Deepfakes can fool people, study finds

In the midst of a strike by screenwriters and actors in favor of limiting artificial intelligence in Hollywood, a study confirms that altered clips can deceive people.

Brad Pitt in The Shining, Will Smith in The Matrix, Charlize Theron as Captain Marvel, and Chris Pratt in Indiana Jones. These are some of the examples of deepfake videos produced by researchers who investigated the potential for creating false memories through supposed remakes of Hollywood cinema.

Based on 436 responses, the team concluded that deepfakes are indeed capable of manipulating the memory of a large number of people. Some even rated the "remakes" as better than the original works.

Deepfakes are videos produced using artificial intelligence (AI) that swap the face or voice of one individual for another. Recently, the tools for creating these clips have become much cheaper and more accessible, which has led to increased discussions about the risks — including political — of disinformation and fake news.

Participants' perceptions:

The test subjects participated online. The study included watching deepfake videos of remakes of films starring different actors. Participants were also presented with clips of real remakes, which included Willy Wonka and the Chocolate Factory, The Terminator, Carrie, and Tomb Raider.

It was then observed that 49% of the participants believed the pieces were real. Many also reported that they remembered the fake remakes and even found them better than the originals — as was the case with Captain Marvel with Charlize Theron (in fact, it is with Brie Larson), in which 41% of the public pointed to the preference.

However, in addition to the videos, in some cases the participants only read the descriptions of the fake films. The rates of false memory built from these texts were equally high, suggesting that deepfake technology may not be as powerful as other tools.

Most of the interviewees highlighted that they were uncomfortable with the use of deepfake technology in the reworking of films. Among the main concerns are the disrespect for artistic integrity and the interruption of shared social experiences within cinemas.

It is worth noting that, currently, actors and screenwriters in Hollywood are mobilized in a strike with the aim of precisely limiting the use of artificial intelligence in film and television productions. They fear that automation tools will replace them with digital clones or ghostwriters.

In a press release, Gillian Murphy, lead researcher and professor at University College Cork in Ireland, argues that these findings can help contribute to these discussions, assist in design, and serve as a benchmark for regulating deepfake technology in films.

“While deepfakes are a major concern for many reasons, such as non-consensual pornography and bullying, the current study suggests that they are not the only ones powerful in distorting our memories. While deepfakes led people to form false memories at fairly high rates in this study, we achieved the same effects using simple text,” Murphy concludes.