The aim is to map how the technology spreads and develops and identify and address some key democratic and political questions.
Speech technology has come of age
By the help of artificial intelligence (AI), machines are rapidly improving their abilities to speak, recognise and understand human language. They are also getting better at deducting information about us, such as age, height, and chronical diseases, based on how we speak.
As a result, we now use our voices for several new purposes. From coffee makers to large industrial machines – you may control them through your voice. Life becomes easier, and society more inclusive, but the development also increases the amount of voice data in circulation in society. What we say and how we say it may reveal sensitive information about us, which might be misused and commercialised for profit. The large tech companies today dominate the market for speech technology and invest heavily to develop it.
Small languages face particular challenges
Shall speech technology become a force for good, it must understand all languages and avoid discrimination. Today, it works best with large languages, such as English, although other languages are slowly catching up. Succeeding with speech technology on a large scale, requires sufficient access to voice data and computational resources.
Important questions remain
We zoom in on some of the most pressing political issues:
- How to secure access to voice data for developers, while protecting privacy?
- How to make sure voice data retrieved from open sources are not misused?
- How to train AI needed for speech technology, while keeping a low carbon footprint?
Project plan
The NBT will start by publishing a policy brief on the topic of speech technology and voice analysis. Later, the project will expand and examine opportunities and challenges of large language models (LLM).
