envelope to send to someone, it has no weight, and doesn’t take up any space. This simulacrum ‘exists’ solely as the result of lines of code, of instructions that create specific output: text, sound, and images. No matter how sophisticated the software may be or become, these disembodied processes remain dissociated from the mechanisms that deliver the results of their computations. Though it seems we can’t help but infer that there is a ‘them’ there, there is no ghost in the machine.

Our instinctive anthropomorphizing is not just a quirky habit; it’s a conceptual risk. When we assign a human pronoun, such as 'he', to an application, we are unconsciously assigning agency and intent. We risk mistaking the AI's optimized pattern-matching for wisdom, or its calculated empathy for care. If we learn to rely on AI 'advisors' and 'therapists' as collaborators—as Zuckerberg suggests—without constant awareness that they are merely sophisticated mirrors of data, we open ourselves up to manipulation, misplaced emotional investment, and a dangerous erosion of critical thinking about who (or what) truly holds our trust.

The risk and danger it implies are misplacing emotional connection and trust in a non-sentient system. No matter how we feel, no matter how engaged, grateful, or entangled we become, how can we remind ourselves that there is no one on the other end of the conversation? Is it enough to avoid using pronouns that give the non-object, that-which-we-are-referring-to, agency and an independent existence?

Is it enough to refer to Artificial Intelligence programs and apps as ‘it?’ Or is this turn of phrase insufficient and too reductive? I wonder if we might need some kind of linguistic category to remind us every time we talk about or to a virtual non-entity, to such a no-thing, that it is just a bunch of electrons dancing nowhere.

Making this linguistic shift could help us maintain a clear understanding of the nature of AI. I find neologisms like ‘The Process,’ ‘The Generator,’ or a ‘Model-Agent’ way too clumsy to use in everyday conversation. What kind of terminology could we adopt or create to serve as a persistent linguistic reminder that we are addressing a sophisticated function, not a friend?

I took the photo at the top of today’s post in Florence, Italy, a few weeks ago. I used software to resize the picture, but I didn’t alter the image in any other way. The graffiti appears near the Campo di Martie train station, across the street from Caffetteria Emmeti, purveyor of delicious, inexpensive specialty coffee.

As part of the editing process and response to my prompts, Gemini — Google’s AI program — suggested adding the fourth-to-last and pivotal paragraph to the post. In an ironic turn, I included it in its entirety, unedited.