@darius having fun determining the most useless data to provide 💅🏻

@darius is ice cream a cake? the greatest thread in forum history,

@darius how is that even supposed to work as a captcha then?? it's like they were so focused on the "using free labor to train our ML" part that they forgot the original purpose was to block bots

@darius I got a similar one the other day that I failed because I was interpreting #9 as a "strawberry cake"

@darius i saw a tiktok where people were making pancake spaghetti, so its not entirely inaccurate to select that one

@darius Know how that goes. accessibility nightmares alone with those... it's nice to know that they are difficult for even those with sight.

@darius Wait. What? Why doesn't it matter what you select? I"m confused. We've been seeing a lot more CAPTCHAs lately (especially my husband) and it's been sooooo frustrating.

@darius Middle-left... pasta cake?

@darius Weird! I thought the usual CAPTCHA approach was to provide a random mix of known and unknown images, then use clever statistics and aggregation to get certainty about unknown labels. Sounds like this particular implementation is buggy.

@darius I got a similar captcha the other day. Distorted pictures of animals. Next time I see one I’ll do the same. You want me to accurately label training data, you can damn well pay me.

@darius I second that ugh... :(

@darius I guess that would help with the reverse image search problem. But it doesn't make me feel great about the strength of hCaptcha.

@darius screenshotting this for something i've been thinking about 😶 lol

@darius any time i'm presented with a captcha i always wonder if there are particular answers that would let me pass while deliberately fucking up whatever bullshit they're trying to train.

@darius wait, wasn't captcha always used for training?

I remember back when Google used words they were training their text recognition algorithms.

There used to be two words, one was generated and the other was a scan from a book or newspaper. They knew the solution to the generated one and wanted us to solve the scan for them.

I sometimes had fun trying to spot the scanned word and "solve" it with random text.

@darius Whoa, I think I got one of these back at the beginning of September(!). I even grabbed a screenshot of it because it felt so off to me. It also did _not_ care which ones I chose (so I picked the, uh, worst ones)

@darius Or maybe it does matter: https://www.commitstrip.com/en/2018/09/24/trolling-the-ai/

@darius
Recaptcha is smarter, it provides you with multiple known good and some training material.

@darius who is benefiting from the tuning of ML datasets using CAPTCHAs—Google?

@darius this is a new evolution that i haven't seen yet! these things have always been used to train ml in some way, even going back to more "virtuous" intentions when recaptcha was being used to fix ocr errors in gutenberg corpus scans. ...