Like others have said already, bots could likely learn to play those easily ... but I'm more concerned about people with disabilities / illnesses that would make playing these games hard, painful or even impossible. Someone who has parkinsons or arthritis for example might be able to click a big square in an image to solve a captcha, but might have trouble to "fine-tune" their movements fast enough to play a minigame that effectively locks them out of the community if they fail, especially if there is a timer involved.
There are two issue posts on the Lemmy github about the captcha options they considered. It is an interesting read. I had no idea there were so many types or even the existence of embedded options. I thought all were 3rd party and most were google, but I was wrong. Still, there are recent Lemmy posts by the devs basically saying the only option to effectively control the bots is by requiring a valid email for account creation.
With AI capabilities now, surely it’s pretty easy for an AI to follow a set of instructions like: create an email, check email, click link in email…etc - is that correct? Or put another way - why would email verification stump ML so consistently if it’s trained to create emails and do the process
Training an AI to play snake or other simple games is not hard. Making it stop at a specific score might make it slightly harder, but not much. Then you just need to read the text from the screen either, which is trivial. No, not hard for a bots to get past. It might slow actual humans more than bots.
It's definitely trivial for an AI to solve the "game" or task, I think an interesting question would be whether you could filter them by checking how efficiently they do so.
I'm thinking something like giving two consecutive math tasks, first you give e.g. 1+1, then you give something like 11 + 7. While probably all people would spend a small, but detectable, longer amount of time on the "harder" problem, an AI would have to be trained on "what do humans perceive as the harder problem" in order to be undetectable. That is, even training the AI to have a "human like" delay in responding isn't enough, you would have to train it to have a relatively longer delay on "harder" problems.
Another could be:
Sort the words (ajax, zebra) alphabetically
Sort the words (analogous, analogy) alphabetically
where the human would spend more time on the second. Do you think such an approach would be feasible, or is there a very good, immediate reason it isn't a common approach already?
With that idea, you (the captcha maker) would also have to write some code that computes how long humans should take to do a task (so that you can time the user and compare that with what your code spits out). Whatever code you write, the bot makers could eventually figure out what you wrote, and copy that.
To put it another way, when you say "humans would spend more time on the second task" with your two examples, you would have to write specific rules about how long humans would take, so that your captcha can enforce those rules. But then the bot makers could use trial and error to figure out what your rules were and then write code that waits exactly as long as you're expecting.