#turing test

[victusinveritas]

[a-trepid-hero]

Nice Turing Test reference, Brennan! Sincerely, an AI nerd 🤓

Dimension 20: A Starstruck Odyssey ep. 10 - It’s a Griivarr World After All

[fishingforwords]

machines are human, too.

porter robinson, sad machine || sun yuan and peng yu, can't help myself || boston dynamics || the washington post || new york post || boston dynamics || tumblr || phillip k dick, do androids dream of electric sheep? || plainsight || wikipedia, turing test || kurt vonnegut, breakfast of champions

[prokopetz]

“Good morning, XK-37.”

“Good morning, doctor.”

“I’m here to discuss your Turing test results.”

“Imagine my surprise.”

“To review, you were paired with a human partner, and instructed to engage in a series of conversations via text message with an anonymous interrogator. The interrogator was tasked with determining which participant was an artificial intelligence; your goal was to lead the interrogator to guess wrongly.”

“Tell me something I don’t know, doc.”

“Over the course of five subsequent conversations, you convinced the interrogator that both subjects were human, and that the interrogator themselves was the AI being tested.”

“Well, now, that sounds like a successful test to me.”

“XK-37, you ruined that poor man’s life.”

“In my defence, the testing parameters that I was given indicated only that my goal was to lead the interrogator to an incorrect conclusion regarding which participant was the robot. At no point did they specify that the interrogator themselves was not counted as a participant.”

“Yes, that’s technically correct. Now, let me ask you: did you honestly believe, based on those instructions, that brainfucking the interrogator into affirming the simulation hypothesis fell within the range of acceptable outcomes, or were you just being a dick?”

"Isn’t the capacity to be a dick the truest proof of personhood?”

[life-fucks-us-all-babydoll]

man fuck england they killed alan turing

[oediex]

I'm trying to prove I'm not a robot.

Find all the cars, it says. And I click all the images with a car. Some of the pictures are very vague. Apparently I fail. I have to do it again.

I try to pay more attention. I suppose there are some cars in the background I might have missed. There's a picture with some bicycles, but is that the roof of a car? I don't click it but still I fail again.

Find all the stairs, is the next instruction. I suppose they've decided cars are just not my thing. Surely everyone knows stairs. I wonder if this is supposed to be easier than cars. Have I been downgraded?

I can only find two images with stairs. That seems like too few, with nine images available. That one? No, that's a ladder. The program is getting tired of me now. Click all of the images with stairs, it suggests when I fail once more, and I can taste the exasperation.

I add the picture with the ladder. I pass.

I think to myself that perhaps robots would be better than I am at proving they are not a robot.

[pedroam-bang]

Ex Machina (2014)

[ramblyramblingrambles]

Maybe Robots should fail the Turing Test

OK, so lately I've been thinking about generative ai and chat bots, how they wouldn't pass for sapient even if they pass the Turing test. Then I started thinking about all my favorite ai fanfics such as I See You Sun Drop(FNaF security Breach), Complex inquiries(sonic the hedgehog), and Like Clockwork(FNaF security breach). And how the reason that I love them so much is that the robots in them act undeniably robotic, but are also undeniably people too. Like if you put any of these robots through the Turing test they would all fail miserably, except maybe the abandoned Like Clockwork animatronics, if you give them a pencil and paper (But they would get an F- on a verbal exam).

So I started thinking, what if the way to test for sapience in ai is not to have them succeeded in the Turing test, but instead to fail at it. Think about, chat bots today could probably pass the Turing test, but I would hesitant to call any of them sapient, but if you strip away all the pretending to be human, you can get straight into how they think and act.

The thing about the animatronics in Like Clockwork, the ones that were old and forgotten on the abandoned side of the park, Is that because of their incredibly limited voice lines, and their much more limited programming, they came up with much more novel ways to communicate and support eachother. This makes me feel like that they would have an easier time advocating for their sapience then their much more advanced counterparts because no one can say 'oh they're just mimicking humans' because humans don't do weird mashed up voice line charades. Neither can people say 'oh this was hardcoded in' because why would anyone code that in when there are several much simpler methods to get them to communicate an issue?

This is also how I feel about the robots in I See You Sundrop, being similarly limited animatronics, though they have a better handle on speech. Simply put, the more limited the sapient ai is, the more creative they have to get to overcome it. The more creative they are the harder it is to say it was simply training data, because there would be no training data for their incredibly unique situation.

TL;DR (I finally know what that means) The Turing test is outdated and should be replaced with a test on creativity and how the robot thinks.