– Reflections on Artificial Intelligence – A Guide for Thinking Humans

This is the third 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.

Mitchell’s book is rich with the history of AI. It traces names, milestones, and turning points across decades of research and ambition. And as you move through that history, one name surfaces repeatedly: IBM. The company wasn’t just a contributor to the early AI movement; for a long time, it was the movement. Mitchell’s account makes it clear that IBM led the charge to build machines that could not only compute, but think.

That ambition reached its most visible moment in 2011.

The camera pans across a Jeopardy! game show stage. Watson sits at the center, powered by algorithms and confidence. To its left is Ken Jennings—human, witty, quick. Watson wins.

And in that moment, the old dream of building a mind felt newly alive.

When you watch the game in progress in the above video, you notice a few things. Watson doesn’t just answer questions. It parses puns, deciphers idioms, and stitches together fragments of history, language, and probability. It even makes conversational comments like “Let’s finish this category.” You can hear the audience laughing in the background. The machine isn’t just computing. It seems to be reasoning.

While reading Mitchell’s chapters on Watson, I was also enrolled in IBM’s Introduction to AI course on Coursera. Listening to the instructors, I found myself asking: where is IBM now? In a landscape now led by OpenAI, Google, and Anthropic, IBM seems like a distant presence. Not gone, but no longer central. A ghost, perhaps, of its former stature. Their voice, once the authority, is now an echo.

What IBM attempted with Watson was not just technologically impressive. It was conceptually brave.

They weren’t building a chatbot. They weren’t optimizing for virality or engagement. Rather they aimed to create a machine that could assist doctors, support scientists, reason through complex problems. It was an attempt to build a different kind of intelligence; one grounded in epistemology, not entertainment.

But such dreams, when turned into products too early, often run into trouble.

Watson struggled in hospitals. Diagnosing illness turned out to be far messier than parsing trivia. Medical language is dense with ambiguity, and the nuance of care, its emotional texture, its interpretive flexibility proved difficult to encode. The system faltered. The promise began to unravel.

Today, IBM watches as others dominate the conversation. ChatGPT writes your emails. Gemini summarizes your documents. Claude replies with thoughtful prose. These tools are fluent, responsive, and optimized to delight.

But what IBM tried was different. More fragile. More ambitious. Not a talking machine. A thinking one.

Watson wasn’t designed to charm. It didn’t mimic human tone or style. Instead, it emphasized structure, logic, and evidence. It aimed to be rigorous, not witty.

In that sense, Watson was a philosopher launched into a marketplace that had already fallen in love with poets. And we, the users and the culture around us, chose accordingly. We valued fluency over fidelity, eloquence over epistemic depth.

But not all ghosts are failures. Some are reminders. Some are blueprints.

Watson was a whisper of what AI could have been, and what it still might become, if we learn to prize understanding over performance, and if we can bear the awkward silence that sometimes accompanies real thought.