There was a sense of excitement and inspiration among the 170 participants of the AI in Production and Distribution workshop – organized by the EBU AI and Data Initiative and hosted by BBC in Salford/Manchester, 27-28 November – a mood that reflected the enormous potential for media innovation with AI and ML that shone through in the presentations and demonstrations.
The conference discussed how public service media are embracing AI and Machine Learning (ML) as enablers for new services, increased productivity and reduced cost, and posed one overarching question: how can public service broadcasters mitigate real and perceived issues related to AI? A line-up of expert presenters discussed ethical questions, ideas for governance, strategic questions around the use of commoditized AI/ML services, the cost of bringing expertise in-house, as well as technical trends and innovations in the field.
EBU Members can access the video recordings and PDF files of all presentations (after logging in) on the 'AI in Production and Distribution' event website.
A packed agenda
Day 1 of the conference put the focus on audience engagement, analytics and ‘ethical AI’, with the BBC, RTS, NPO, YLE, FTV, and Imperial College of London each reflecting on how AI – and algorithms in general – can be put in the service of media PSMs values. One very tangible example was the use of recommendation systems for better audience engagement, and the question of how to keep humans at the centre of relevant decision making. A panel of experts discussed this and other issues, such as unintended bias in AI and ML algorithms. The conference raised the notion that, in order to make AI ethical, it and its decisions need to be explainable.
Another key topic was the practical challenge and complexities of bringing the power of deep neural networks to end-user devices. The conference discussed the use of so-called convolutional neural networks as a promising avenue, and edge computing as a way to keep processing close to the data while reducing bandwidth and data-sharing issues.
Day 2 then continued with a deeper dive into case-studies of applications and relevant technologies, with Al Jazeera’s Grant Franklin Totten giving the keynote and an overview of how that global media organization puts AI in the service of its journalists to improve value, relevance and reach.
France’s ‘Institut national de l'audiovisuel’ (INA), one of the designated guardians of French cultural heritage, presented their work on using AI to detect gender bias in content, and NGO ‘First Draft News’ on the use of AI for truth and trust. Other presentations dove into topics such as language services, video compression, AI-aided cinematography, lip-reading and animal recognition in the wild.