Deepfake

The deepfake unit is probably the biggest thing we will cover all year due to the rapid advancement of AI. I used to fall for those black-and-white Ring camera clips where an animal walks up and rings the doorbell. They look like security footage, the grainy quality feels authentic, and the idea is just funny enough that you want it to be real. I watched them, shared them, and never questioned where they came from.

What I didn't understand then is that the format itself does a lot of the persuading. Low-quality video feels honest. We associate grainy footage with a real camera and a real moment, so we lower our guard. A fake that borrows that look gets a free pass from our brains.

Today, I'm much better at catching these. Working in social media on Instagram has trained my eye. I see an enormous amount of content every day, and over time I started noticing the small things that give a fake away: movement that looks slightly off, details that don't hold up on a second look, and accounts that post this kind of content again and again with no source. Repetition taught me what real footage usually looks like and what it doesn't.

But the class made me think about what that skill is worth. I can spot the doorbell animals now, and the technology will keep improving. The tells I rely on today may not work next year. The more lasting habit is the one from our confirmation bias discussion: ask whether I want a video to be true before I decide whether it is.

My biggest takeaway is that experience helps, but it isn't a permanent shield. Staying skeptical has to be a habit, not a one-time lesson.

AI use: I used Claude to help organize my notes and edit this post. The reflections are my own.