CAPTCHAs

funnily (or not, depending on your perspective) they often include animals that in some cultures/regions could become food. Your example contains 2. Somehow I don't think they would accepted answers.

More like 3.

And people could eat pianos. Would they? Probably not… but it’s still a possibility. lol

Yup, I did consider that! Good remark.

Here's a new one.

You need to do three of these to pass. I wonder that this can still work, considering that for a while now models have been great at animal identification, but I guess that's why these are so abstract and malformed!

It looks like it's growing another head.. although I guess the main head is growing out of the shell anyway lol

Heh yeah, I hadn't paid that much attention to it yet.

I don't think I've seen this one before! I wonder if it's mostly just analysis of how the user solves the CAPTCHA? Because the ostensible task would be pretty easy for a bot, I think.

every night

too quick is my guess

Wow, that's annoying!

https://www.schneier.com/blog/archives/2023/08/bots-are-better-than-humans-at-solving-captchas.html

I don't agree with Bruce's blog post title, given the abstract and conclusions of the paper, though

Looks kind of interesting, although I don't really care for reading it completely at the moment.

Here's a weird one on GitHub.



I didn't understand what I was supposed to do at first, which makes sense, because the prompt was poorly translated (probably automatically). I didn't notice any mistakes except on the first one, though, and then I had to redo the entire thing. That's quite frustrating... anyway, it seems quite easy to train a model to do this. So I suppose this is one of those CAPTCHAs where the point is to study the input data and see if the reaction times, mouse movement etc, match human users.
To make matters worse, when I'd done it, I still couldn't proceed setting up 2FA, as it said it was unable to verify my CAPTCHA response...

I asked ChatGPT if I am a robot and it said maybe, so that's how I voted.

This one's also interesting, seen today and yesterday at one of the EMC voting sites:

Another one that seems pretty easy to me to automate. But I guess correctness of the response isn't of much importance to hCaptcha? I am curious to know the details of how it works, but I assume they don't want to share that, lest people exploit it.

Here's a new one.

I'm going to ask if there's a course at my university that includes CAPTCHAs as part of the course material, actually...

I just had one (can't recall on which captcha service, sorry) on which I made a mistake twice in a row, then passed the test. Peculiar! I guess it does make sense though, if you consider that AI models can probably easily perform most of these tests as well.

There not being one might be caused by this.

This one I could actually imagine being challenging for some models.

I wonder if many people who are not on a Minecraft server but do regularly use the web also have considerable experience with CAPTCHAs.