With the rise of AI-generated content in the last few years, it can feel like AI is everywhere you look. As the media landscape we know today becomes increasingly polluted with AI-generated content, it’s unavoidable, and sure, some of us aren’t phased by this, but perhaps maybe we should be…
What if the line between AI and reality was completely blurred? That may sound extreme or daunting, but it seems to be the path we are heading, or rather tumbling headfirst down. From political content to pop culture rumours, you may have already seen some of these fake videos without even realising. So, as AI permeates the internet and floods social media feeds, how can you protect yourself from all of the misleading information?
Well, here are some of the best ways to spot these AI-generated videos.
1 – Watermarks
Some of you may be familiar with the popular open AI video model Sora, popularised heavily by a TikTok trend involving fake videos of celebrity Jake Paul. If you’re not, Sora is a generative AI model that creates videos with a watermark that floats around the screen, similar to other AI models such as Veo. While this is one way you may be able to identify AI-generated content, it isn’t exactly foolproof.
In recent times, there have been various cases of watermark censoring on AI videos, whether that means covering up logos with a sticker, emoji, or completely blurring it out with other AI and or editing software. This is why it’s important to keep an eye out for weird blurring, warping, and suspiciously placed emojis in content you see day to day.
2 – Debunking deepfakes
Deepfakes may seem like something primarily made for the sake of humour but the truth is they aren’t always so light-hearted, and can actually be one of the most misleading forms of AI. Imagine seeing your favourite celebrity endorsing or doing something harmful, only to find out the video you saw was completely fake. To avoid that possibility, some of the most important things to look at when trying to determine if a video is AI or not are unnatural features such as:
Facial realism: unnatural smooth skin, unusual wrinkles, poor blending on features, abnormal hands, or contrast between age-identifying features such as skin or hair.
Eyes: unnatural blinking and eye movement.
Lighting and physics: inconsistent shadows and unrealistic glares, particularly on people wearing glasses.
Lip movement: unsynced or unnatural movement of people’s mouths.
3 – General flaws:
Finally, we have the general flaws that AI has yet to perfect. An AI-generated video can no doubt pose as extremely accurate and of high quality; however, it’s important to notice some of the more obvious indicators that give away the video’s authenticity. Think about camera angles. Ask yourself some simple questions about what you’re seeing: Who’s filming the video and why? Look out for technical mistakes such as distorted, strange, or incomplete-looking objects or body parts. Finally, rely on common sense to recognise scenes that may seem perfectly staged or suspicious, which are often a red flag.
All this being said, not all AI-generated content is created in bad taste; it’s still important to be able to identify when content is AI-generated, especially considering the speed at which AI is evolving. As the line between AI content and reality continues to blur, being a passive viewer just won’t cut it anymore. However, the good news is that awareness can be reclaimed as a form of power, and paying attention can protect you from a digital landscape flooded with AI content.
