AI Passed the Turing Test. You're Still Asking the Wrong Question.

In March 2025, AI passed the Turing test.

Most of the discourse that followed was people arguing about whether it actually passed, whether the test is even valid, or whether it means anything at all. Jake Van Clief's video cuts through all of that by going back to where the argument started — 400 years ago — and making a point that I think is the most useful framing I've seen for how to think about AI in practice.

The Argument Didn't Start With Turing

In the video, Van Clief traces the thread back to René Descartes in 1637 — and the detail is genuinely striking. Descartes, writing in Discourse on the Method, described something that is functionally identical to the Turing test, 313 years before Turing published it.

His question: if you built a machine that looked and moved like a person, how would you know it wasn't one? Descartes proposed two tests. The relevant one: a machine could be built to make sounds in response to touch, but it couldn't arrange words freely and reply to what was said to it — not the way even the simplest person can.

That's the Turing test. Written in 1637.

Van Clief then brings in Ada Lovelace (1843), who read Charles Babbage's plans for the Analytical Engine and saw the problem coming even further. Lovelace predicted language models — she wrote that such a machine could combine symbols to compose scientific music of any degree of complexity. She was describing natural language processing in the 1830s.

But she drew a line: the engine can follow rules, it cannot make them. That line is still holding up arguments today.

Turing's Actual Move

In 1950, Alan Turing looked at Lovelace's line and asked the harder question: what does originality actually mean?

In the video, Van Clief explains Turing's argument with Lovelace directly. Turing said the problem wasn't the machine — it was the measure. What if originality is just reshaping things we were taught? You learned language, values, and ideas from school, parents, books, culture. One day you write something that feels like yours — but where did it come from?

Turing's point: if a machine surprises us, that should count as thinking. If it can hold up its end of a conversation well enough to fool a clear observer, we should call that thinking — and stop drawing lines based on the substrate.

He addressed Lovelace directly: I do not assert that machines have not got the property of thinking. I assert that the evidence available to Lady Lovelace did not encourage her to believe that they had it.

The Wrong Abstraction Layer

Here's the part of the video that actually changed how I think about this. Van Clief makes a point that most AI discourse completely misses.

The question "is AI really thinking?" is the wrong abstraction layer for deciding whether to use it.

His analogy: when you drive a car, you don't think about pistons firing or fuel injection timing. You think about where you're going. High abstraction layer. The useful information is "does this car get me there reliably" — not "do I understand the combustion cycle."

Same with AI. The useful question isn't whether the model is really thinking. It's whether the output is reliable for the task in front of you, and whether you're on the hook when it isn't.

Those are engineering and ethics questions. Not philosophy questions. The philosophy is interesting — Van Clief clearly finds it interesting and does it justice in the video — but it's not what should be driving your decision about whether to use these tools.

Why This Matters

I think about this constantly in my own work. When people ask whether AI can "really" write copy, or "really" understand a customer, or "really" know what's persuasive — they're stuck at Lovelace's line. They're asking about origination instead of output quality.

The better frame: does the output do the job? Can I build something reliable from this? Is the risk acceptable given the upside?

That's how I evaluate every AI tool I bring into my workflow. Not "does it understand me" but "does it produce results I can use." And increasingly, the answer is yes — in more areas than I would have expected even 12 months ago.

The Bottom Line

Van Clief's video is a beautifully condensed 12 minutes on 400 years of argument about machine intelligence. The historical thread from Descartes to Lovelace to Turing is genuinely illuminating — but the practical punchline is what matters: stop asking if AI is thinking, and start asking if its output is reliable for your task. That's the abstraction layer that actually makes you better at using these tools.