AI in the UK: ready, willing and able? Contents

Appendix 7: Note of Committee visit to BBC Blue Room: Monday 20 November 2017

The Select Committee on Artificial Intelligence visited the BBC Blue Room on 20 November to see the work the BBC was doing in relation to artificial intelligence.

Six members of the Committee were in attendance,648 as was Dr Mateja Jamnik, Specialist Adviser to the Committee.

Matthew Postgate, Chief Technology and Product Officer at the BBC, and his colleagues, began by providing an overview of recent developments in artificial intelligence, and some of the challenges the BBC face in this area. Matthew noted that recruiting skilled people was a perennial challenge, with only around 10,000 people in the world estimated to be capable of programming neural networks. He also discussed the challenges posed by AI developments elsewhere in the world, especially in the USA and China, and why it was important that Europe consider its own cultural and ethical values, and develop its own AI systems which reflect these values.

The Committee was shown a number of technical demonstrations of what other media companies were doing with AI and machine learning. This included Netflix’s efforts to hyper-personalise the shows they present to their customers, right down to the way in which particular shows were presented. They also demonstrated an AI service, Lyrebird, which could replicate individual voices, using a small amount of data.

The discussion moved on to what the BBC was doing with AI. The Committee was told of experiments using machine learning for end-of-show credit detection, which the BBC hoped would reduce the amount of work that humans needed to put into a time-consuming and tedious activity. The BBC was also exploring the use of AI to augment the work of human camera operators, with a system, trained on years of BBC archival footage, which could begin to select appropriate shots of certain kinds of shows. It was anticipated that, with time, these kinds of ‘AI directors’ could increase the productivity of the BBC considerably, allowing it to cover more events, especially on a regional or local level.

Finally, there was a short discussion about related issues, including how to address bias in datasets, the BBC’s Data Science Research Partnership, which connects the BBC with universities, and the BBC’s policies on user data. The BBC’s approach to engaging with the public on AI was also discussed, and the BBC’s efforts to lead public debate, while also providing AI-powered services which embodied an ethical approach to AI, were highlighted.

648 Lord Clement-Jones (Chairman), Baroness Grender, Lord Holmes of Richmond, Lord Levene of Portsoken, Lord Puttnam and Baroness Rock.

© Parliamentary copyright 2018