AIACUREF AI: A CURE ORBAUMO FOR BAUMOL’S DISEASE? LSDISEAS EAIACURE FORBAUM OLSDISEA SEAIACUR CREATe Working Paper 2024/8 GILLIAN DOYLE SABINE BAUMANN AI: A cure for Baumol’s disease? Gillian Doyle and Sabine Baumann * Abstract The production of outputs in cultural industries, including film and television etc, is said to suffer from Baumol’s disease in that, because of its emphasis on creative labour elements which cannot readily be mechanised, replicated by computers nor streamlined, it is prone to aboveaverage inflation. However, recent developments in generative AI, with its capacity to assist in generating text, images and sound, have established unprecedented opportunities to automate and support aspects of creative work across media industries. What does this imply for Baumol’s cost disease? This paper examines recent developments in AI, and analyses to what extent, by potentially reducing costs and raising productivity in inflation-prone media content creation activities, these technologies challenge conventional theory and effectively counteract Baumol’s disease. Key Words Artificial Intelligence; Generative AI; Baumol’s disease; media labour markets; cost inflation; job displacement * Gillian Doyle is Professor of Media Economics at the University of Glasgow, based in the Centre for Cultural Policy Research (CCPR), and Sabine Baumann is Professor for Digital Business at Berlin School of Economics and Law and Scientific Director at OFFIS Institute for Information Technology. 1 1. Introduction Referred to as ‘the jewel in the crown of cultural economics’, Baumol’s cost disease embodies the proposition that ‘the failure of technical progress in the arts to keep pace with technical progress in the economy as a whole, while wages nevertheless rise everywhere more or less at the same rate, necessarily implies irremediable cost inflation in the arts’ (Blaug, 2001: 131). The production of outputs in cultural industries, including film and television etc, is said to suffer from Baumol’s disease in that ‘because creativity is inherently labour-intensive and because labour costs tend to rise more quickly than others, costs across these sectors will tend to rise at a faster rate than inflation’ (Doyle, 2013: 101). Named after US economist Will Baumol, Baumol’s disease refers to the ‘inevitable’ escalation of real costs that occur in labour-intensive industries such as the arts (Towse, 1997) where, because of the emphasis on specific sorts of creative labour, ‘flexibility does not exist’ to substitute capital for this input (Towse, 2020: 416). According to Baumol, growth and indeed the survival of sectors such as performing arts is threatened by the persistence of above-average inflation because labour elements cannot readily be mechanised, replicated by computers nor streamlined. However, recent developments in generative Artificial Intelligence (AI) have established unprecedented opportunities to automate and support aspects of creative work across media and creative industries. The capacity of generative AI to assist in generation of text, images and sound means this technological tool has significant potential to transform and disrupt processes of video, audio and text creation. What does this imply for Baumol’s cost disease? While some media executives highlight concerns about how developments in AI might encourage copyright infringement by, for example, text chatbots and image generators usage of content (Criddle et al., 2023), others acknowledge the positive potential and ‘a massive opportunity to optimise creative work’ (Read, cited in Pitel and Storbeck, 2023). The embrace of AI as a tool to support creative work, although controversial, appears to offer significant opportunities to automate and support production of an array of creative outputs, from advertising campaigns and animations to songs and storylines for the printed page and digital screen. This paper examines recent developments in AI, and analyses to what extent, by potentially reducing costs and raising productivity in inflation-prone media content creation activities, this tool challenges conventional theory and effectively counteracts Baumol’s cost disease. In questioning the validity, in a contemporary context, of Baumol’s influential ideas about impediments to productivity growth in creative industries, AI technologies extend earlier work 2 whose concern is theorisation of media production in the digital era, such as by Shapiro and Varian (1999) and Dwyer (2015), and it breaks new ground by focusing specifically on the highly disruptive capability of AI. Although Baumol’s disease has previously encountered scepticism, for example from Cowen (1996), and although earlier research has highlighted how AI is likely to displace jobs (Arntz, Gregory and Zieharn, 2016; Frey and Osborne, 2013; Nedelkoska and Quintini, 2018) and transform creative industries (Caramiaux, Lotte and Geurts, 2019; McKinsey, 2018), little or no previous work has focused specifically on how the growing integration of AI-powered techniques pose a highly potent challenge to assumptions about labour-intensive human work within media and creative industries which lie at the core of Baumol’s theory. With regard to methods, this paper draws on analysis of secondary source literature, particularly in the realms of media and cultural economics and media business studies, and on reports and specialist press coverage of the development of AI and of media and creative industries. In the sections that follow, we first of all assess the history and significance of Baumol’s cost disease and then we examine the recent rise of AI and the ways in which AI is transforming processes of creativity and co-creation with media and creative sectors. Finally, this paper homes in on its central question: does AI provide a cure for Baumol’s cost disease? 2. Baumol and Costs in Arts, Creative and Media Industries The concept of cost disease was first introduced by Baumol and Bowen in their influential book Perfoming Arts: The Economic Dilemma (1966). The authors gathered and compared data for both manufacturing (the so-called ‘dynamic’ sector) and performing arts such as opera, ballet and orchestras (the ‘stagnant’ sector) in the US and UK, identifying how costs and prices were rising more quickly in the latter than the former (ibid). Baumol pointed out that, because of technology-driven and other advances and opportunities for efficiencies, the amount of labour needed to produce ‘typical’ manufactured goods has generally declined over time. But, in other sectors ‘which, by their nature, permit only sporadic increases in productivity’ such gains are not available (Baumol, 1967: 416). The output of, for example, a singer or a dancer or a string quartet is embodied in and identical to their labour, so there is little prospect of streamlining or automating this in ways that help increase output per hour. Baumol observed that growth lags behind in sectors where there is little scope to substitute capital for labour (with corresponding increases in output per man-hour) and therefore the economy as a whole is subject to ‘unbalanced’ growth. Baumol and Bowen’s ideas on cost inflation in the arts have remained influential over time, inspiring and ‘directing empirical studies, theoretical analysis and animating debates [about 3 public funding of arts] in an era of federal budget constraints’ (Tiongson, 1997: 118). The notion that the survival of live performing arts such as opera, concert and theatre is threatened by persistently above-average cost inflation lends weight to arguments favouring subsidisation of the arts from the public purse. Over the years, several studies have confirmed the existence of cost disease in sectors including, for example, orchestras in the US (Felton, 1994). However empirical evidence has not always been conclusive: for example, a study of inflation in performing arts in the UK carried out by Peacock, Shoesmith and Millner in the 1970s found no evidence of Baumol’s cost disease (Heilbrun, 2011: 71; Towse, 1997). A study of festivals by Frey found little sign of cost disease in this sector (Frey, 1996). Throsby argues that, while many empirical studies have confirmed the existence of cost disease, it remains that most companies find ways of mitigating this so that its impact on performing arts is ‘muted’ (Throsby, 2010: 69). Whether or not a secure diagnosis of cost disease is arrived at may hinge partly on how exactly ‘production’ and ‘outputs’ are defined. It also depends on whether and how initial processes of content creation are distinguished from reproduction and/or distribution of content. In saying why he doesn’t believe in Baumol’s cost disease, Cowen argues that, even if a contemporary string quartet playing a Mozart composition requires the exact same labour and time (four players playing for 40 minutes) as back in Mozart’s time, productivity will have increased because of innovations in electronic reproduction which mean that the quartet’s output can be now enjoyed by many more listeners. When measured in units of consumption, productivity has ‘skyrocketed’ (Cowen, 1996: 208). However, a counter-argument is that recordings and reproductions, however much they extend opportunities for consumption, are not in fact the same as live performance. Walter Benjamin’s pioneering essay on cultural criticism elucidates how, for example with recorded films as opposed to live on-stage performances, the connection between performer and audience is compromised – ‘mechanical reproduction of art changes the reaction of the masses toward art’ (Benjamin,1935:10). The distinction between initial production of original content (with associated live performances of creative labour) versus reproduction and/or onward distribution of such content is acknowledged in later iterations of Baumol’s work where he notes that, whereas content distribution activities (e.g. broadcasting) are liable to benefit from technological advances, performance of dedicated creative labour in the initial production stage resists standardization or streamlining (Baumol, 1993: 20). Preston and Sparviero argue that to ‘give justice to’ the most important aspect of Baumol’s disease – the intuition that some sorts of creative labour are not replaceable by new technologies - the definition of Baumol’s disease ought to focus more sharply on the production stage and on the ‘creative inputs’ or 4 characteristics of creative labour which are irreplaceable by new technologies, such as formulation of original ideas and concepts (Preston and Sparviero, 2009). Baumol’s notion that some types of creative labour cannot be replaced by new technologies clearly applies to media and other creative content production as well as performing arts. The persistent need for novelty and innovation within media content makes production of it a process that is heavily and inescapably reliant on human creative labour as an input. The fact that ‘each television programme, film, newspaper and magazine edition must offer messages, images or stories that are novel and unique’ and associated need for creative labour has very significant implications for the economics of media content production’ (Doyle, 2013: 100). Making media content, drawing on unique inputs from, for example, storytellers and performers, is an expensive business with high original or so-called ‘first copy’ costs and the creative labour involved in production cannot readily be automated. This is not to say that media companies are devoid of opportunities to capitalise on economies of scale or to achieve some forms of standardisation within production activities (Dwyer, 2015). On the contrary, economies of scale are a common feature of the industry. Economies of scale exist within processes of media content production because, for example, television producers with a number of ‘returners’ or re-commissions of shows are able to achieve better rates of overhead recovery (Doyle, Paterson and Barr, 2021). More broadly, as earlier work in economics of media has shown, economies of scale are a prevalent feature of the whole vertically integrated process of supplying media content to audiences because, while first copy production costs may well be high, once these have been covered, the marginal costs involved in facilitating consumption of said content by additional viewers and/or audience segments are typically low (Doyle, 2013). Another economic feature of media content production is uncertainty about how much demand there is likely to be for an as yet unproduced item of content or, as US economist Richard Caves refers to it, the law ‘nobody knows’ (Caves, 2000:3). Therefore, as earlier work on economics of media production has identified, the high costs and uncertainties involved in production encourages and indeed necessitates use of strategies aimed at risk reduction (Picard, 2011). These include content portfolio strategies (Collins, Garnham and Locksley., 1988; Hoskins, McFadyen and Finn, 1997) and processes of imitation, repetition and re-use of formulas that have proven successful with audiences (Bielby and Bielby, 1994). Although media companies have developed numerous strategies to counter and mitigate risk, it remains that a core challenge is the ongoing and constant need for novelty and innovation within 5 new items of content which, in turn, have been taken to imply that human ingenuity and creative labour are indispensable components of media content production. The problem with reliance on creative labour inputs, from writers and storytellers to actors and musicians, is that, historically at least, these have resisted automation. However, as the following sections will show, current developments in AI are challenging the supposition that creative labour inputs are irreplaceable. 3. The Rise of Artificial Intelligence Questions surrounding what constitutes AI and where the origins of related technologies lie have inspired debate. While frequently misconceived as being just a single system, AI in fact denotes a wide set of technologies (Shin et al., 2022) one of which - generative AI – has recently gained prominence through the rise of Large Language Models (LLMs) such as ChatGPT. AI technologies are seen as ‘intelligent’ in that they emulate human ways of interpreting external data, learning and adapting behaviour to achieve targeted objectives and tasks (Duan et al., 2019; Guzman and Lewis, 2020; Haenlein and Kaplan, 2019; Nilsson, 2009; Shin et al., 2022). Some authors identify broad stages in the development of AI (Haenlein and Kaplan, 2019). Whereas ‘narrow’ (also called weak) AI systems are those that have been built to complete a single, narrowly defined and structured task (e.g. image recognition or language translation), more highly evolved ‘general’ AI systems are capable of learning and adapting based on limited external data and have potential to tackle problems from a number of differing domains (Goertzel, 2014; Leavy, 2023; Russell, 2021). At an even more advanced evolutionary level, ‘superintelligent’ AI systems display cognitive, emotional and social skills, can develop thinking skills of their own and are capable of surpassing human intelligence. Several events in the development of AI paved the way for its application in media production. While the concept of self-organizing devices that might relieve humans of unwanted chores can be traced to the ancient Greek philosophers (Nilsson, 2009), the more recent foundation of AI as an academic field of research dates to the seminal 1956 Dartmouth conference organized by John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon (McCarthy et al., 1956). Scholars in this field sought to understand how human brains produce complex patterns through connected brain cells and reproduced this, in the context of machines, in the form of nodes and links for artificial neural networks (McCulloch and Pitts, 1943; Rosenblatt, 1957). The advent of increasingly powerful computers in the 1950s and 1960s encouraged a focus on tasks, for instance solving puzzles, playing chess or answering simple questions, that machines could be trained to carry out (Garvey, 2021). AI grew in sophistication and, in 1966, Joseph Weizenbaum 6 presented ELIZA, one of the earliest chatbots that demonstrated how a computer programme could engage in natural language conversation (Natale, 2019) – a key milestone. This set the stage for future AI systems capable of generating human-like text. From the 1980s onwards, computer assisted journalism and so-called robotic journalism evolved further as the use and integration of algorithms within news production and as automated summation and narration became more commonplace, fuelled by increased computing power and the availability of increasingly large datasets (Latar, 2018). Early forays into AI-powered journalism by organizations such as The Washington Post and The Associated Press, who started experimenting with automatically generated content for tasks such as writing financial reports and sports stories, hinted at the potential for AI technology to transform content production (Moran and Shaikh, 2022) In 2018, OpenAI released their GPT-1 (generative pre-trained transformer) language model – a tool that uses machine-learning techniques to generate natural language. Using books as input data, GPT-1 was trained to predict the next word in a sentence. Subsequent iterations in 2019, 2020 and 2023 offered major improvements in ChatGPT’s ability, when prompted, to generate contextually relevant text across a range of topics (Murgia, 2023). That the platform is free to use and generates coherent text that readily resembles human speech has propelled its rapid adoption, notwithstanding concerns in some quarters about the implications of its widespread usage (Roe and Perkins, 2023). Since AI content generation tools can create images, audio and audiovisual material as well as text, their emergence and growth have vast implications for media and entertainment content production as well as journalism. 4. AI in the Media AI has been a major transformative force, changing the way media content is produced and distributed. AI systems can be seen as ‘creative rocket fuel’ for media businesses and examples abound of how it is assisting in media content creation processes (Connock, 2023: 8). Drawing on reports, academic papers and specialist press coverage of recent developments we now consider how AI is affecting job roles across a number of sectors of media, highlighting how career paths are being reshaped, which roles are subject to diminution, how creative processes are being enhanced and where new skill sets are required. This analysis is not intended to provide a comprehensive or in-depth survey of AI systems but rather to illustrate how AI is impacting across the board on media industries, activities and jobs, from targeted advertising to digital stand-ins and from synthetic voices to LLMs to recommendation engines. A key point of interest is to what extent AI is displacing versus supporting and augmenting artistic labour. 7 Film In film, AI has demonstrated its immense potential to undertake a variety of tasks along the value creation process from pre-production and production through to distribution – see Table 1. For example, advanced algorithms can analyze scripts for narrative structure, character development, and potential audience appeal, helping producers and studios make informed decisions about which projects to invest in (The Economist, 2024, Wu, 2024). AI can also be used to produce scripts and offer alternative ways of storytelling through film involving profound changes in narrative delivery and potential for connection with audiences (Suich Bass, 2023). In post-production, AI is increasingly being used for tasks including editing, visual effects (VFX) and colour grading. Its usage can reduce editing time from days to a few hours and it can enable automation of some repetitive tasks or assist in creating complex visual effects, such as rotoscoping in Everything Everywhere All at Once (2022), which traditionally required extensive human labour (Suich Bass, 2023). Dubbing for foreign language versions can be automated through AI, including lipsyncing (O’Connor, 2022; O’Connor, Grimes and Riddle, 2023). Tools such as Speechify can rapidly clone voices with minimal training data and then generate new audio with those voices (O’Connor, 2022; Speechify, 2023). AI applications not only streamline workflows and enhance efficiency, but also open up new creative possibilities. For example, while shooting the film Here, AI was used to swap the faces of actors Tom Hanks and Robin Wright for automatically de-aged versions during their live performances with no additional compositing required (Rubin, 2023). The same technology can be used to create photo-realistic avatars (Hemphill, 2023). AI also plays a significant role in film marketing and distribution. By analyzing viewer data and market trends, AI can help studios target their marketing efforts more effectively ensuring that films reach their intended audience (Klein, 2019; Farchy and Denis, 2020). AI-driven analysis can predict box office performance and streaming success, guide distribution strategies, and optimize release schedules (The Economist, 2023a). In an industry characterized by unpredictability combined with high production costs (Caves, 2000), the ability to use predictive analysis to improve marketing and mitigate risks is clearly strategically advantageous. Advancements in AI technologies have significant implications for job roles. For example, script analysts can operate more speedily and effectively by drawing on AI tools to quantify specific elements of structure, theme, plot and dialogue which are likely to be well received by audiences (Townsend, 2024). But, while this enhances productivity, human judgement remains an indispensable component of the process, albeit that script analysis now requires skills in usage 8 of the AI tools that complement and augment traditional functions. So, in line with Acemoğlu and Restrepo’s observations about the history of automation, the displacement effect is ‘counterbalanced by technologies that create new tasks in which labour has a comparative advantage’ (Acemoğlu and Restrepo 2019: 4) – while AI tools save time, they also ‘reinstate’ or increase demand for labour in non-automated tasks some of which require new specialist skills – see Table 1. Automated production and editing tools empower creative directors, editors or CGI technicians as they enhance creativity and improve efficiency (Khan, 2023; O’Connor, Grimes and Riddle, 2023). For example, extras or minor roles may soon be substituted by digital stand-ins and, while animals may be among the first to be replaced (The Economist, 2023b), over the longer term many roles, aside from those played by a few superstars, could involve digital doubles (O’Connor, 2022; O’Connor, Grimes and Riddle, 2023). So the integration of AI in filmmaking undoubtedly involves potential for diminution in some traditional roles. Nonetheless, leading practitioners and experts take the view that, when it comes to high quality film and television content production, ‘generative AI won’t be a direct substitute for artists but it can be a tool that augments their capabilities’ (Schomar, 2023). For example in script-writing, AI tools can speed up the process, thus reducing costs, but human input is still centrally required to ensure coherence and quality, ‘underlining that AI is far from replacing the depth and authenticity of human creativity’ (Cheng, 2024: 52). In short, while AI offers valuable opportunities for technical enhancement, the need for human creativity, intuition, and storytelling remains paramount in the film industry (O’Connor, Grimes and Criddle, 2023; Townsend, 2024). Television AI has impacted on television production and post-production in much the same way as film, bringing valuable enhancements and cost-efficiency improvements – see Table 1. In creative processes such as scriptwriting, use of AI tools can yield data-driven insights that, in turn, influence content creation decisions (Khan, 2023; Wu, 2024). In addition, AI algorithms are being used by content service providers to enhance content selection decisions and to provide personalized content recommendations to viewers, in order to improve discoverability and so increase engagement and satisfaction (Doyle, 2018; Khan, 2023). In the realm of television advertising, AI is also used extensively to facilitate targeting, enabling more precise placement based on individual viewer interests and preferences. Targeted approaches increase the relevance and effectiveness of advertising campaigns (Pitel and Storbeck, 2023). 9 AI is used in various ways in television content production, including for example, to control cameras and adjust lighting during shoots, or to perform some of the repetitive aspects of video editing, thus freeing up producers to perform more creative tasks. Use of AI in television content-making, because it allows different sorts of technologies such as ‘virtual production, motion capture and real time graphics’ to be combined, also facilitates new creative possibilities, such as AI-driven ‘synthetic humans that move, speak and react within a virtual studio’ (Wormwell, 2023). But, while AI can enhance productivity and creativity in numerous ways, the core function of story-telling remains one in which human input is seen as indispensable. As television producer and industry expert Evan Shapiro put it: … it’s very unlikely that a great screenplay is going to be written by a piece of AI, because generative AI does not stay awake at night worrying about the relationship with its father. The imperfections that make us human are what make us able to make art. (Shapiro in Bell, 2024). In the television sector and elsewhere across media, an ever-growing reliance on AI has raised a range of concerns, such as about privacy, transparency, uses of and monopolization of data, pluralism and diversity of content (Doyle, 2018) and, of particular pertinence here, potential job displacement. But most evidence thus far suggests that, rather than actually displacing the need for human input, the development of AI has been the accompanied by emergence of new job roles and the need for new skillsets – see Table 1. As with film, automation has resulted in attrition of some roles while at the same time increasing demand for others. For example, at present there has been growing demand for television professionals skilled in AI programming and management and in advanced CGI and VFX skills (The Economist, 2024). In short, AI is seen as augmenting and enhancing the productivity of television producers – ‘allow[ing] creatives to be even more creative – and more productive too (Wormwood, 2023) – rather than displacing them. Journalism and News Content Production As discussed earlier, AI has been used in journalism and news production from the 1980s onwards to assist, for example in sifting and analysing relevant information and identifying patterns or in summarising or creating news narratives (Latar, 2018; Moran and Shaikh, 2022). Currently, the main uses of AI in news media include research, creation of texts and images plus editing (Murphy, 2024). AI can provide support across the entire process of news creation, including inception of novel ideas, data analysis, drafting, proofreading, visualization and factchecking (Caswell, 2023; Chan-Olmsted, 2019). 10 Automated news writing, especially for some forms of straightforward reporting such as financial updates or local news (Alim, 2023), has become commonplace, as has, within investigative journalism, reliance on data analysis to provide insights about complex datasets (Guzman and Lewis, 2020; Porlezza, 2023). AI is used to support automated news content curation for targeted audiences. AI tools have been developed to optimize workflows in news production by means of automating routine tasks such as formatting, distribution, and social media dissemination (Wilczek, Haim and Thurman, 2024). Many such AI-driven mechanisms alleviate the burden of repetitive duties, thereby potentially giving content creators greater time and autonomy to concentrate on the more creative aspects of content generation (Baumann, 2021; Caswell, 2023). But the extent to which AI yields positive gains depends on circumstances and, as Simon (2024: 18) argues, if, for example, ‘something produced by AI ends up needing to be laboriously checked by a human, or if its output cannot be fully trusted’ then potentially its use in news production may decrease rather than improve efficiency. The use of AI in news production has triggered a range of concerns, for example in relation to ethics and transparency in its usage and about the ways that AI-based recommendation systems tend to reproduce and amplify biases (Shin et al., 2022). Another major concern is how journalistic content is currently used to train LLMs. Journalists, news providers and other content creators argue that their work is being used unfairly by AI platforms that appropriate the monetization of user data and attention surrounding that content in ways that eventually threaten to put creators out of business (Faroohar, 2024). Some news publishers such as Axel Springer, owner of titles such as Politico, Bild and Business Insider, reached content licensing deals with Big Tech companies over compensation for use of their content to train AI software in 2023 (Thomas and Murgia, 2023). But many remain concerned about lack of fair compensation and ‘appear torn between litigating and negotiating’ about use of their content (Tobitt, 2024). The use of AI within news content production is impacting on cost-efficiency and jobs. As generative AI becomes increasingly integrated within everyday news production, an example of one of the roles in which the need for laborious human input may be diminished is basic copywriting (Murphy, 2024). But in many respects ‘the technology doesn’t supplant human practices so much as it changes the nature of the work’ (Diakopolous, 2019: 1). Although AI is ‘shaking up’ the industry, it is unlikely to replace ‘roles where nuance is needed’ any time soon (Murphy, 2024). As Caswell argues, despite growing reliance on AI, ‘[t]he most valued skill in an AI-empowered news organisation will likely be the same as it has been in traditionally configured news organisations’ – human editorial judgement (Caswell, 2023). 11 Table 1: Role of AI in Different Media Industries and Effects on Jobs Media Industry Role of AI Examples of AI Applications Job Roles Likely to Be Replaced Job Roles Likely to Be Enhanced New Skill Sets Required Film • Analysing and forecasting market and audience trends • Script analysis for predicting box office and streaming success • Economic assessment of talent (cast and crew) • Increasing efficiency and creativity in post-production • Enhancing visual storytelling with AIcreated realistic characters • Automated special effects and postproduction enhancements • Algorithmic content recommendations • Synthetic voice creation/ voice cloning/automatic voice-overs • Automated lip-syncing for dubbed versions • De-aging actors in real-time without further editing by high-resolution, photorealistic face-swaps • Digital character creation and animation • Replacing real animals by life-like AIcreated versions • Script analysts • Some postproduction roles like basic editing • Some performance artists like background staff or voice-narrators • Creative directors • CGI technicians and animators • AI programming and management • Advanced CGI and VFX skills Television • Streamlining content production processes • Automated editing for news shows • Personalized tailoring • Gauging audience engagement and preferences for show development • Audience analysis for targeted advertising • Personalized content recommendations for viewers • AI-driven audience analytics and show ratings. • Analyse protagonist emotions to place appropriate ads • Routine editors (for news, etc.) • Basic content curators • Show producers • Audience analysts • Data analysis and interpretation • AI-driven content management Journalism and News Production • Generating routine news content quickly • AI-assisted real-time news aggregation • Data analysis in investigative journalism/Uncovering patterns and insights in complex datasets • Curating news content for targeted audiences • Automated writing of financial reports and sports updates • Use of generative AI to write local news • Reporters for routine news • Basic editing • Basic factcheckers • Investigative journalists • Data journalists • Skills in AI tools for data mining • Advanced analytical skills 12 Use of AI requires new sorts of professional skills on the part of news creators - see Table 1. While the need for human reporters to conduct routine sorts of news coverage is diminishing, investigative journalists need to acquire new skills, for example in analyzing large datasets that may uncover trends and yield worthwhile stories (Guess, 2024). The news production industry needs journalists and news creators that are proficient not only in news gathering and communication but also in data literacy and use of contemporary AI tools and techniques. As illustrated in Table 1, AI technologies are being used in numerous ways across media industries. While little research has been conducted thus far which quantifies precisely how deployment of AI in the media is effecting employment, costs and/or productivity, the evidence presented in this brief survey confirms how AI is making an impact on all key sectors and stages in supplying media, from creation to editing and distribution, and how it is transforming the resources and skills needed to operate successfully in the digital era. These developments have naturally triggered fears that AI-driven automation may precipitate workforce displacement. A recent IMF study predicts that AI technologies will affect 40 percent of all jobs worldwide, in particular highly-skilled jobs (Cazzaniga et al., 2024). The media industry is no exception (O’Connor, 2022). While circumstances are still evolving, as Table 1 shows the growing use of AI has altered the dynamics of the media work landscape, in some cases displacing traditional (often, but not always, repetitive or formulaic) functions and activities that can now be automated. While this can potentially give rise to labour market frictions and job transition difficulties (Moldenhauer and Londt, 2019), the use of AI is seen by many across the media industry as ‘more of an opportunity than a threat’ (Thomas, 2023a). The consensus view among industry experts is that human creativity is still the crucial driver of media content creation (O’Connor, Grimes and Criddle, 2023). So, rather than obviating the need for human input, AI needs to be integrated appropriately and ethically alongside human creative labour - ‘AI-human collaboration is the key to addressing challenges and seizing opportunities created by generative AI’ (Nah et al., 2023). Uptake of AI typically involves a managed integration of new AI tools within content creation and production processes, supported by investment in new systems, training and recruitment (Bradshaw, 2023). Content creators, editors, producers and organisations are generally keen to adopt relevant AI tools because they recognise the potential of AI to deliver cost savings and/or enhancements in creativity and productivity (Pitel and Storbeck, 2023). 13 5. A Cure for Baumol’s Disease? The use of AI as a tool to support creative work, although not without its problems, evidently facilitates major opportunities to enhance and support production of a variety of media outputs, including content for advertising campaigns, news stories, film, music, publishing and broadcasting. The positive potential of AI to boost economic productivity, including not least across the media industries, is widely acknowledged by policy-makers (DSIT, 2023). At the same time, copyright infringement is a significant and recurrent concern. Many media organisations complain ‘that tech firms have been using their news stories and features to train chatbots without a licensing deal in place’ (Thomas, 2023b). AI offers considerable potential for unauthorised exploitation, distortion, re-mediation and monetisation of professionally crafted original media works, including news content, whether in recycled or fake variants. Even so, AI is increasingly present in all stages of making and supplying media (Murphy, 2024). As the evidence presented in Section 4 shows, the use of AI to automate some of the timeconsuming and repetitive tasks that are integral to content creation and production, from research and pattern identification to creation of mock-ups and tailoring, is increasingly prevalent. As is typical of any efficiency-improving technology, AI is cutting down on the need for human labour to perform certain tasks involved in media production. Integration of AI tools can allow journalists, story developers and other content-creators to offload some of the more time-consuming functions involved in their work, thus freeing up more time for them to engage in editorial creativity, to the benefit of both productivity and content quality (Khan, 2023). So media organisations evidently have much to gain from integrating AI tools judiciously into processes of creating and supplying content. Does this mean that at last Baumol’s cost disease has been cured? Alas not. Certainly, the arrival of AI challenges banishes any notion that all forms of creative labour inputs are irreplaceable. But the central thrust of Baumol’s argument is that because labour-intensive work in creative industries cannot be replicated by computers, and because labour costs tend to rise more quickly than others, these sectors are prone to persistent above-average inflation. We would argue that, even though the use of AI tools has increased significantly amongst media organisations and can yield a myriad of efficiencies and cost-savings, the core creative work involved in content creation remains heavily dependent on human creative labour. Although AI is a powerful tool, its risks and limitations are such that, inevitably, humans need to remain ‘in the loop’ (Zysman and Nitzberg, 2024). Echoing the findings of earlier research which suggests that human-AI collaboration is what drives improvements in productivity (Nah et al., 2023; Sowa, 14 Przegalinska and Ciechanowski, 2021), our analysis of how AI tools are used in media content creation indicates that, rather than substituting for human ingenuity, they typically play a support role. For example, where AI is used in news production it generally requires extensive human intervention to safeguard and ensure the coherence and accuracy of ensuing news content. Likewise, where AI is used to help suggest storylines, plot twists and dialogues for example for television, it nonetheless requires human input and editing by creative scriptwriters and other media professionals to ensure the quality and coherence of the ensuing content. Human creativity remains at the core of content production (Cheng, 2024; Schomer, 2023). This concurs with recent case study-based research findings indicating that the use of AI by creative teams is not replacing human labour (Erickson, forthcoming 2024) and also with Connock’s position that while AI systems provide ‘rocket fuel’ for media businesses they are not, thus far, replacing human artists (Connock, 2023: 8). So creativity within media content-making industries remains inherently and obstinately labour-intensive. As a consequence, Baumol’s proposition that these industries are susceptible to higher than average cost inflation still holds true. That Baumol and Bowen’s ideas about cost inflation in arts and creative industries (1966) still remain of relevance 50+ years on stands as tribute to their pioneering work in economics. Be that as it may, the potential for generative AI to further transform media creation processes and media businesses over coming years should not be under-estimated. As AI is increasingly adopted, key challenges for media businesses include how to protect their intellectual property assets, develop specialist AI skills, balance machine learning with human creativity, and use AI ethically. The agenda for policy-makers is no less daunting, bearing in mind AI-related concerns about misinformation and transparency, copyright and licencing and potential market dominance. Thus, there is little doubt that the research agenda in this area will continue to evolve and flourish over coming years. 15 References Acemoğlu, D and Restrepo, P (2019) Automation and New Tasks: How Technology displaces and reinstates Labor, Journal of Economic Perspectives, 33 (2): 3-30. Alim, A (2023), Daily Mirror publisher explores using ChatGPT to help write local news, Financial Times, 19 February, accessed at: https://www.ft.com/content/4fae2380-d7a7-410c-9eed91fd1411f977. Arntz, M, Gregory, T and Zierahn, U (2016), The Risk of Automation for Jobs in OECD Countries: A Comparative Analysis, OECD Social, Employment and Migration Working Papers, No. 189, OECD Publishing, Paris. Baumann, S (2021), ‘Media Management and Business: Old vs. New’, in Mahoney, M and Tang, T (Eds), The Handbook of Media Management and Business, Lanham: Rowman & Littlefield: pp. 407-425. Baumol, W J and Bowen, W G (1966), Performing Arts, the Economic Dilemma: A Study of Problems Common to Theater, Opera, Music and Dance, New York: Twentieth Century Fund. Baumol, W J (1967), Macroeconomics of unbalanced growth: the anatomy of urban crisis, American Economic Review, 57(3): 415-26. Baumol, W J (1993), Health care, education and the cost disease: a looming crisis for public choice, Public Choice, 77(1): 17-28. Bell, M (2024), A stellar RTS panel presents its predictions for the TV sector for 2024, Television Magazine, 6 February, Accessed at: https://rts.org.uk/article/ash-atalla-claire-enders-evanshapiro-dan-clays-and-patrick-holland-their-2024-tv Benjamin, W (1935), The work of art in the age of mechanical reproduction’ Trans. Harry Zohn in Arendt, H (1968) (Ed), Illuminations: Essays and Reflections. New York: Schocken Books, pp217251. Bielby, W and Bielby, D (1994), All hits are flukes: institutionalized decision-making and the rhetoric of prime-time program development, American Journal of Sociology, 99 (5): 1287-1313. Blaug, M (2001), Where are we now on Cultural Economics? Journal of Economic Surveys, 15(2): 123-143. 16 Bradshaw, T (2023), Gaming industry puts generative AI to the test, 28 June, Ft.com. Accessed at: https://www.ft.com/content/58ddd586-0c06-4d67-9215-54b14b4fc1b1. Caramiaux, B, Lotte, F and Geurts, J (Eds) (2019), AI in the Media and Creative Industries, New European Media. Caswell, D (2023), AI and Journalism: What’s next? Reuters Institute Report, accessed at: https://reutersinstitute.politics.ox.ac.uk/news/ai-and-journalism-whats-next. Caves, R (2000), Creative Industries: Contracts between Art and Commerce, Cambridge, MA: Harvard University Press. Cazzaniga, M, Jaumotte, F, Li, L, Melina, G, Panton, A J, Pizzinelli, C, Rockall, E J and Mendes Tavares, M (2024), Gen-AI: Artificial Intelligence and the Future of Work, Staff Discussion Notes, No. 2024/001, IMF Publications. Chan-Olmsted, S (2019), A Review of Artificial Intelligence Adoptions in the Media Industry, International Journal on Media Management, 21(3-4): 193–215. Cheng, G (2024) Research on the Displacement Impact of Artificial Intelligence on the Film Industry, Highlights in Business, Economics and Management, 28: 48-53. Collins, R, Garnham, N, and Locksley, G (1988), The Economics of Television: The UK Case. London: Sage. Connock, A (2023), Media Management and Artificial Intelligence: Understanding Media Business Models in the Digital Age, Abingdon: Routledge, Cowen, T (1996), Why I do not believe in the cost-disease: Comment on Baumol, Journal of cultural Economics, 20(3): 207-214. Criddle, C, Murgia, M, Thomas, D, Nicolaou, A and Pitel, L (2023), AI and media companies negotiate landmark deals over news content, FT. com, June 17, Accessed at: https://www.ft.com/content/79eb89ce-cea2-4f27-9d87-e8e312c8601d. Department for Science, Innovation & Technology (DSIT) (2023) AI regulation: a pro-innovation approach (Command Paper Number: 815), UK Government White Paper on AI, 29 March, London: DSIT. Diakopoulos, N (2019), Automating the News. How Algorithms are Rewriting the Media, Cambridge, Massachusetts: Harvard University Press, Doyle, G (2013), Understanding Media Economics (Second Edition), London: Sage Publications. 17 Doyle, G (2018), Television and the development of the data economy: Data analysis, power and the public interest, International Journal of Digital Television, 9 (1): 53-68. Doyle, G, Paterson, R and Barr, K (2021) Television Production in Transition: Independence, Scale, Sustainability and the Digital Challenge, Series: Palgrave Global Media Policy and Business. Springer: Cham, Switzerland. Duan, Y, Edwards, J S and Dwivedi, Y K (2019), Artificial intelligence for decision making in the era of Big Data – evolution, challenges and research agenda, International Journal of Information Management, 48: 63–71. Dwyer, P (2015), Theorizing media production: the poverty of political economy. Media, Culture & Society, 37(7): 988–1004. Erickson, K (2024 forthcoming) AI and Work in the Creative Industries: Digital continuity or discontinuity? Farchy, J and Denis, J (2020), ‘Artificial Intelligence’ in Towse, R and Navarrette Hernandez, T (Eds) Handbook of Cultural Economics, (Third Edition), Cheltenham, UK: Edward Elgar, pp 38-46. Faroohar, R (2024), Content Creators fight back against AI, 8 January, Ft.com Accessed at: https://www.ft.com/content/39e74df4-3adc-4e8f-8b85-f8561284e138. Felton, M (1994) Evidence of the Existence of the Cost Disease in the Performing Arts, Journal of Cultural Economics, 18(4): 301-2. Frey, B (1996), Has Baumol’s Cost Disease disappeared in the performing arts? Ricerche Economiche, 50: 173-182. Frey, C and Osborne, M (2013), The Future of Employment. How Susceptible Are Jobs to Computerization? Working Paper, Oxford: Oxford Martin. Garvey, S (2021), The “General Problem Solver” Does Not Exist: Mortimer Taube and the Art of AI Criticism, IEEE Annals of the History of Computing, 43(1): 60–73. Goertzel, T (2014), The path to more general artificial intelligence, Journal of Experimental & Theoretical Artificial Intelligence, 26(3): 343–354. Guess, R (2024), How AI could act as a boost for Investigative Journalism, VOA News, 10 January 2024, Accessed at: https://www.voanews.com/a/how-ai-could-act-as-boost-for- investigative-journalism/7434364.html. 18 Guzman, A L and Lewis, S C (2020), Artificial intelligence and communication: A Human–Machine Communication research agenda, New Media & Society, 22(1): 70–86. Haenlein, M and Kaplan, A (2019), A Brief History of Artificial Intelligence: On the Past, Present, and Future of Artificial Intelligence, California Management Review, 61(4): 5–14. Heilbrun, J (2011) Baumol’s Cost Disease, in Towse, R (Ed), A Handbook of Cultural Economics (Second Edition), Cheltenham: Edward Elgar, pp67-75. Hemphill, J. (2023), Robert Zemeckis’ Latest Tech Move: De-Aging Tom Hanks with AI, Indiewire, accessed at: https://www.indiewire.com/features/general/robert-zemeckis-latest-tech- move-de-aging-tom-hanks-with-ai-1234805275/. Hoskins, C, McFadyen, S, and Finn, A (2004), Media Economics: Applying Economics to New and Traditional Media, Thousand Oaks, CA: Sage. Khan, A (2023), Analysis: How the Sector is Making Use of AI? Broadcast Magazine, 29 June 2023. Klein, J (2019), Perils and Promise in the Modern Media Industry, Brown Journal of World Affairs, 25(2), accessed at: https://search.ebscohost.com/login.aspx?direct=true&db=buh&AN=137810 889&site=ehost-live. Latar, N (2018), Robot Journalism: Can Human Journalism Survive? New Jersey: World Scientific. Leavy, B (2023), Understanding the fundamental economics of AI, Strategy & Leadership, 51(2): 17–23. McCarthy, J, Minsky, M, Rochester, N and Shannon, C (1956), A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence, accessed at: http://jmc.stanford.edu/articles/dartmouth/dartmouth.pdf. McCulloch, W S and Pitts, W (1943), A logical calculus of the ideas immanent in nervous activity, Bulletin of Mathematical Biophysics, 5: 115–137. McKinsey (2018), ‘Creative Disruption: The Impact of Emerging Technologies on the Creative Economy’, World Economic Forum White Paper, London: McKinsey & Company. Moldenhauer, L and Londt, C (2019), Leadership, Artificial Intelligence and the Need to Redefine Future Skills Development, Journal of Leadership, Accountability and Ethics, 16(1). 19 Moran, R and Shaikh, S (2022), Robots in the News and Newsrooms: Unpacking MetaJournalistic Discourse on the Use of Artificial Intelligence in Journalism, Digital Journalism, 10(10): 1756-1774. Murgia, M (2023), ChatGPT maker OpenAI unveils new model GPT-4, Ft.com, March 14, accessed at: https://www.ft.com/content/8bed5cd7-9d1e-4653-8673-f28bb8176385. Murphy, H (2024), Media groups look to AI tools to cut costs and complement storytelling, FT.com, 26 March, Accessed at: https://www.ft.com/content/c581fb74-8d85-4c08-8a46a7c9ef174454. Nah, F., Zheng, R., Cai, J., Siau, K., and Chen, L. (2023), Generative AI And ChatGPT: Applications, Challenges, And AI-human Collaboration, Journal of Information Technology Case and Application Research, 25(2): 1-28. Natale, S (2019), If software is narrative: Joseph Weizenbaum, artificial intelligence and the biographies of ELIZA, New Media & Society, 21(3): 712-728. Nedelkoska, L and Quintini, G (2018), Automation, skills use and training, OECD Social, Employment and Migration Working Papers, No. 202, OECD Publishing, Paris. Nilsson, N J (2009), The quest for artificial intelligence, Cambridge: Cambridge University Press. O’Connor, S (2022), Actors worry that AI is taking centre stage, Ft.com, August 30, accessed at: https://www.ft.com/content/7c26a93f-88ec-4a50-8529-7df81af86208. O’Connor, S., Grimes, C and Criddle, C (2023), AI shakes up way we work in three key industries. Professional services, filmmaking and coding are among the first to use the technology in everyday operations, Ft.com, June 18,Accessed at: https://www.ft.com/content/b34ef958df28-41a1-b30f-dfaf46ff2a80. Picard, R (2011), The Economics and Financing of Media Companies, New York: Fordham University Press. Pitel, L and Storbeck, O (2023), AI is an opportunity for creative businesses, says Bertelsmann boss, FT.com, May 25, 2023, Accessed at https://www.ft.com/content/1a02c92c-b5e6-454da303-44b7db04e306. Porlezza, C (2023), Promoting responsible AI A European perspective on the governance of artificial intelligence in media and journalism, Communications, 48(3): 370–394. 20 Preston, P and Sparviero, S (2009), Creative Inputs as the Cause of Baumol's Cost Disease: The Example of Media Services, Journal of Media Economics, 22 (4): 239-252. Roe, J and Perkins, M (2023), ‘What they’re not telling you about ChatGPT’: exploring the discourse of AI in UK news media headlines, Humanities and Social Sciences Communications, 10(1): 1-9. Rosenblatt, F (1957), The perceptron, a perceiving and recognizing automaton Project Para, Cornell Aeronautical Laboratory. Rubin, R (2023), Tom Hanks and Robin Wright will be digitally de-aged in new Robert Zemeckis movie, Variety, 31 January. Accessed at: https://variety.com/2023/film/news/tom-hanksrobin-wright-digitally-deaged-robert-zemeckis-movie-1235507766/. Russell, S (2021), The history and future of AI, Oxford Review of Economic Policy, 37(3): 509–520. Schomer, A (2023), Will Generative AI Supplant or Supplement Hollywood’s Workforce? Variety, April 17, Accessed at: https://variety.com/2023/tv/news/will-generative-ai-supplant-orsupplement-hollywood-workforce-1235568343/. Shapiro, C and Varian, H (1999), Information Rules: A Strategic Guide to the Network Economy, Boston: Harvard Business School Press. Shin, D, Hameleers, M, Park, Y J, Kim, J N, Trielli, D, Diakopoulos, N, Helberger, N, Lewis, S C, Westlund, O and Baumann, S (2022), Countering Algorithmic Bias and Disinformation and Effectively Harnessing the Power of AI in Media, Journalism & Mass Communication Quarterly, 99(4): 887–907. Simon, F (2024), Artificial intelligence in the news: how AI retools, rationalizes, and reshapes journalism and the public arena, Columbia University: Tow Center for Digital Journalism. Sowa, K., Przegalinska, A and Ciechanowski, L (2021), Cobots in knowledge work: Human – AI collaboration in managerial professions, Journal of Business Research, 125: 135-142. Speechify (2023), AI Voice Generator. Create content with our natural sounding, human-quality AI voice generator, accessed at: https://speechify.com/voiceover/?landing_url=https%3A%2F %2Fspeechify.com%2Fvoiceover%2F. Suich Bass, A (2023), AI will transform every aspect of Hollywood storytelling, But it will also cause more friction, The Economist, November 13, accessed at: 21 https://www.economist.com/the-world-ahead/2023/11/13/ai-will-transform-every-aspect-ofhollywood-storytelling?. The Economist (2023a), Large, creative AI models will transform lives and labour markets, The Economist, April 22, accessed at: https://www.economist.com/interactive/science-andtechnology/2023/04/22/large-creative-ai-models-will-transform-how-we-live-and-work?. The Economist (2023b), AI is stalking the last lions of Hollywood: The first actors to lose their jobs to artificial intelligence are four-legged, The Economist, December 20, accessed at: https://www.economist.com/christmas-specials/2023/12/20/ai-is-stalking-the-last-lions-ofhollywood?. The Economist (2024), AI is transforming the entertainment business: That means opportunities and anxieties, The Economist, January 4, accessed at: https://www.economist.com/films/202 4/01/04/ai-is-transforming-the-entertainment-business?. Thomas, D (2023a), AI use in advertising an ‘opportunity’ not a threat says WPP chief, April 27, FT.com. Accessed at: https://www.ft.com/content/1cf9604d-3568-4588-bb4a-178d6eacada6. Thomas, D (2023b), British media and creative industries quizzed over AI risks to copyright, FT.com, August 21, Accessed at: https://www.ft.com/content/76680815-b313-44a7-b12ba57ebfcb34bd. Thomas, D and Murgia, M (2023) Axel Springer’s OpenAI deal sets new template for media ties with Big Tech, December 15, FT.com accessed at: https://www.ft.com/content/f308c81d-27034f0b-bf87-bf0187379ffa. Throsby, D (2010), The Economics of Cultural Policy, Cambridge: Cambridge University Press. Tiongson, E (1997) Baumol’s Cost Disease Reconsidered, Challenge, 40(6): 117-122. Tobitt, C (2024), Who’s suing AI and who’s signing: Publisher deals vs lawsuits with generative AI companies, Press Gazette, 29 May. Accessed at https://pressgazette.co.uk/platforms/newspublisher-ai-deals-lawsuits-openai-google/. Townsend, S (2024), The rise of AI in film: How AI script writing is changing the game, Medium, 5 March. Accessed at: https://medium.com/@channelasaservice/the-rise-of-ai-in-film-how-aiscript-writing-is-changing-the-game-04509a15c9fe. Towse, R (Ed) (1997), Baumol’s Cost Disease. The Arts and Other Victims. Cheltenham: Edward Elgar. 22 Towse R (2020), Performing Arts, in in Towse, R and Navarrete Hernandez, T (Eds), A Handbook of Cultural Economics (Third Edition), , Cheltenham, UK: Edward Elgar, pp. 415-20. Wilczek, B., Haim, M and Thurman, N. (2024). Transforming the value chain of local journalism with artificial intelligence, AI Magazine, 2024: 1-12. Wormwell, R (2023), The Future of AI in TV, Broadcast, 4 October. Accessed at: https://www.broadcastnow.co.uk/comment/the-future-of-ai-in-tv/5186552.article. Wu, W (2024), Application of Intelligent Algorithms and Big Data Analysis in Film and Television Creation, Scalable Computing: Practice and Experience, 25 (3): 1882-1893. Zysman, J and Nitzberg, M (2024), Generative AI and the Future of Work: Augmentation or Automation? BRIE Working Paper 2024-2. 23 AIACUREF ORBAUMO LSDISEAS EAIACURE FORBAUM OLSDISEA SEAIACUR Centre for Regulation of the Creative Economy School of Law / University of Glasgow 10 The Square Glasgow G12 8QQ www.create.ac.uk 2024/08 DOI: 10.5281/zenodo.13167298 CC BY-SA 4.0 In collaboration with: