According to a new study published in the Proceedings Of The National Academy Of Sciences Of The United States Of America (PNAS), people’s ability to detect fakes was no better than a random guess, & they actually rated the fictional faces as more credible than the real ones.
The researchers created 400 fake faces with an equal split of black & white faces, & 100 faces from each of four ethnic groups: Caucasian, East Asian, & South Asian. In this study, the researchers compared each of these with real faces pulled from the database that was originally used to train the GAN, which had been judged similar by another neural network.
A crowdsourcing platform called Amazon Mechanical Turk was used to recruit 315 participants. The participants were asked to judge 128 faces from a combined dataset & determine whether they were fake or not. Their accuracy rate was only 48%, which was worse than the 50% you would expect from a random guess.
The authors’ end the paper saying: We, therefore, encourage those developing these technologies to consider whether the associated risks are greater than their benefits. If so, then we discourage the development of technology simply because it is possible. If not, then we encourage the parallel development of reasonable safeguards to help mitigate the inevitable harms from the resulting synthetic media. Safeguards could include, for example, incorporating robust watermarks into the image and video synthesis networks that would provide a downstream mechanism for reliable identification (14). Because it is the democratization of access to this powerful technology that poses the most significant threat, we also encourage reconsideration of the often laissez-faire approach to the public and unrestricted releasing of code for anyone to incorporate into any application.
Click here to read the study.
