• by Steven Rosenbaum , Featured Contributor, Yesterday

What if artificial intelligence isn’t destroying truth, but exposing how little we ever agreed about what truth actually is? To answer that question, I went straight to the best source I could find.

Steve Fuller, Auguste Comte Professor of Social Epistemology at the University of Warwick in England, and author of “Post-Truth: Knowledge as a Power Game,” has spent decades challenging conventional ideas about knowledge, expertise, and intellectual authority.

In a recent Zoom conversation, he offered a provocative argument: AI may be forcing us to reconsider not only how we establish truth, but what we mean by it.

Consider AI hallucinations. We generally understand them as failures, instances in which a machine confidently produces information that isn’t factual. Fuller sees something more complicated, particularly when AI reconstructs someone’s ideas using words that person never actually spoke.

"There’s all this stuff that I’ve said out there, and you can imagine AI trying to answer a specific question, kind of mixing and matching half of one sentence and half of another, “Fuller said. “But to be honest with you, it still might end up producing something that effectively is the truth -- that basically is the sort of thing this person would say.”

His defense comes from his own discipline. Philosophers routinely speak for thinkers who have been dead for centuries.

“We’re always talking about what Plato said, and Aristotle said... and we’re always kind of representing what all these guys are saying... as if there are... interlocutors right now. We do have philosophers, I should say, especially in the analytic tradition... who will say they can actually say what Kant meant better than Kant can.” His conclusion: “AI may be just a very good analytic philosopher.”

That’s a startling proposition. For a journalist, quotation marks promise actual words. Fuller is asking whether a machine can invent those words yet faithfully represent the underlying ideas. I’m not convinced the distinction excuses a fabricated quotation, but it’s a challenge worth examining.

His next concern was what happens when we become so dependent on AI that we lose the ability to evaluate its answers.

"There’s been a substitution of skill set taking place whereby relying more on AI, we don’t feel we need to kind of do the traditional stuff,” he said.

“Back in the old days, when we had this thing called education, people were actually required to read books and stuff, even if most of the stuff in the book wasn’t relevant for answering the question on the exam. But now we live in a world where the only time people want to know anything is when they have to answer an exam question or something comparable to it,” he explained.

Fuller isn’t advocating a ban on AI. He uses it himself and requires his students to generate competing ChatGPT answers, then evaluate which is better. His point is that critical thinking requires enough independent knowledge to recognize when a machine is wrong.

But what about the companies creating these increasingly powerful systems? Surely they understand and control the technology they’re selling?

Fuller isn’t so sure.

"What they have control over is a market… They don’t necessarily have control over the products they’re producing,” he said.

“And that’s what makes it a bit different from an automobile... Because at least with a car, you know what a car can and can’t do, and you can regulate it in a pretty straightforward way... But with AI... because it’s a much more interactive thing and because the product itself is able to transform itself... we’re talking about a product that the producer cannot fully control."

Toward the end of our conversation, I asked Fuller the question we’d been circling for most of an hour: What is truth?

He distinguished individual facts from our larger understanding of reality. We can establish whether it’s raining outside. The harder question is how we assemble those facts into a shared picture of the world, particularly now that the arbiters we once turned to, science and religion among them, have lost their authority.

"The fact [is] that epistemic authority’s kind of disappeared, and now everybody’s putting forward their own standards, basically -- and depending on how many followers you have on social media, that becomes the standard of truth,” Fuller said.

“Post-truth is by no means a pluralist paradise, because you might say, okay, if we all live in our own worlds, then it’s going to be cool... But the point is, this is not the world... we still have the conflicts, just like we did back in the old days."

Fuller sees real intellectual possibilities in AI’s disruption of established authority. I’m less comfortable with his willingness to blur the boundary between interpretation and accuracy. But his larger argument is difficult to dismiss.

We're introducing machines capable of producing persuasive answers at extraordinary scale, just as our shared standards for deciding what's true are fracturing.

The trouble with truth isn't that we've stopped searching for it. It's that we increasingly disagree about how we'll know when we've found it.