Mastodon

You Don't Know Me! My latest GPT

Posted by Matt Birchler
— 2 min read
You Don't Know Me! My latest GPT

Sometimes you post something to social media and it escapes your friend group and goes some level of viral. This leads to tons of people who don't know you seeing something you posted. They likely have zero context over who you are or what you believe. They don't know if you're joking or are being serious. They just have these few words and they're going to make a judgement about you in a flash.

This happened to me this weekend and it wasn't fun. I had people suggesting I thought privacy wasn’t important and that I must not understand why a certain software feature existed. It was far (far!) from the worst anyone has experienced with a social media post, but it did bum me out. The killer thing for me was that reading the original post, it was clear to me what I meant, but I could also see how someone who didn't know me could see that single post and think I believe things I don't. No social post can include all context and will omit things for brevity, and strangers may see those omissions as intentional, and not just for the sake of brevity. And I'm sure I've done this to other people over the years…assume the worst of strangers with bad takes online, right?

My new GPT

It’s called I Don’t Know You! and you can provide it with the text (and images, if you’d like) for something you’re about to post to social media. It will then tell you what positive and negative reactions people might have to the post if all they knew about you was this one single post.

When I fed in the post above, this is what it returned:

Is it kinda generic? Sure, but the negative reactions listed there are pretty well in line with what people replied to me with, and I wonder if I would have clarified the post a bit more if I had thought about the negative reactions more.

Here's another one with a deliberately bad post about President Lincoln:

I've configured the GPT to look for blatant factual inaccuracies as well, so if you write something clearly false it will try to warn you.

Of course we should be considering these things whenever we write things online, but how often do we write a social post, think “this is a banger”, and send it off without much thought?

Still a proof-of-concept

And as with basically all of these GPTs, this would be better if it was built into Mastodon/Threads/Ivory/etc. so that if I wanted, I could get this sort of alert in the posting interface, and not have to remember to go somewhere else. Again, GPTs feel like generating proof-of-concepts of features that will one day be built into “regular” software.

But hey, if you're interesting in playing with this and you have ChatGPT Plus, you can check it out now.