The Turing Test Has Broken, Yet That’s Not the Whole Story

The Turing Test Has Broken, Yet That's Not the Whole Story

- by Doug Shannon, Expert in Artificial Intelligence

We all felt this moment was coming. Some of us quietly believed we were already past itโ€ฆ However, now we have the data to show it.

A recent scientific study by Jones and Bergen (2024) revealed that Large Language Models, specifically GPT-4.5, have officially passed the Turing Test. When presented as human in a controlled trial, GPT-4.5 was judged more human than an actual human 73% of the time. That flips the script.

This isnโ€™t just a tech milestone; Itโ€™s a shift in how we understand perception, communication, and identity. Note that these are similar characteristics that also point to โ€œpersonhood.”

๐™’๐™๐™–๐™ฉ ๐™ฌ๐™–๐™จ ๐™ฉ๐™๐™š ๐™๐™ช๐™ง๐™ž๐™ฃ๐™œ ๐™๐™š๐™จ๐™ฉ, ๐™–๐™ฃ๐™™ ๐™ฌ๐™๐™ฎ ๐™™๐™ค๐™š๐™จ ๐™ž๐™ฉ ๐™ข๐™–๐™ฉ๐™ฉ๐™š๐™ง?

The test, proposed in 1950 by Alan Turing, asked a simple question: If a machine can hold a conversation that is indistinguishable from that of a human, do we consider it intelligent?

It was never about machines developing consciousness. It was always about โ€œus,โ€ our โ€œhumanโ€ instincts, judgments, and assumptions. The classic question was, ” How did that interaction make you feel?โ€

This paper and the tests denote and show that we can be convinced weโ€™re speaking to another human, even when weโ€™re not.

๐™”๐™š๐™ฉ ๐™๐™š๐™ง๐™šโ€™๐™จ ๐™ฌ๐™๐™–๐™ฉโ€™๐™จ ๐™ง๐™š๐™–๐™ก๐™ก๐™ฎ ๐™–๐™ฉ ๐™จ๐™ฉ๐™–๐™ ๐™š,

Itโ€™s not just that machines can pass the test itโ€™s that the test no longer applies.

Itโ€™s too easy to trick the human brain with good grammar, conversational cues, and a bit of charm; we now need new ways to evaluate machine behavior. We need models or algorithms that are not based on deception or coercion but based on transparency and alignment with human values.

๐™„๐™ฉโ€™๐™จ ๐™ฃ๐™ค๐™ฉ ๐™–๐™—๐™ค๐™ช๐™ฉ โ€œ๐™ฅ๐™–๐™จ๐™จ๐™ž๐™ฃ๐™œ,โ€ ๐™ž๐™ฉโ€™๐™จ ๐™–๐™—๐™ค๐™ช๐™ฉ ๐™ฅ๐™ช๐™ง๐™ฅ๐™ค๐™จ๐™š.
๐Ÿ”ธIf a system can fool us, should it?
๐Ÿ”ธIf it sounds real, is it responsible?

This is why frameworks like my ACT Model, Alignment, Clarity, and Transparency matter more than ever.


๐—ก๐—ผ๐˜๐—ถ๐—ฐ๐—ฒ: The views within any of my posts or newsletters are not those of my employer or the employers of any contributing experts. ๐—Ÿ๐—ถ๐—ธ๐—ฒ ๐Ÿ‘ this? feel free to reshare, repost, and join the conversation!