Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

What's funny is that the AI needs to be trained...yet it already knows which choices are correct and incorrect. So, is it really necessary?

Edit: This is a joke, I am joking.



I don't know how the image captchas work but the old-fashioned type-the-words captchas asked you one it knew the answer to and one it didn't. By giving unknown words to multiple users and finding a consensus they could move words from the unknown to known set.


If they do it the way that the OCR recaptcha works, it allows some new ones to go through and uses consensus to classify them.

So most of them will have already been classified and those are used to test your integrity (and verify you) but an occasional new one will be presented that won't count towards your verification and if enough people agree on it it'll be classified.


I suspect it knows the correct choices for some of the little pictures it shows you, but for others you are training it.


This is the natural question to ask. I finally decided that they probably compare your answers to several other humans to validate training data.


> Edit: This is a joke, I am joking.

The voters seem to have formed a consensus that it was not a joke, unfortunately, so your humor has failed the test.

(This was a meta-joke, and I too am joking.)




Consider applying for YC's Fall 2026 batch! Applications are open till July 27.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: