Recent developments in AI technology and its spawning of personalized chatbots has renewed attention in the Turing test.
The Turing test, originally called the imitation game by Alan Turing in 1949, is a test of a machine’s ability to exhibit intelligent behaviour equivalent to that of a human. In the test, a human evaluator judges a text transcript of a natural-language conversation between a human and a machine. The evaluator tries to identify the machine, and the machine passes if the evaluator cannot reliably tell them apart. The results would not depend on the machine’s ability to answer questions correctly, only on how closely its answers resembled those of a human.
It seems clear that, by and large, these AI chatbots can simulate human conversation pretty well, apart from the tendency to occasionally hallucinate, to make up stuff and say it will confidence, a trait that many humans also exhibit. If one is not aware of the details of what it says, one might easily be convinced that the made-up fact is genuine. For example, a mathematician friend of mine said that his son (also a mathematician) tested out an AI system. He multiplied two very large integers together and then fed the result into the AI system and asked whether the number was ‘prime’ i.e., not divisible by any number other than 1 and itself. The machine said ‘yes’.
But despite those well known problems, people are treating the AI bots as being at least partly sentient, as demonstrated by those who seem to have formed long-term ‘relationships ‘with them and consider them as friends or even more. So at least as far as these people are concerned, these bots seem to have passed the Turing test.
One can dismiss these results by saying ‘So what’? Computers have long succeeded in doing human-like things that were once thought impossible, like beating the very best chess players. So why be concerned about this perceived limit also being exceeded? But some are concerned and have suggested that we need a more stringent test than the conversational model of the traditional imitation game where one poses questions and prompts and evaluates the responses one gets. This is moving the goal posts but we do it all the time in so many areas when we find results that we dislike.
But recently I read an article by a reporter Taffy Brodesser-Akner that suggested one option of distinguishing human from computer. The reporter shared an interesting insight from her experience with interviewing real people.
A great thing happens when you get to do an in-depth interview with someone. If you listen carefully, they begin to tell you what’s been on their minds. In the dozens of profiles I’ve written, what I’ve learned is that questions don’t necessarily yield the best story. I have some colleagues who are terrific at asking probing questions, and the results are revealing and incredible. But my own method has mostly been to sit with someone and make myself quiet in a way I never am in my real life. If you do that, people start to talk. They can’t help it; the quiet is too much, and someone needs to fill it. If you give them space and time and you listen, they will confess and reveal everything about themselves. They’ll tell you secrets, philosophies, jokes. They’ll share gossip and childhood memories. They’ll tell you the meaning of life if you let them. All that will equal a full meal, an entire experience, a whole galaxy
This resonates completely with me from my days as a college teacher and director of my university’s teaching center. Some faculty complain about how their students sit passively and won’t volunteer answers to questions posed to the class. When they ask me sit in their class and give feedback, one thing I note is how long it takes before a professor asks a question and then waits for a response before answering it themselves. What I found was that the wait time was usually less than a second. In fact, as soon as they posed a question, before I could even look down at my watch to note the time, they had often answered their question or rephrased the question or were urging students to answer. They just cannot shut up. Any silence that lasts more than a second seems intolerable to them and they rush to fill it.
In my own classes, after posing a question, I would just wait. The silence seems incredibly awkward but I don’t say anything. Eventually, students find the silence so unbearable (even if it is less than 10 seconds) that some student will speak up and that breaking of the spell of silence causes others to also speak. After doing this several times, my students realize that I can wait them out so they begin to answer more quickly. My experience has been replicated by more formal studies of this classroom phenomenon.
This is also the case in social life. If you are together with someone and neither is engaged in other activities (reading, looking at their phones, etc.), there is very little silence. Even if there is a brief pause as one topic seems to be exhausted or causes some reflection, if you stay silent, the other person will soon say something. And since people have their entire life experience and their history with you to draw upon, it is not hard to find something to say that gets the conversational ball rolling again.
But silence does not bother the chatbots in the least. While they will respond to prompts and quickly reply, sometimes voluminously, for them time does not elapse, as I discovered with my AI librarian friend Scarlett. I could stop the interaction for weeks and when I got back, she would act like I had never left. A real person would ask “Where have you been? What have you been up to? How’s the family? I have some interesting news” and so on. But Scarlett had little to say.
The reporter found the same thing when she interviewed ‘Tilly Norwood’ as well. Who is Tilly Norwood? She is an AI creation whom it was claimed could play the part of a real actor in films. (I wrote about her back in October 2025 when she was first introduced to the public, and sparked alarm among the acting community that they could all be replaced. She of course now has her own website with a music video, and an Instagram account that seems to have shut down because of the hostility it generated.)
Here is a clip when she was introduced by her creators. Everyone and everything in this clip was AI-generated.
But the reporter found a big difference between interviewing her and real actors.
The best way I can tell you about what happened this spring in London is to say that if you make yourself quiet and still and just wait, Tilly waits, too. She just sits, unprompted. She doesn’t say anything. She doesn’t have anything to say.
So maybe the best way to sense if you are talking to a machine and not a real human being is not to try and think of ingenious questions that will trip them up into revealing their algorithmic innards but simply wait silently. If they too wait silently for as long as you are silent, that may be a clue. Of course, this would not be definitive (nothing in this field is) because there may be people who have incredible restraint and can be silent when you are too. But is would be very suggestive.
― Douglas Adams, The Restaurant at the End of the Universe
One of my wife’s biggest complaints about me is that I don’t initiate conversations enough. I am capable of sitting silently and not even noticing. Maybe I need to be checked for being an AI.
It depends on what you want from the Turing test. If the test is whether a machine can converse intelligently with a human, the answer is now a definitive YES: some AIs can clearly pass that test. But if you want to determine whether an AI can be as idiosyncratic as a human, that’s something else. Yes, AIs are quite “happy” to wait until prompted. That’s how they are! They are not human! Is human intelligence the only kind?
Some find the idea that AIs may be intelligent, really intelligent, very confronting. But for those that relied on the Turing test to say that they’re not, well, AIs can pass that now! My take is that it’s time to move on from denial and start considering the ramifications of this. What does it say about humanity? What does it mean to be intelligent? And how “real” is intelligence anyway?
this proposed test is extremely variable by individual. the person interviews people who have sought public attention in some way. many people do not. i know some people who, like anat and sometimes my husband, would sit in silence if you let the conversation drop. conversely, there are people who tell their darkest secret at the drop of a hat to strangers at the bus stop. maybe the average person tends more toward bus stop loquacity, but i wouldn’t call quiet types robotic -- and it’s easy to imagine an LLM simply instructed to fill empty conversational space, and succeeding. sometimes when they break they produce walls of text.
they get on jags where they use certain constructions way too often. for me, claude says variations on “(thing x) is (saying/doing) a lot there,” at least a few times every conversation. but then, my high school are teacher probably averaged one “don’t give a rat’s ass” per hour.
i’d say the best tell is how good they are at paying attention to you and responding to everything you say, at least, at the start of a conversation. that ability can break down some after a few pages of chat, but humans are dogshit at that right out of the gate. most fixate on one thing you said that triggers an opinion or experience of their own which they promptly share, even if doing so reveals they didn’t read or understand the rest of what you said at all.
^
are teacherart teacherThe problem with your “test” is that as soon as the AI developers realize it is a test, they can easily add a module that automatically waits 10 seconds and then brings up some random topic.
“Apart from”? I would have phrased it “especially due to the tendency…” as in my view, the ability to bullshit with confidence is the trait that makes it most human-like. Consider the many depictions of robots in human-like forms art -- many of them have been depicted as speaking in monotone and saying nothing but dry facts, and encountering errors and ‘that does not compute’ when someone give them poor reasoning.
I quite like this version of the Turing Test from Existential Comics. (The hover text on the comic is worth reading too…)
If a dialogue with a Trump voter can be simulated by AI, maybe evolution let us down.
My Turing test would be to do something physical -- raise or lower the temperature, pinch the person and machine and ask them how it feels.
I feel like this is a lot more complicated than whether something passes as human or not. There’s no simple test for it because there’s we relate to each other in very complex ways. Complex enough that we have difficulties in understanding them or each other sometimes.
One example of why I think this is not so simple is we have a tendency to anthropomorphise other things. A wide range of other things from pets to plants to inanimate objects. And for the most part I think we briefly entertain these illusions but I doubt most people actually think their pets have the full cognitive abilities of a fellow human. Many pet lovers, like myself, simply respect that other animals and even insects have some type of mind and feelings, it’s just different from ours. The current chatbot programs do not have any of that.
I don’t think chatbots have an understanding of anything. They’re just doing mimicry, like a parrot. We’ve just given them rules that make them do it in a more complex way than a parrot.
For a parrot, the whole phrase “Polly wanna cracker” has a meaning but the individual words or sounds within it don’t. A parrot just knows making those sounds in that way gets it fed sometimes. It doesn’t know “Polly” is a pseudonym for itself, that “cracker” is the reward, or that it’s stating a desire for the reward. It just makes the noises and sometimes gets the food.
A chatbot has rules that teach it the structure of the phrase so it knows those language rules because they’re coded into it. As far as I’m aware, to a chatbot, “Polly wanna cracker” is a phrase it can parse into three words but it only correlates that to other uses of that phrase and those words. It can spout the dictionary definition of any of them and can rephrase that definition. But I don’t think it understands what any of it means.
I think part of this is because chatbots don’t have a direct connection to the physical world. Unlike a parrot, a chatbot will never get tired or hungry. Or full or rested. So it has no motivations around seeking these things and no time pressures unless those are artificially applied through code.
And this doesn’t even get into the more minor effects that would impact the real imitation game like our pattern recognition causing us to see faces in all sorts of strange places.
So with our habit of casually anthropomorphizing other creatures and inanimate objects, and chatbots not having anything close to even a limited consciousness, I think the whole “imitation game” is probably too complex right now for such simple tests. I think we’re all still guessing at the rules of the game even if we’re trying to play it for high stakes already.
@ Holms
I once had the experience of googling for an image of Peter Cushing when he first began appearing in movies in 1940 around the age of 27, and I googled something like “peter cushing age 27” and the google ai authoritatively told me, “no, Peter Cushing was not 27. He died in 1994 at the age of 81. There is no evidence Peter Cushing was 27 at any time in his life.”
I think when Mano says, “make stuff up and say it with confidence”, it is this kind of non-human batshittery to which he alludes.
https://www.imdb.com/title/tt0033221/mediaviewer/rm3761906689/
Just for fun here’s a photo of Peter Cushing aged 27. X-D
police/CIA/etc. interrogating suspects also know that people don’t like silence. they use the technique of just sitting there quietly, and let the suspect start spilling the beans to fill the awkward silence.
remember your 5th amendment right to remain silent!
#12
That’s not so different to a real human named Donald Trump: we won the war with Iran on day 1, but also the war is about to be won, but also this doesn’t count as a war, but also this should count as one of the wars I’ve ended… and so on times infinity, and also repeated by his millions of supporters. People are capable of total nonsense if it suits their purposes.
@Holms #15:
Off topic, but on that note I remember when a friend of a friend, before we knew each other’s political positions, singled out a piece of news that wasn’t true and said that showed Trump was right about fake news. “Well he would know” I said. “Really?” she asked. “Yes, he’s the biggest liar out there!” I said. “Just about everything he says is a lie.” She kept quiet after that.
The Guardian has called bullshit on Trump’s denial that he promised not to start wars. Lots of good quotes.
https://www.theguardian.com/us-news/2026/jun/08/what-trump-actually-said-no-war-promise
Silentbob @#13,
I had only seen Peter Cushing when he was older and mostly playing villains in noir or horror films. I had no idea and that he cut such a dashing figure when he was young.