I've studied deepfakes for more than 25 years. Here's why AI is making it nearly impossible for you to know what's real.

In today's digital landscape, distinguishing between authentic and fabricated content has become increasingly challenging. Digital forensics expert Hany Farid, who has dedicated over 25 years to studying deepfakes, warns that our collective perception of reality is deteriorating. As AI-generated material becomes more sophisticated, it often eludes detection, leaving the average internet user vulnerable to misinformation. Farid, a former professor at the University of California, Berkeley, and co-founder of the digital forensics startup GetReal, emphasizes that while professionals can analyze content using advanced computational tools, the general public lacks the resources and expertise to do so effectively. As a result, they are often left to navigate a maze of misleading images, videos, and audio clips. Reflecting on his academic journey that began in 1999, Farid recalls a time when the digital landscape was still nascent. Back then, the manipulation of digital evidence was not a significant concern, but as technology evolved, so did the potential for misuse. Farid's early work laid the foundation for digital forensics, a field that has since grown in importance as citizen journalism and social media proliferated. Initially, identifying fake photographs was manageable, as manipulation required skill and often left telltale signs. However, advancements in AI have changed the game. Today, anyone with basic technological skills can produce convincing alterations in mere moments. The gap between what is real and what is artificially created has narrowed significantly, making it difficult to discern authenticity. Farid notes that while earlier AI-generated content was limited in duration, recent developments allow for increasingly longer and more intricate videos. The technology now creates outputs that are visually indistinguishable from reality, posing serious challenges for verification. Although there are still physical inconsistencies that can be detected, the process of authenticating content is labor-intensive and time-consuming, often taking an hour or more. This time frame is problematic in a fast-paced internet environment where misinformation can spread rapidly. Farid’s team often finds itself conducting analyses post-publication, with viral content already racking up millions of views. The implications of misidentifying real versus fake content are profound, affecting legal matters, media integrity, and geopolitical decisions. Perhaps most concerning is the potential erosion of a shared reality. Farid highlights that society is increasingly polarized, engaging in debates over fundamental truths. The decline of a common understanding complicates democratic discourse, as differing perceptions of reality can lead to chaos rather than constructive disagreement. The challenge lies in navigating this new digital realm while striving to uphold a collective sense of truth.