Anyone else noticed there’s never a captcha for “Select all images that contain pedestrians”? Seems significant somehow.
Finally explained why Teslas are hitting all pedestrians.
I feel like not many people know that captchas train ai to be used to recognise things and for cars.
Man I’ve been getting a lot of the “images with cross walk” lately.
First of all, why the fuck you care about classifying a cross walk in your dataset. It’s a damn computer - if it recognizes a hazard it shouldn’t take more than a couple of instructions to throw the brakes.
Second, why the hell are you telling me I didn’t select all the images with cross walks? Clearly I can tell and your software can’t.
Americans wouldn’t know what that was so they can’t ask us
I really hate these text formats. It messes with the default reading order humans have had for millenia.
Why is the punchline before the start of the joke?
Yes the microblogging quote function should have an option to appear at the top aswell as the at the bottom. Depending on context, top is more practical.
By this metric, I’m afraid I’m only partially human 😅
The most annoying part is when you know that you clicked on every applicable image/square, but enough people didn’t click on one that it now tells you that you’re the one in the wrong.
All I’m trying to do is save motorcyclists’ lives. I don’t care what anyone says, body extremities should be considered as part of the motorcycle. Last thing we need is self-driving cars clipping people’s arms because they don’t recognize it!
And then failing the test somehow and having to click all the squares with stairs
And then failing that one too because you take literally any measure to protect your privacy. And then the next one.
Is that why I get so many capchtas on firefox?
It definitely seems to be the case for me. VPNs and other privacy settings can trigger the aggressive captchas as well.
It’s fun to imagine a self-driving car somewhere basing its decisions on the results of those captchas. “Come on, someone tell me where the traffic light is, I really need to know, like, now, this is important, hurry up!”
There really is always a relevant XKCD.
I understand that it’s not even the pictures that matter so much as how you move your mouse to select.
Erratically selecting all images with the traffic lights. Otherwise they might think you are a robot.
Only humans can identify the machines that identify humans.
But what about the tiny corner of the traffic signal in the other box? What about pedestrian crossing lights? Do they count? Sometimes they do, sometimes they don’t. It’s infuriating.
I had “select all objects that are lighter than this” and the reference was a dog and I was supposed to pick tennis balls and not ships. This was new to me but I managed quite well
also sliding puzzle pieces into place
AI can solve logic/image captcha. So, yeah.
I hope so, we’ve been training it for years.
AI generates the challenge, AI can solve it. What are we even doing here?
We are the monkeys touching the thing on the screen to get a peanut or in this case a cookie.
