How Serious Is Netflix’s Use of AI in ‘What Jennifer Did’?

The big picture

• Several entertainment projects, among them
  true detective,
  Late Night with the Devil
  i
  Civil war
  they have been assumed to have used AI-generated imagery.
• The Netflix documentary
  What Jennifer did
  has faced backlash for allegedly using AI to create fake footage of convicted murderer Jennifer Pan.
• The use of artificial intelligence in documentaries raises concerns about trust and authenticity, as viewers may question the validity of all images presented.

In the entertainment industry, there have already been multiple controversial uses of artificially generated images since the available technology leapfrogged in 2023. The trend started with Disney using a series of artificially generated images artificial in the credit sequence for secret invasion last year. Recently, there has been a seemingly never-ending barrage of further criticism directed at the most recent season of Max. The real detectiveas well as movies Late Night with the Devil i Civil war, each of which made use of AI imagery in various ways. However, the most recent scandal is even more alarming. Netflix's latest true crime documentaryWhat Jennifer did It has been alleged that he used AI to create what are presented as real photographs of his subject, convicted murderer Jennifer Pan. The allegations were first made public by Futurism and only later addressed by the documentary's producer, meaning that any use of AI was intended to go undetected. The use of AI footage in a documentary, where there is a different bond of trust between creator and audience, raises even more concerns. about how this technology is used and what will come from its growing use.

What is “What did Jennifer do”?

What Jennifer did was directed by Jenny Popplewellhis second Netflix true-crime documentary after the 2020s American Murder: The Family Next Door. This time, the theme is a Canadian crime story. In 2010, 24-year-old Jennifer Pan was at home with her family when three men broke into the house in an apparent burglary. His mother, Bich Ha Pan, was killed and his father, Huei Hann Pan, was left in a coma. Police investigating the murder initially assumed Jennifer was a victim. But while investigating the incident, details came to light that cast doubt on his story, including security camera footage that showed the killers entering Pan's home but appeared to show no signs of forced entry. When her father awoke from the coma, he said he saw her whispering to the attackers in a friendly manner.

An investigation into Pan's life revealed that he had spent several months lying to his parents about attending college, to the point of fabricating college transcripts and student loan documents. In fact, he had not graduated from high school and his acceptance to Metropolitan University of Toronto had been rescinded. Pan eventually connected with the killers through text messages obtained from his phone and made a partial confession. She was later convicted of murder and attempted murder, as were the three assailants she allegedly hired to kill her parents so they could split their potential inheritance.

Did 'What Jennifer Did' Use AI?

What Jennifer did follows the investigation from the perspective of the detectives assigned to the case, who are interviewed extensively. This means that, at the beginning of the documentary, Pan is presented as she would have appeared the first time: an ordinary 24-year-old who was the random victim of a horrific crime. As the police begin to investigate Pan, three candid photos of her are shown to illustrate a friend's positive description of her as “happy, bubbly, confident and very genuine”. All the pictures show Pan smiling, and in two of them, he shows peace signs to the camera.

One of these images is suspected to be generated by AI due to multiple irregularities in Pan's hands (hands are known to be a challenge for many AI image generators to get right). Also, in the background behind her, a bookshelf appears to contain multiple strange objects that cannot be identified. A second photo, a close-up of Pan, is suspected to have been AI-generated due to an irregularity in his gum that cannot be identified elsewhere. The irregularities have been cataloged thoroughly.

At the time of writing, the only response from any of the documentary's creators to these allegations has come Jeremy Grimaldi, who is credited as an executive producer. Grimaldi is also a crime reporter who covered the Pan trial and wrote the 2016 book A Daughter's Deadly Deception: The Jennifer Pan Story. In response to an inquiry from the Toronto Star, Grimaldi denied that Pan's image had been altered while acknowledging that something had been done to the background of the image to disguise the identity of the person who had provided the original photograph.. “Any filmmaker is going to use different tools, like Photoshop, on movies,” Grimaldi said. He added: “Jennifer's photos are real photos of her. The foreground is exactly her. The background has been anonymized to protect the source.” This statement is unclear as to what method was used to “anonymize” the background of the image. Nor does it explain the many signs that suggest both the background and foreground of the image in question are AI-generated.

Why does it matter if a documentary uses AI-generated footage?

To some, it may not seem like a big deal that these images were generated by AI. After all, they only appear on the screen for a few seconds. And while Jennifer Pan is appealing her conviction, it's unclear how a picture that was meant to show how normal she was on the surface would sway a judge or jury. As many argue, documentaries already create a distorted image of reality simply through the editing process by reflecting the director's point of view. True-crime documentaries already augment reality through the extensive use of re-enactments, which they combine with archival footage. Is this new use of technology to create a false image important?

Probably the biggest reason it's important is that even a handful of AI footage in a single documentary makes it hard to trust. none of the images shown in the documentary in question. It also casts a shadow of doubt on all the information presented in the documentary. watching What Jennifer did with the knowledge that it's supposed to have an AI image or two makes all the photos used look suspicious. This makes sense, because while most of us have a pretty tuned-in idea of ​​which footage from a true-crime documentary has been played in a studio, even when it's not labeled, it can still be very difficult to tell the difference an AI image of the real thing. thing Online, people have said they won't trust any of Netflix's true crime documentaries anymore.

Specifically, it may appear that these particular images are not this harmful There's already a picture of Jennifer Pan smiling that no one is arguing about, so what's the difference in generating a few more to fill the frame? But even in this seemingly neutral decision, there are hidden consequences. There's a difference between having just one photo of a happy person (which looks like it might be a selfie) and having several. There is a difference between an image taken in a club and one taken in a bedroom, as the AI-generated image appears to be, especially since Jennifer Pan is alleged to have conspired with an ex-boyfriend to commit the crime. Was he the one who took the photo?

The growing use of AI is worrying

These may seem like extremely minor concerns, but imagine these questions multiplied by thousands if using AI images in this way becomes commonplace. The last thing worth noting is this There's still a lot we don't understand about how image-generating AI works. The technology is often referred to as a “black box,” because even its creators understand very little of what goes on inside it. We've built hardware and software that can take billions of images and recognize enough patterns to re-approximate them on command, and it can adjust itself based on human feedback. Although we call it “artificial intelligence,” it actually works nothing like a human mind. Image data used to train on social networks such as Friendster and MySpace were often made publicly available by the public. Could an AI, asked to fill a bedroom wall with “typical photographs”, generate a picture of you?

What we do know is that the temptation to use this software is strong. Imagine using AI in a documentary about a killer who used Photoshop to create a fake life for himself, ignoring the irony! It is socially discouraged to use unlabeled AI in any form. But that may not be enough to stop these images from bleeding into our media, gradually eroding our ability to know what's real.

What Jennifer did is available to stream on Netflix.

Watch on Netflix