Sora 2: A New Age of Taxidermy

To say generative AI has progressed rapidly would be an understatement. The days of an AI-generated, barely recognizable Will Smith clumsily eating spaghetti are long gone. Now, we have hyperrealistic videos of Martin Luther King Jr. as a DJ and Michael Jackson hosting Instagram Live events. Many of these infamous videos were brought to you by Sora 2. OpenAI’s newest prompt-to-video model is the newest such digital canvas for our culture’s surreal imagination.

Despite its impressive computing power, Sora 2 failed to distance itself from the controversy that AI often carries. The battle over creativity, misinformation, and intellectual ownership continues in our cultural conversations about AI. However, Sora 2 fans the flames of a larger issue: the gross resurrection and puppeteering of the dead through AI-generated content. Yes, those AI videos of Tupac Shakur and Kobe Bryant hanging out might be funny. However, having temporary control over a person’s likeness is a harrowing thought. Where are the ethical boundaries on generating someone’s image, if there are any at all? What happens when AI’s hyperrealism is used maliciously? This rising, problematic genre is what many on social media are calling “AI necromancy,” or “digital taxidermy.”

Unfortunately, AI has begun to cross the threshold of cultural normalcy. You can especially see instances of AI in our advertising. Recently, the online betting platform Kalshi went viral for its AI-generated commercial, which premiered during the NBA Finals. According to a report by NPR, the ad, created by AI content creator P.J. Accetturo, “racked up more than 3 million views on Kalshi’s X account.” Regardless of the response, the ad clearly found an audience. It proved to companies that AI has a place in marketing.

It’s Sora 2’s open policy regarding the dead that has strengthened anti-AI sentiment. In an interview with PCMag, OpenAI told the publication that while they “block depictions of public figures,” they “…allow the generation of historical figures.” This opens up a myriad of ethical dilemmas. It turns the likeness of a human being into a digital puppet for entertainment. No longer is Bob Ross’s image just the art and kindness he left behind. It can be a fake video of him promoting AI, uncomfortably mimicked to an unsettling degree of detail. The deceased are used as props for amusement or agendas. It turns legacy into public domain, and people into clay dolls to humiliate. It’s reviving someone and shaping them into a desired figure and form, a digital taxidermy job.

Not to mention the effect those “harmless” videos have on those still grieving. Zelda Williams, daughter of the late Robin Williams, recently denounced the rampant AI-generated videos of her father that people were sending her. According to E! News, Williams posted a firm statement on her story, saying, “Stop believing I wanna see it or that I’ll understand, I don’t and I won’t.” She also emphasized how much her father would hate generative AI and her disgust with the idea: “You’re not making art, you’re making disgusting, over-processed hotdogs out of the lives of human beings….” Bernice King, daughter of MLK Jr., retweeted the quote on Twitter, making it known that the same idea applies to her father.

The idea is not exclusive to celebrities either. In another report from NPR, Christopher Pelkey was shot and killed by Gabriel Paul Horcasitas at a red light in 2021. In May of this year, Stacey Wales, Pelkey’s sister, used an AI-generated video of her now deceased brother to address the courtroom in an impact statement during Horcasitas’s sentencing. While Wales noted that the video was “healing” and the intentions were harmless, the ethics and legality moving forward are unprecedented. What if deepfakes are used as evidence, or AI recreations of the dead are used to persuade juries? The notion is unsettling to say the least.

This genre of digital taxidermy is one of perversion. Its potential is incredibly volatile, especially when it comes to AI’s misinformation problem. Take, for example, the short video of Jackson and Tupac talking below. It looks uncanny in how convincingly real it seems, aside from the monotone AI voice. While some can easily identify AI from a mile away, not everyone has the capacity to verify truth in the age of fast-paced information streams. AI makes it easy to mislead and lie. Digital taxidermy can extend beyond entertainment. Now, anyone can manipulate the lives of historical figures, making them say things they’d never say or do things they’d never do. It becomes a slippery slope of misinformation and fabricated realities. It can be a grim world to look forward to.

But hope remains. According to a Variety report, the Creative Artists Agency has been pushing back against OpenAI and Sora 2 for encroaching on the intellectual property of celebrities and artists alike. In a public statement, the CAA said, “It is clear that OpenAI/Sora exposes our clients and their intellectual property to significant risk,” and that compensating artists for their likeness and property is a “fundamental right.” While the intentions are not wholly anti-AI so much as they’re pro-compensation, the fight to regulate AI continues. And in a world that embraces AI tighter and tighter, that’s all we can do for those who can’t opt out.