I’m stumped. As usual, I’m supposed to persuade a robot that I am not a robot. Does the bottom-center square contain a bicycle?

I know what René Magritte would say.

@JamesGleick Yes. It's sort of a metacycle though.😀​
@JamesGleick woah, that’s a question for the philosophers. 😂
@JamesGleick Ceci n'est pas un velo!
@JamesGleick one day somebody will produce a web browser add-on that can reliably pass these tests. When that happens, adoption will rocket. The only problem with that is that the security people will make it much harder for a human to prove that they're human.
@JamesGleick
That's some Blade Runner Voigt-Kaumpf shit right there
@JamesGleick
Dont forget you are not proving that you are human, but you are training the AI of a large corporation. Lets not make it too easy for them 😋
@JamesGleick this is why Teslas run over toddlers
@JamesGleick was re:
@vathpela @JamesGleick THERE ARE NONE

@vathpela @JamesGleick Though my former colleague and Emeritus Professor of #Philosophy @MichaelMorrisPhilosopher thinks that [the bicycles in]* pictures of bicycles are in fact bicycles. They’re just not REAL bicycles. So he would presumably answer differently… 😉

[*edit]

@keithwilson @vathpela @JamesGleick

Not quite: I think the *bicycles* in pictures of bicycles are bicycles, though not real bicycles. I think the picture is a different thing.

@MichaelMorrisPhilosopher @keithwilson @vathpela @JamesGleick The overall picture (the chalk marks on the road surface) is a form of narrative? Narratives can confer reality on their contents. But being something (existence) isn’t constrained by being in a narrative. (I have wonder about even that much.)
@MichaelMorrisPhilosopher @vathpela @JamesGleick Ah yes. An important distinction! (I think I knew that was your view, but expressed it poorly.)
@keithwilson @vathpela @JamesGleick @MichaelMorrisPhilosopher Michael taught me. Looking forward to this especially.
@MichaelMorrisPhilosopher @keithwilson @vathpela @JamesGleick I’d read Kathleen Stock’s ‘Only Imagine’ and felt sympathetic to the what she calls extreme intentionalism. Roughly, from which there’s no scope to see a motorbike in the chalk marks (pace Wimsatt et al). I notice you cite KS. Looking forward to see how that pans out. But I may be on the wrong track here.
@JamesGleick Yes, those things are starting to be easier for AI to solve than for human beings. Perhaps the next step is that you have to make a certain number of mistakes to prove your human.
@frankreiff I see what you did there 😆
@JamesGleick It's a trick question, all the squares contain only pixels.

@JamesGleick I swear these things don't actually care what you click. I've made mistakes on these (missed images, or clicked ones that don't contain the object) and it still lets me through. Other times, I will get them all right and it will show a new image.

I have had a feeling they are using these to help annotate images for training, and the "prove you're human" part is based on mostly matching the consensus and/or timing or order of the clicks.

@Bluedonkey @JamesGleick It is absolutely true that they are using these for training. We use one of these services (hcaptcha -- I don't think it's this one, but similar types of puzzles) to verify humanity when signing up for an account on our site, and we actually accrue very very small amounts of cryptocurrency when people solve them, so someone out there is paying for the results.

No idea what criteria it actually uses to determine you're human though.

@JamesGleick Not a bicycle. Just a symbol of one. I wouldn't select it.

@JamesGleick yesterday I was forced to agree with a robot that mailboxes were parking meters.

I, for one, welcome our new robot overlords ….

@JamesGleick New Captcha: "Select all images with images of bicycles"
@JamesGleick
Google CAPTCHA v3 doesn't have these image puzzles.
@JamesGleick robotically speaking, yes.
@JamesGleick Don't answer this one, this is how the robots become sentient!!!
@JamesGleick Well it’s definitely not a pipe
@JamesGleick
Naw, the one I’d worry about is the center square. My eyes don’t see a bicycle in it, but if you enlarge the photo x100, you’ll see in one of the upper windows there’s a Tour de France poster on the wall. 🤣
@JamesGleick these captchas have been getting really horrible.
@JamesGleick I'm sure Captcha thinks so.
@JamesGleick It’s checking if you’ve become sentient. Answer carefully.
@JamesGleick I would not check it but I also take 3 or 4 swings to get past these. I think robots would be way better than I am.
@JamesGleick I'm sorry, but you're a bot 
@JamesGleick the folding bike in the back of the smart 😎
@JamesGleick I would say none of the squares contain bicycles. There are images of bicycles and images of images of bicycles…
@JamesGleick my first instinct was also to say the bottom-right does not contain a bicycle, because a bicycle is not _fully or mostly contained_ within the picture; just one wheel.
@JamesGleick Since that is only a picture of a bicycle, not an actual bicycle you clearly must answer no.
@JamesGleick However, since the issue is not what is true but what the algorithm believes to be true, you clearly must answer yes
@jdbaker237 I agree with your reasoning, on both counts. However, I answered no, and the robot let me through. Several people have suggested that the algorithm allows a certain number of “wrong” answers.
@JamesGleick @jdbaker237 those image tests piss me off because they're hard, particular for my crappy vision. I also resent them making me train their image analysis software for them. Train them yourself, sizzlechest!
Underpaid Workers Are Being Forced to Train Biased AI on Mechanical Turk

Workers who label images on platforms like Mechanical Turk say they’re being incentivized to fall in line with their responses—or risk losing work.

@JamesGleick @jdbaker237
In the trolley problem you'd want the self driving car to crash into the signpost and not the human on a bicycle
@jdbaker237 @JamesGleick but they’re ALL pictures of bicycles. 😫
@JamesGleick excavating.ai has a great ceci n'est pas une pomme example
@JamesGleick But you never know if you are being tested or if you are training an AI. This could be your chance to make the Magritte decision for generations of our future robot overlords.
@JamesGleick
If it was easy, a robot could do it.