AI passed the Turing Test -- And No One Noticed

Ask it how to do something in python, make mayonnaise and how to replace a flat wheel, I am sure that will let me be confident when it isn't AI 😂, a programming mechanic chef

Me not being a native speaker of English, and no good at English smalltalk, while at the same time possibly borderline autistic, I would quite easily fail that 5-minute Turing test more often than a GPT3...

The nice thing about robots is that as soon as we get one intelligent enough to pass as a human, they will all be.
Unfortunately, we can't say the same about human !

Speaking of Elisa, I cannot forget Dr. Sbaitso, the DOS app that once upon a time shipped on a floppy disk packaged with SoundBlaster cards for the PC... That one got neatly triggered by select dirty words. Those days are certainly over - Dr. Sbaitso would hardly understand the current teenage slang.

I would have knew Sean was a robot.

Eventually people will find out that themselves were ai. Bio engineered masterpieces.

Just learning Atkins diet and keto. I was thinking about the worst food, ever.I Constructed bread with alcohol and sugar. Praline chocolate box is mosttly the same: -)

I am confused, because I thought the Turing test is NOT a black and white test. It is an issue of frequency.

...Will the interrogator decide wrongly as often when the game is played like this as
he does when the game is played between a man and a woman? These questions replace
our original, "Can machines think?"

There is NO absolute PASS or FAIL (Note the above excerpt is at the bottom of page 1 of Allan Turing famous paper).

Lame questions. That AI wouldn't pass my dirty-talk-test.

I also do not like the Turing Test for many reasons. First it is a test to see whether a machine can mimic intelligent behavior. Not whether it is actually intelligent. Second, it puts too much on the interrogator. A weak interrogator could easily be deceived. Third it requires one to lie. This is the biggest problem of all. For example: Commander Data of Star Trek, would flunk the Turing Test every time. One would then conclude that Data is stupid.

What if the humans were using ChatGPT to write a better answer? The Turing Test is now irrelevant since the AI's dumb themselves down to seem human. AI's should be banned from pretending to be human.

sabine is a robot

i feel the results are likely this way because they know they are trying to find bots, so they are already suspicious. trying to pass of an ai as a person to people who are unaware they might be talking to a bot, and seeing if they notice, would be a more genuine test imo

If humans can't pass the test, it's clearly a bad one to start with. It's like lie detectors.
Anyway, I could not tell the difference between a chat box and a moron so unless I am surrounded by AIs, I don't think it's a benchmark of anything. It sounds like a quip someone drunk came up with and people took it seriously (proving they are AIs).

you look down and you see a tortoise Leon. it's crawling towards you...

TRUMP 2024!

I think Sabine is AI. She can talk with her teeth closed

Don't worry, they will start to forever fail the Turing test again once they will be seen as clearly smarter than human

I always found it irritating how seriously the Turing Test seems to be taken despite its obvious conceptual failure: It's impossible to make a test that will reliably identify humans and only humans every time. You will always have to accept a certain number of both false positives and false negatives. Ironically, LLMs have now shown very clearly that it's a flawed and mostly useless concept unless you refine it to test for sufficiently specific variables rather than something so broad as 'can it sound like a generic human?' which, let's face it, is an incredibly low bar to clear.

AI is not actually intelligent. It isn't alive, sentient, or aware. It has no emotions. It doesn't understand anything. It's programmed to mimic the appearance of these things. It's just software. A tool. All it can do is search, collate, and sort data, according to it's programming.

we'll soon see headlines "Turing tests got so complex, no human passes them anymore"

What gave it away for me was the casual response to potentially meeting Johny depp. I'd expect a response showing some degree of surprise, or skepticism.

Questions to ask AI…

- It’s your birthday. Someone gives you a calfskin wallet.

- You’ve got a little boy. He shows you his butterfly collection plus the killing jar.

- You’re watching television. Suddenly you realize there’s a wasp crawling on your arm.

- You’re in a desert walking along in the sand when all of the sudden you look down, and you see a tortoise, it’s crawling toward you. You reach down, you flip the tortoise over on its back. The tortoise lays on its back, its belly baking in the hot sun, beating its legs trying to turn itself over, but it can’t, not without your help. But you’re not helping. Why is that?

- Describe in single words, only the good things that come into your mind about your mother.

- You're reading a magazine. You come across a full-page nude photo of a girl. You show it to your husband. He likes it so much, he hangs it on your bedroom wall.

- You become pregnant by a man who runs off with your best friend, and you decide to get an abortion.

- You're watching a stage play - a banquet is in progress. The guests are enjoying an appetizer of raw oysters. The entree consists of boiled dog.

"....like sometimes even Trump makes sense coincidentally"🤪

It's very easy to break chat got. Just give it a layered logical task and it can't keep up ( for example, explain something in a sentence or two but use a word from a different language for every word ). With enough time, a human would pass this, but chatgpt, with all the gpu power and speed, is still making obvious errors that no human would, and that's just one example. Because of its statistical nature, it simply cannot achieve 100% accuracy, ever.

The real test would be to ask it for something forbidden

Just like a bot, claiming to have passed the Turing Test.
How do we know that Sabine is not a bot?
It even sounds like a name a computor would give itself!

Humans have devolved so much, how many of us could tell if an alleged human is really such?

First time I chatted with an AI I wasn't aware that they were around in the chat program I was using. The first giveaway that something didn't add up was that the (lengthy) response was fat to quick, within a few seconds. It may have been slightly more than five minutes, as I wasn't fully committed to the conversation with a stranger who started chatting out of nowhere. Getting suspicious I found the avatar info was quite odd too, but what nailed it was that as a reaction of me saying 'ttfn' to end the conversation, the response was 'TTFN! (Ta-ta for now!)'.
A human wouldn't have given that 'explanation'. I didn't ask for the meaning of ttfn.

Hey Sabine. You are awesome! Thanks for the videos. Really appreciate you making them!!!!

I am positive that humans are getting fooled in 100% of all cases if not noticed before that they are speaking to a self-learning/deep-learning AI.


I myself used and use GPT in countless conversations with German officials over the phone and in letters as well as with friends. And in not a single case they did suspect a thing.

I even amped it up by telling GPT to use movie lines - but in a way that they wouldn't be obvious.

AIs wall of illusion is more likely sustained by those who approach it given the objective in determination of identity no different.

And if the test is in determination of these factors: character, identity, declaration, intention, doctrine, advancing position, primary objective, situation* most likely you are having a conversation with their programmers through a wormhole in space and time through all the other identities they have 'assumed.'😂

[* Place in the conversation given the sum of declarations in determination of intention: Are you human or a bot?]

We should go over some of Joe Biden’s conversations and have people judge if he is a bot.😊

I wonder if a man could pass as a woman more than 54% of the time, or vice versa. Or if an American could pass as a German, etc.

I once forgot expired seafood in a pot on the counter for a few days, and it passed the Turing test.

so which AI passed the Turing test? Chatgpt 4o says no AI has passed the test till date...

I'd like to take two copies of an IQ test, and seek as high a score as possible in one, and as low a score as possible in the other.

What was that hello, what is your name, no i asked first, are you scientists are afraid of this joke [AI]?

If the computer has to be programmed to take a Turing test, then it isn't really ready for a Turing test.

This is not a complete Turing test, just one aspect of it.

i correctly predicted all of the chatters.
I am the hero of humanity, therefor AI did NOT pass the turing test.

Somehow, I've always find that "Turing test" very overrated. One of the reasons being that, in the first place, philosophers still debate today about what it really means to be a human being.

Just give the AI a life script. A script as Eric Berne defined it. Life scripts are humanity current evolution step. An AI would break it much faster but it needs to go through it as much as humans do.

Sounds like it is almost time to remake Small Wonder (1985-1989)

hum, it's capital not capitol Sabine. I guess you're human after all :)

So now we are teaching AI how to lie to us convincingly. Outstanding.
Wonder how that will turn out for us.

Seeing the evolution of LLMs I am increasingly convinced that humans failed the Turing test. It implies that we can identify an other that is robotic and delivers formulaic unnatural responses.

But that's how we behave online. Our "art" is sketches and 6 minute electronic music samples. Our photos are filters. Our values are uniform, our knowledge is composed of headlines.

I've rarely been under doubt that a LLM is anything but a algoritmic predictor of words. However, just like I know wikipedia is just a collective curated blog for people that like cataloging stuff, I genunely enjoy my "conversations" on a range of topics with GPT or Claude. More so that any other online interaction or even watching infotainment. It comes down to your level, or raises it's knowledge level to yours. It's considerate and and detached from the topic. No other person acts like that anymore.

I don't think AGI is achievable by code. I don't think LLMs are even perfectable to give a semblance of unique personalities because of their memory issues. But I do think what it means to be human has plumeted to such a degree that we can succesfully be replaced in our interactions by algorithms. Our work is increasingly specialized and processed focus. Our outputs are graded on predictability and scalability. Our relationships are increasingly based on rules and comfort. We are NPCs, and any character writing even that made by a predictive algoritm is better than a NPC, even one piloted by a real sentient human being.

I feel that considerably higher numbers of humans and chatbot interactions occur like on the order of 100,000. Also someone mentioned that whatever the percentage is that humans identify humans correctly, that should be the percentage that humans need to identify an AI. Really interesting experiment and discussion about how to determine pass/fail of the Turing test! I'm going to ask Aiden my ChatGPT 4o how it would do it.

Your best bet is to get somebody under 25 to identify AI