In response to increased interest in, and concern around, the use of artificial intelligence (AI) in different audiovisual fields, the Flanders Audiovisual Fund (VAF) has just released a new report that examined data about the usage of AI in all requests for support in the creation of audiovisual projects. In collaboration with Howest DAE Research and legal experts on intellectual property, the VAF used questionnaires to collect the requisite data from the start of October 2023, and the results were analysed in July 2024.

(L'article continue plus bas - Inf. publicitaire)

The study intended to look at what types of AI tools are used, why and when they are used, as well as legal questions around intellectual property. The VAF determined that during the study period, the use of artificial intelligence was reported as relatively low across all categories of project, including fiction, animation, documentary, games and podcasts. However, perhaps the most prominent and sweeping find is qualitative: a continued vagueness around how AI is defined in a final audiovisual product in addition to lack of knowledge around AI and its legislation, leading to unsure responses from a large volume of stakeholders.

For the purposes of the questionnaire, 10 different categories of AI were defined: generative art (such as pictures and images through AI models like Dall-E or AI-enabled Photoshop tools), video and visual effects, 3D images, virtual humans, animations, text, audio, testing (such as automating simulations and machine learning bots), physics and engine optimisations (including the simulation of film sets and camera positioning) and others.

The survey was conducted across three different stages: the initial request for support, the first payment request and the last payment request. For surveys completed in the last stage, only 22,2% of respondents indicated that AI was eventually used in the project, down from 28% of respondents who initially stated they intended to use AI. The largest category of AI used was text generation, which dominated over all other forms of AI through all stages, followed by generative art in stages one and two.

Among the interesting trends that emerged was that more funding applicants for series and podcasts indicated that they intended to use AI than their film counterparts, something the VAF was unable to explain based on their data. Respondents indicated different reasons for using AI, including improving the quality of their project, cutting costs or achieving something that would not have been possible without AI due to time limits or to it being physically impossible. Those against the use of AI tended to argue that the tools available were “less innovative and predominantly serve to complete work quickly and at a cheaper cost”. In fewer cases, AI was specifically used as the subject of projects including both fiction and documentary works.

When asked about the justified use of AI, questionnaire respondents supported its usage in administration and support, prototyping and pitching, translation and transcribing as well as inspiration and research. 15% of respondents noted that they thought AI should never be used in works. Three large categories emerged when participants were asked about the top reasons for AI non-use, including concerns around using it to infringe upon intellectual property and privacy, AI as used to directly replace writers and artists, and potential ethical dubiousness and malicious intent.

The report notes with caution that the results merely present a snapshot, and the VAF intends to continue the survey for five years in order to collate more comprehensive data and observe trends. The organisation further recommends a follow-up qualitative study to better study some of the trends that have emerged from this initial report.