Publication
Research for CULT Committee. The use of Artificial Intelligence in the Audiovisual Sector
Georg Rehm
European Parliament, Policy Department for Structural and Cohesion Policies, 5/2020.
Abstract
This briefing paper takes a look at the use of AI technologies in the wider audiovisual sector. A survey with ten questions concerning the most important aspects was circulated to 85 contacts at 73 organisations. A total of 22 responses were received. The main findings are as follows: 1. Almost all respondents report broad use of AI technologies, especially for automated indexing, improved content accessibility as well as localisation. AI is used for processing audio or video, language or text data or for knowledge management purposes. 2. Among the technologies used are ASR, TTS, NLP, NER, MT, summarisation, search and recommender engines, content classification, subtitling, vision and metadata extraction (see Appendix 2: Glossary – Terms and Abbreviations). 3. AI technologies foreseen for future use are more experimental and include the automated detection of illegal content and deep fakes as well as flexible curation technologies. 4. There is a big demand for large amounts of training data including labelled, structured and unstructured data, domain-specific training data, acoustic data and data for illegal content. 5. There is also a need for more language technologies for all European languages, including ASR, TTS, MT, content curation services and metadata extraction as well as Linked Data. 6. In terms of policies, it is suggested to focus upon an ethical framework regarding the use and misuse of AI that protects human values and fosters cultural and linguistic diversity. It should also protect against the misuse of AI for false news and misinformation. 7. Regarding opportunities, many respondents suggest concentrating on the AI-based production of high quality content. In addition, AI allows unlimited localisation and makes it possible for a fragmented and culturally diverse ecosystem to survive in a world dominated by capital-intensive ventures based in the US. 8. The consumption of intentionally created false or manipulative content is seen as an imminent danger. It is stressed that, as video is quickly becoming our main means of communication, there is a threat that relates to the use of AI for misinformation and manipulation, which could have an impact on the foundations of our democratic society. 9. The awareness of the European AI tool market varies. Some perceive the market to be non-existent, others perceive it to be highly fragmented. Due to the dominance of non-European technology enterprises, European companies should be supported more. 10. Collaboration at the European level is seen as essential because individual players have limitations and difficulties in using AI technologies. Europe’s multilingualism is seen as crucial: to guarantee inclusiveness and accessibility, tools need to be made available, especially for under-resourced languages.