A new report published by the BFI suggests that around 130,000 scripts from TV and film have been used for ‘training’ by generative AI companies.
Commissioned by the BFI, a newly published report explores the impact of generative AI on the UK’s film and TV industries. It’s somewhat lengthy at over 60 pages, but there are some eye-catching statistics that emerge from its research.
In a section devoted to the thorny subject of copyright, the report estimates that some 130,000 scripts from TV and film have been stolen – sorry, ‘mined’ – by companies such as OpenAI. This represents a tiny segment of the writing and imagery that tech firms scrape from the internet and feed into their gen-AI platforms. Essentially, vast tranches of human knowledge and creativity are fed into the machine without the creators of that work receiving compensation.
The report, titled AI in the Screen Sector: Perspectives and Paths Forward, was written by an organisation called CoSTAR (Convergent Screen Technologies and performance in Realtime). It emerges as the UK government has made repeated attempts to push through a proposal that will essentially allow tech companies to continue to train their models on copyrighted work without seeking consent from the owner or paying them for its use.
There has been vocal opposition to the proposals, not least from filmmaker and peer Baroness Beeban Kidron, who’s led a bloc in the House of Lords that has voted down the government’s proposals a total of four times to date.
The BFI and CoSTAR’s report echoes the concerns eloquently put by Baroness Kidron and other critics: that gen-AI actively competes with flesh-and-blood writers and other creatives, and that the companies behind it are feeding off entire industries without putting financial benefits back into them.
“The existing training paradigm for generative AI models poses a threat to the fundamental economics of the screen sector and its ability to create value from making and commercialising new IP,” one section of the report reads. “Sources of AI training data include scripts from more than 130,000 films and TV shows, YouTube videos, and databases of pirated books. As generative models learn the structure and language of screen storytelling – from text, images and video – they can then replicate those structures and create new outputs at a fraction of the cost and expense of the original works.”
In response, the report recommends that the UK should position itself as a “world-leading IP licensing market”, with the UK’s Copyright Licencing Agency acting as a broker between British creators and the gigantic tech companies that want to use their work for training.
Interestingly, the report also chimes with the House of Commons’ own findings. In April, the government’s Commons Culture, Media and Sport Committee published a report which found that “the rapid growth of generative AI technologies threatens their earnings and future employment opportunities.”
Time will tell whether Keir Starmer’s government will stop and listen to the growing number alarm bells ringing over its existing proposals. Whatever happens, the future promised by gen-AI has such wide-ranging and profound implications that you can feel CoSTAR’s report struggling to get to grips with them all. There’s the environmental impact to consider; the societal implications of mass unemployment and how those people pay their bills; the ethical questions over stealing artists’ work; the question of how AI is affecting education, and perhaps even how our brains work in generations to come.
Whatever happens, the use of gen-AI is evidently taking root to some degree in the UK’s creative institutions. According to CoSTAR, “Organisations including the BBC, BBFC, and BFI are actively experimenting with generative AI, demonstrating sector-wide curiosity and openness.”
Suggested product
SPECIAL BUNDLE! Film Stories issue 54 PLUS signed Alien On Stage Blu-ray pre-order!
£29.99
