At the end of the nineteenth century human society seemed to be on the brink of profound behavioural change spurred on by advances in technology. Over the previous fifty years innovations had occurred in communications, in transport, and, less benignly, in weaponry. Pioneered by the Lumière brothers and Georges Méliès in Paris, popular culture was about to be transformed by the advent of cinema – ominously enough one of the first pieces of film, ‘The Devil’s Castle’ (1896) was a horror movie depicting a winged demon, initially in the form of a bat and played by Méliès himself, flying into a haunted tower. Improvements in mechanical reproduction altered the way in which information was consumed, whilst ‘trick’ photography called into question whether one could really believe one’s own eyes.[1]

Fast forward another twelve or thirteen decades, and humanity is again confronted with a period of rapid advancement in how information is processed and disseminated. Within the last twelve months the ability of Artificial Intelligence programmes to almost instantly create convincingly-written text of any length and on almost any subject had progressed with dizzying speed. We are witnessing a communications revolution of equal significance to the popularisation of the internet at the end of the 1990s and evolving at a much more rapid pace. As historians this will have profound consequences for how we all carry out work, and how we assess our students’ work.[2] As at the time of the Victorian fin de siècle, responses to this technological change vary from intense optimism to profound concern, and as was the case at the turn of the twentieth century, this change is being pushed forward for primarily commercial reasons. Combined with other forms of AI that are capable of replicating physical characteristics and the human voice, it may be that we are entering a period of ‘two-tier’ truth, where essentially any information or imagery encountered online must be viewed as potentially or even probably fraudulent. The ramifications of this are enormous.

However, that is not the primary focus of the present discussion. Instead, I will draw connections between ChatGPT and its emerging AI cousins and an earlier source of societal anxiety, the Gothic vampire of late-Victorian popular fiction, as it appears in Bram Stoker’s seminal Dracula (1897). In the 1890s there was a definite shift in how technological progress was viewed, and this was reflected in the literature of the time. The presentation of modernity had moved on from the romances of Jules Verne in the 1870s and 1880s, where changes in communication and transport were essentially positive, a means of bringing humanity together, to the early novels of H.G Wells, where science was often malevolent – a means of torture and mass murder, and a challenge to human identity.[3] Of all the horror texts written in the 1890s, Dracula is probably the most concerned with modern technology. Stoker’s great innovation, as critics have noted, was to bring his vampire to London in the present day, rather than setting his story in a far-off land in the early modern period. Thus, a whole armoury of contemporary tech is available to Dracula’s opponents, including typewriters to create multiple copies of reports, phonographs to record one’s voice, a telephone for instant communication, and the latest carbines for the final chase in Transylvania.[4] Coupled with this, the contemporary phrenological theories of Cesare Lombroso are referred to, as is the rise of the New Woman.[5] As Jonathan Harker notes, whilst trapped in Dracula’s castle, ‘It is nineteenth century up-do-date with a vengeance.’[6]

Stoker intended his creation to be exactly contemporary. Today, of course, this means that we instantly locate the novel as archaic, an articulation, a freeze-frame even, of the hopes and fears of Stoker’s own era. How then does this Victorian vampire anticipate AI text creators? Firstly, like ChatGPT, although Dracula very closely resembles a human, he is in fact non-human, a soulless vessel driven only by programmed instinct. Having seen his host climb down the castle walls like a monstrous lizard, Harker wonders: ‘What manner of man is this, or what manner of creature is it in the semblance of man?’[7] Later, as Professor Van Helsing explains the nature of their foe to his comrades, he describes the Count as ‘brute, and more than brute; he is devil in callous [sic], and the heart of him is not.’[8] He is both more than and less than human. Yet Dracula, like text-generating AI, is very good at resembling a human (or the work of a human), and this is what gives him much of his power. He is not like the Frankenstein Monster, a grotesque figure whose physical appearance marks him as an outsider; Dracula can move through London society (he adopts a pseudonym, ‘the Count De Ville’ to aid him in this), he can even (unlike in some of the film adaptations of the novel) walk around in the daylight.[9] Like ChatGPT, when examined closely, his non-human nature (no reflection in a mirror, hair on the palms of his hands) may become apparent, but at first glance there is nothing to distinguish him from the tide of humanity upon which he feeds. That something might be familiar at first glance and yet unsettlingly different on close inspection is at the core of the Gothic as a genre.

Dracula also resembles these AI systems in his ability to learn, to absorb new material, collate it, and exploit it, and to do this beyond the limits of human capacity. Critics of the novel have stressed how important the collation of information is in the quest to defeat Dracula, with the Count as a reactionary anti-modern force.[10] Yet Dracula himself feeds on information and, unlike his mortal opponents, he has an unlimited amount of time to do so. Confronted in his lair in Piccadilly, the vampire taunts his opponents: ‘My revenge is just begun! I spread it over centuries, and time is on my side.’[11] The first thing that Jonathan Harker notes when he arrives at the Count’s castle is the amount of material on Britain, including legal textbooks and railway timetables, that Dracula possesses. He makes use of Harker to improve his own English pronunciation and grammar, he desires that it would be flawless, indistinguishable from that of a native speaker.[12] In other words, Dracula has been doing his homework, and, unlike humanity and like AI, there are apparently no natural limitations on how much research he can carry out.

Given recent reports in the media, it might appear that the total domination of ChatGPT and its rivals over every element of business, study and entertainment reliant on the use of text is imminent. The economic ramifications of these new forms of text-generating AI are real and concerning, as is the speed with which these programmes are developing. Yet, as with the vampire, there are in fact limitations. As Van Helsing says, after listing Dracula’s powers: ‘He can do all these things, yet he is not free. Nay, he is even more prisoner than the slave in the galley, than the madman in his cell.’[13] Dracula is in fact bound to a quite specific geographical area by the need to rest in a coffin lined with earth from his Transylvanian home. He can only enter a residence if someone from within invites him to cross the threshold. He cannot cross running water, and there are ways of preventing him from gaining access to an abode.[14] Similarly, whilst within its domain of text available online AI is wholly dominant, and there is nothing within the reaches of the internet it theoretically cannot make use of, this does not mean that it is uncontainable. From the perspective of a historian, for example, ChatGPT cannot go into a physical archive and spend a week working through correspondence that has not been digitalised. ChatGPT cannot arrange an interview and record and then contextualise someone’s experiences based on it. ChatGPT cannot read a book that is not available online. ChatGPT can fabricate something approaching what a historian might write having carried out that physical research, and it can of course use material an academic has written which is available online (including this blog) to formulate a more convincing text. At the current point, text-generating AI can now produce grammatically flawless written English, and advance within confines a coherent (although not necessarily correct) answer to a question. But it cannot formulate truly original, creative, contrarian, transgressive, or taboo-breaking original arguments. Like the vampire, there are certain boundaries which (at the moment) it cannot cross.

Dracula is in the end defeated. Although his opponents make use of modern technology throughout the novel, to beat their undead enemy in the end they must abandon modernity – to kill the nosferatu one must use the methods of the Middle Ages. Although it might at times seem an attractive option to be able to return to some sort of pre-internet arcadia, before online shopping, newsfeeds, echo chambers, and huge amounts of digitalised historical primary material, this is not going to happen. We must make our peace with AI-generated text, one way or another. There are some vampires you cannot stake.

---

[1] Elizabeth Ezra, Georges Méliès: The Birth of the Auteur (Manchester: Manchester University Press, 2019) chapter one.

[2] ‘AI Bot ChatGPT Stuns Academics with Essay-Writing Skills and Usability’, The Guardian, 04 December 2022 (accessed 05 May 2023).

[3] See Mark R. Hillegas, The Future as Nightmare: H.G Wells and the Anti-Utopians (Oxford: Oxford University Press, 1967) pp.10-14, Steven Mclean, The Early Fiction of H.G Wells: Fantasies of Science (Basingstoke: Palgrave Macmillan, 200), chapter four and chapter five.

[4] Valerie Clemens, ‘Dracula: The Reptilian Brain at the Fin de Siècle’ in Elizabeth Miller (ed.) Dracula: The Shade and the Shadow (Brighton: Desert Island Books, 1998) p.192.

[5] See Clive Leatherdale, Dracula: The Novel & the Legend (Brighton: Desert Island Books, 1993) p.228, Carol A. Senf, ‘Dracula: Stoker’s Response to the New Woman’ Victorian Studies vol. 26, no. 1 (1982): 33-49.

[6] Bram Stoker, Dracula (London: Penguin Books, (originally published 1897, this edition published 1994), p.49.

[7] Stoker, Dracula, p.47.

[8] Stoker, Dracula, p.283.

[9] Leatherdale, Dracula: The Novel and the Legend, p.107.

[10] David Schmid, ‘Is the Pen Mightier Than the Sword? The Contradictory Function of Writing in Dracula’, in Miller (ed.) Dracula: The Shade and the Shadow, p.114.

[11] Stoker, Dracula, p.365.

[12] Stoker, Dracula, p.31.

[13] Stoker, Dracula, p.287.

[14] Christopher Frayling, Vampyres: Genesis and Resurrection from Count Dracula to Vampirella (London: Thames and Hudson, 2016).

---

Dr Daniel Renshaw is a Lecturer at the University of Reading, specialising in migration, diaspora and identity in Britain and Europe from the beginning of the nineteenth century to the present day.

All comments and opinions presented in this article are that of the author.

We have made every effort to abide by UK copyright law but in the instance of any mislabelling of images, please contact the author of the blog post