Peers2peers as House of Lords digital regulation chiefs visit Cambridge
A quartet of Cambridge technology companies have welcomed the House of Lords as part of an inquiry into the work of digital regulators by the Communications and Digital Select Committee.
The aim of the inquiry is to investigate the effectiveness of digital regulation, building on the Committee’s previous report Regulating in a digital world, which was published in March 2019.
In ‘Regulating in a digital world’, the Committee found that regulators had failed to keep pace with advances in digital technologies. There are over a dozen regulators with a remit covering the digital world and this fragmentation has led to both gaps and overlaps in regulation. New regulation was too often driven by responding to newspaper headlines rather than strategic thinking, the committee felt.
Sound technology pioneer Audio Analytic, serial innovator Microsoft Research and Machine Learning specialists Invenia Labs and Myrtle.ai were all visited by the committee.
Lord Gilbert of Panteg, Chair, Communications and Digital Select Committee, said: “Visiting Cambridge, Audio Analytic and the other companies provided a fascinating insight into emerging technologies and their uses. It will be very useful for our work on digital regulation, on which we will produce a report next month.”
Dr Chris Mitchell, CEO and founder of Audio Analytic talked about the company and its sound recognition technology and discussed its best practice approach to data collection and management.
“Audio Analytic has built Alexandria™, the world’s largest audio dataset for machine learning with over 30 million labelled recordings, across 1,000 classes and featuring 400 million metadata points.”
There was also a tour of the company’s Sound Labs which include a large semi-anechoic chamber and a fully anechoic chamber. These chambers are designed to absorb reflections of sound and insulate from external noise sources, acting like a green screen for audio recording, which is essential for data collection and augmentation.
Dr Mitchell said: “Machine learning is heavily dependent on data. It needs to be realistic, relevant and diverse in order to build reliable AI systems but in order to avoid the serious legal and reputational risks involved, the data you collect and manage has to be sourced correctly.”
“Regulators around the world are starting to ask questions about where data comes from and how it is used for AI systems. As the global regulatory landscape around data evolves, I expect that regulators will demand traceability and transparency, which is what we are prepared for.
“We were delighted to host the House of Lords committee and to talk about data and artificial intelligence and how we’ve managed to build large datasets that comply with regulatory requirements.”
In April 2021, Audio Analytic published a white paper on the subject of audio data management for machine learning, which highlighted the issues and set out a number of principles which the company follow.