Everybody talks about how often “AI is wrong” but what about how often you and I are wrong? Or when something we are reading is wrong.

How about we give AI a chance to show us how wrong we are about something? Doesn’t that seem fair? (Wise, even!) I made a little button that will do just that.

I made it a little silly, but really – ought we not consider counter-arguments whenever we can? You can click this bookmarklet/button when you’re on any web page (that you trust) and it will copy that page’s content to your clipboard and open ChatGPT. Hit paste and a well-crafted prompt will be submitted along with the text of the article. Counter-evidence, with source links, will be the output.

If you highlight text on a page, that text will be what it considers. I did it with this LinkedIn post while writing it, for example, and read a number of reasons this is a bad idea.

But I think it’s pretty cool! I’m hoping it will enable me to be wrong less.

Grab it here: https://lnkd.in/gmZJRN-2

Caveats: Use at your own risk. You’ll have to think for yourself about whether you buy these counter arguments. And watch out for untrustworthy sites that might use prompt injection to sneak something malicious into what you submit to ChatGPT.