"God does not need a teacher."

Senior Director Uwe Hermann reflects on a year gone by, and looks forward to Eriksholm's 2018 ambitions. This year, Artificial Intelligence (AI) plays an important part.

In October 2017 the German weekly magazine ”Der Spiegel” came with a provoking, maybe even alarming headline in its science section, titled “God does not need a teacher”.

The author elaborated on the fact that many of the real revolutionary inventions of mankind, like the first telephone, went unnoticed for many years and, only in retrospect, maybe many decades later, it became apparent that this was when a technical revolution started.

In this sense, the article argues that 2017 might not be remembered for the various stupidities of some world leaders, but for being the year when “it happened”: the first singularity. “AlphaGo Zero”, the Artificial Intelligence (AI) developed by Google’s daughter company Deepmind, learnedhow to play GO, in just three days, without human intervention. The method behind that feat is called “enforcement learning”. Essentially, the system played against itself and thereby learned the game.

Ke Jie, the world’s best human GO player commented that AlphaGo Zero is playing like a “GO-god”, rather than another human player. So indeed this “god” did not need a trainer.It learned all by itself.

How does this tie into our work at Eriksholm Research Centre? – Well, our mission is to make audiological discoveries with the potential to provide significantly enhanced end-user benefits in future hearing care. To this end, we pull on a global network of expert scientists around the world in our endeavour to probe into unknown scientific territory, where we have expectations of potential use for the benefit of the hearing impaired. Artificial Intelligence is one such area in which we have been working for years. One of the promises is to find even better solutions based on DNNs (Deep Neural Networks) e.g. for solving the proverbial “cocktail party challenge”, which means to filter the wanted acoustic signal out of a babble of voices and noise.

Another AI application is eHealth and personalized solutions, where we optimise e.g. hearing instrument settings for the personal preferences of the users. The global progress in AI research has been a boost to our research, so - in my personal book - 2017 became the year when I needed no longer explain why AI is relevant to the future of hearing healthcare. It became self-evident, like bursting through the sound barrier.

Looking forward to 2018, and to the continuous close and valuable collaboration with experts from all over the world, I expect continued rapid developments in this AI field in conjunction with Brain-Research, Bio-feedback, Sensor-Fusion, eHealth and many other disciplines at Eriksholm.

So, stay tuned! We will have many more exciting stories to tell in 2018!