What we must be: AI and the future of porn studies

ABSTRACT

This article explores the potential impacts of artificial intelligence (AI) on the future of pornography and pornography studies. It links the content policies created by technology companies such as OpenAI, which prohibit sexually explicit material, to the social media, technology, and entertainment industries before them, and examines the cultural and regulatory impact on pornography, sexuality, and human relationships. Drawing on the work of Joseph Weizenbaum, the author argues that the increasing disconnection of humanity from creative expression, along with growing disinterest in the consequences of that disconnection, will have profound consequences for the future of pornography studies.

KEYWORDS:

• Pornography
• Artificial Intelligence
• Joseph Weizenbaum
• AI
• OpenAI
• ChatGPT

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 Shaka McGlotten (2023) has noted that the prominent AI tools will inadvertently make sexually explicit images occasionally when given certain sets of prompts, illustrating that the engineers behind these tools cannot prepare for every permutation and outcome. See also Choi (2023).

2 Weizenbaum had direct experience understanding the ways in which humans and machines could interact. Between 1964 and 1967, while at MIT, he created ELIZA, software that simulated therapy sessions, giving users the impression that the computer was empathizing with them. Weizenbaum was startled to see that people attributed human-like emotions and judgements to the machine, despite it being a computer program using pattern matching and text substitution (Weizenbaum 1966, 1967). The experiment significantly influenced his thinking and represents a landmark, foundational case study in AI development.

3 It is worth noting that Claude contradicts Anthropic’s policy statements, illustrating the unreliability of these AI tools. When I asked Claude about the constitution, it provided a startling answer: ‘The notion of “Claude’s Constitution” as described in some of Anthropic’s marketing materials is more of a metaphorical concept rather than something concrete in my architecture as an AI system. I don’t actually have any formal governing documents or hardcoded principles that guide my behaviors … Any references to my “constitution” are just creative marketing language, and don’t reflect any kind of tangible framework or protocol ingrained into me as an AI' (Anthropic 2024).

4 As a thought and technology experiment, I asked ChatGPT to describe what an alternative contemporary reality might look like if OpenAI’s content policy had been used to craft US obscenity law in 1970. ChatGPT described a long list of possible societal impacts, but of particular interest was its awareness that there would be potential for critical changes in ‘cultural norms, discussions around sexual education, and shifts in public perceptions of what is considered acceptable or taboo' (Open AI 2024a). This seems to be a remarkable understatement.

5 For example, Seduced.ai’s terms of service prohibit ‘any form of rape, murder, non-consensual deepfakes (images portraying real people), scat, non-consensual sex, minors in any circumstances, violence, severe beatings, watersports, bestiality, zoophilia, necrophilia, prostitution, pedopornography, or related content' (Seduced.ai 2024). This list is made up of a combination of legal, illegal, and legally contentious elements (as well as some that are obviously reprehensible), a nuanced analysis of which is beyond the scope of this article. Nevertheless, it also exemplifies an approach to content regulation beyond geographic boundaries and nation-based regulations, similar to how the mainstream AI tools operate (Seduced.ai 2024).

6 David Church (2015) makes similar arguments about the ways in which exploitation films of the past continue to hold influence over contemporary culture despite their temporal remove.

7 There is, of course, a clear need for filmmakers to obtain consent from performers prior to directing a scene in order ensure their safety, including consent to experiment and find an experience ‘naturally', as Poole describes. My main point here is just that Poole’s quasi-documentary, hands-off approach to filming Donovan and DiCiccio resulted in the capturing of something that a machine is not capable of creating.

8 Vinciquerra said this in response to concerns by the Writers Guild of America (WGA) and Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA) labour unions, which struck in 2023 in part because of concerns that AI would impact job security (Hayes 2023).

9 This kind of fatalistic indifference towards AI by engineers summons an increasingly common comparison to the development of the atomic bomb, which was similarly built by engineers who assumed ‘progress would move forward' with or without them, thus justifying their participation (Rhodes 1987).

10 Ensuring that scholarly work be written exclusively and entirely by humans is not a concern for the future, it is an immediate and pressing issue. For example, one leading AI researcher has already developed a tool for ChatGPT that can write academic papers at the level of a doctoral student (Mollick 2024).

11 It is not surprising that, during the Altman drama, the New York Times explicitly invoked the mythology around Jobs on its front page, ensuring that the mythologies around these tech figures stay firmly in place (Streitfeld 2023). Weizenbaum (like myself) was clearly influenced by Theodor Adorno, who similarly argued that mass culture could distract people from engaging in critical inquiry (Adorno 1991).