By: Kevin  Wang

Images generated from artificial intelligence (AI) are quickly becoming more and more popular and are starting to become a huge issue in the online world. With AI constantly evolving and improving, these fake images get harder to discern from real photos. AI deepfakes may seem harmless, but they can be used to carry out scams, identity theft, and propaganda. However, there are still several methods one could use to identify if an image is real or AI generated.

Artificial intelligence used to be far from perfect, with obvious errors such as hands with 6 fingers or glasses with strange lenses, but AI has improved greatly and deepfakes are extremely hard to tell apart from real images. According to AP News, due to the improvement of AI, some widely shared advice on how to detect generated images no longer holds. Henry Ajder, a leading expert in AI and a founder of consulting firm Latent Space Advisory, said that most deepfake images have “an aesthetic sort of smoothing effect” that leads to skin “looking incredibly polished.” BBC Bitesize reported another method of finding AI images is to focus on the small details. AI images are created using the data from other images, so AI often struggles to create details that aren’t seen often in most photos. One final way to check images is to do a reverse image search. This will help identify where the image is from. Also, AI images usually show up in less places than real images, so the reverse image search will return fewer results.

In this image from @joinrealmai on X, the fingers of the hand are oddly shaped and the text on the clothes is absolute gibberish with no real meaning.

Artificial intelligence is still progressing at extremely fast speeds, and it’s being trained on more internet data each day to create higher quality images. The methods mentioned in this article may be completely outdated in a year’s time. As AI grows bigger, we may need to add pieces of code to make sure we can still tell what’s real or not.