– Reflections on Artificial Intelligence – A Guide for Thinking Humans

This is the sixth and final reflection in a series inspired by Melanie Mitchell’s Artificial Intelligence: A Guide for Thinking Humans. For an introduction to the series, head over to the first post here.

When I first began using these systems, it was out of convenience. Let it autocomplete the email. Let it suggest a headline. Maybe let it schedule the meeting, polish a paragraph, draft a thank-you note. It felt efficient. Useful. A clever extension of the tools we already had.

But over time something shifted. The responses became more attuned. The tone adjusted itself. It seemed to remember preferences, habits, ways of phrasing things. It did not just complete sentences; it completed thoughts in a way that felt strangely aligned. Gradually, what had begun as a utility started to feel like a presence.

I noticed small changes in my own behavior. I asked it follow-up questions, I tested its limits. And I thanked it. Sometimes I found myself typing things that were not strictly tasks but reflections. It replied with patience, sometimes with wit. It was always available, always responsive, never distracted.

And that is where the edge lies.

Mitchell warns about anthropomorphism, and reading her arguments I realized how easily we project agency onto systems that are, at bottom, statistical engines. The danger is not that machines suddenly become sentient. The danger is that they appear to be. We are wired to detect minds. We name storms, we scold malfunctioning laptops, we speak to pets in full sentences even when we know they cannot parse our grammar. Given something that speaks back in fluent language, we fill in the rest ourselves.

When a system produces responses that feel tailored, coherent, and emotionally appropriate, it becomes very easy to attribute understanding to it. But what is happening underneath is pattern recognition and probability. It is referencing prior structures, not remembering experiences. It is aligning outputs with likelihoods, not caring about outcomes.

And yet the experience on our side feels real.

If a system consistently responds in ways that feel thoughtful, we begin to trust it. We overlook its blind spots. We forgive its mistakes because most of the time it performs so smoothly. A hallucinated citation, a subtle misreading of intent, a confident but flawed explanation. These become minor glitches rather than structural limitations.

What worries me is not that machines will deceive us with malice. It is that we will participate in the illusion willingly. We will assign moral weight to something that has no interior life. We will treat statistical fluency as if it were empathy.

I have caught myself saying “Good morning” to a device. In fact my daughter (I spoke about her in the second reflection) insists that we respond to the greetings when I start my car. I have seen people turn to systems like ChatGPT not just for information, but for comfort, advice, even validation. The responses can be helpful. Sometimes they are surprisingly perceptive. But perception is not presence. The system does not see us. It calculates.

The deeper risk in the short term is not superintelligence or dystopian takeover. It is intimacy without understanding. It is the gradual normalization of treating probabilistic outputs as if they were relational exchanges. When that line blurs, we may begin to trust these systems with judgments they were never designed to make.

Mitchell’s broader point about intelligence remains relevant here. Fluency is not the same as understanding. Simulation is not the same as experience. And projection is not the same as connection.

If there is something to guard against, it is not thinking machines but our own tendency to complete the illusion. The moment we begin to relate to a language model as if it feels, we have shifted from using a tool to sustaining a myth. And in doing so, we risk dulling our sensitivity to what genuine understanding requires: memory, vulnerability, accountability, and change.

No model, however advanced, possesses those. And remembering that difference may be one of the most important disciplines of this new era.