Photo by — @jankolar

Listening For or With AI

Big Technology and Test Data Without Clear Consent

To enable voice technology to work you need training data. In some cases it means you need a lot of labeled training data from a variety of different places, especially if you are operating as a global company who wants to understand a variety of people. Increasing with the growing complexity of voice controlled interactions, and seeing new ones that actually to some degree work such as Amazon and Siri, there has been a growing consciousness about the rush to the best technology in voice control. There are great aspects to this and some more questionable aspects, I will mention one of each.

Voice Control — the Great

A use case close to my heart is for Parkinson. I could see is for my father who has Parkinson, for him it will get increasingly difficult to interact with technology. If for example in addition to this there are specialised versions of Amazon or Google Home that understand different type of voices I could see this being an enabler for my dad.

Photo by — @stefano_intintoli

Voice Control — the Questionable

Most large technology companies have been capturing and listening to audio from users’ devices.

“In April, Bloomberg reported that Amazon was employing thousands of people to listen to what users say to its Alexa-powered speakers. Then, in July, Belgian broadcaster VRT obtained more than 1,000 audio recordings captured by Google, and the company admitted that the recordings underwent human review.Last month, the Guardian revealed that Apple captured what its users were telling Siri, and reviewed those recordings to grade Siri’s responses.”
VICE 14th of August 2019

Voice assistants offer utility and convenience, but can come with trade-offs. It is likely that the voice messages is categorised to improve the product, yes, however it is done in all likelihood to make it easier to sell you products.

This is no great surprise, we are talking about the most commercial technology companies in the world. However not asking permission and doing it in secrecy is kind of a double problem in this regard.

Photo by — @nicolasjleclercq


This is #500daysofAI and you are reading article 178. I write one new article about or related to artificial intelligence every day for 500 days.

If you have a collection with empirical examples on the use of AI from different locations around the world I would very much like to redo this article. I have found a few, but I am looking for more.

AI Policy and Ethics at Student at University of Copenhagen MSc in Social Data Science. All views are my own.

AI Policy and Ethics at Student at University of Copenhagen MSc in Social Data Science. All views are my own.