This time of year is good for forecasts. Some of them are disconcerting, because they raise fears that we’d rather leave asleep. That’s what I thought when I read an article about the widespread recording of all our conversations. That’s the stuff of George Orwell, you say, and that’s what I thought until I looked more closely, or should I say listened more closely…. Then I realized we’re in a new era where the aim is not just to capture everything we say with the aid of mass storage. The challenge is to get the machines to understand the content of these recordings using more and more complex search algorithms and analysis.

Anyone interested in intelligence (artificial or not) knows that speech recognition is a key component of it. After half a century of small jumps of progress in this domain, the field of study has grown, in area and depth. Today, the biggest advances are being made in deep neural networks, notably in what are called “recurrent neural networks”, essential for learning, and for machines to learn man’s spoken language.

It’s not surprising that this research is carried out mainly by Google, Apple, Amazon and Microsoft, because the machines that can hear us talking are not only our telephones and computers, but everything connected to what’s now called the Internet of Things, which might one day be called the Internet of Ears.

A French version of James Somers’ article is carried in the January edition of “Courier International”. It raises an interesting question: if this is happening, how is this general recording of what we say going to change the way we speak?