Everything is one, everything is connected: The coffee machine with the light switch of the heater with the door lock, and with the Internet.

In the house of Tomorrow, a voice command of the residents, to the bath water to let in, a gesture to control the TV. When you do, what you watch, what are the wishes you Express, and commands you to give – all of which are valuable information for companies like Amazon and Google. These exploit the data and make a billion. With its voice assistant, Alexa and co. you come to us so close as never.

It is therefore hardly surprising that around half of the language assistants, who already hold today in the house, come from these two companies. The technology market analysts Canalys, the total number of such devices is expected to increase worldwide from the current 100 million to more than 300 million by the year 2022. Almost three-quarters of them will be equipped with Amazon’s or Google’s technology, appreciate the industry experts.

privacy advocates prepare these Figures. Because as soon as the everyday wizard will be activated by means of a keyword (such as “Alexa” or “Ok Google”), to transfer unfiltered recordings, including background noises to their respective Cloud – centralized, often distant server structures with huge storage capacities. What happens there with the data, get the users.

“Nobody wants all the things that happen on-site, one-to-one transfer somewhere and be stored there,” warns Thomas Bendig, research coordinator at the Fraunhofer group for information and communication technology in Berlin. It is absolutely necessary that such information remained with the user.

In the Cloud the data would be used, too, in order to improve the machine learning Algorithms, “in order that Alexa can better understand and react,” according to Amazon upon request. And: “We allow our customers to talk about data recovery for training purposes in the Alexa settings.” To a third party, there is no voice data to be passed on. In December it came after a report in the magazine “c’t” is a breakdown: on request, sent to files a customer found him totally foreign records, the group spoke of an “unfortunate event” as a result of “human error”.

There are already ways to leave the Material of the users. A method of Edge – Computing-computing power at the edge (Edge) of the Cloud. The data are collected in a decentralised manner close to the user and processed. “This may apply, for example, the normal Internet router, we are in the house”, stated Andreas Seitz, a researcher at the chair for applied software engineering of the technical University of Munich – even if commercially available routers were not yet able to do so.

it is Possible, however, that such processes run directly on the Hardware on which the language assistants are installed. A company that goes this route, is the Start-up Snips based in Paris and New York. Snips provides language assistance systems for business customers, which they can build into their devices. The privacy of the users have top priority, stress the responsible.

“It just feels wrong that a group has access to a microphone that is in their own home,” says Snips-co-founder, edge of Hindi. The memory and processing power to the machine understanding voice commands, necessary, in the case of Snips, therefore, not on distant server farms, but on the devices themselves, to the business customers update to the speech recognition of the Start-ups. Voice and background noise remain in place – without that someone can unintentionally have access to it.

“There are a number of companies and research projects that pursue such approaches,” says Bendig. Because it’s not just for the home use processing of data, beyond the Cloud is attractive. Also, the industry has long since become aware of.

in addition to the protection of information, speed plays a role. In the case of smart machines, computational processes have to run often in real time, said Wolfgang Furtner, who is responsible for Edge Computing at the Munich-based semiconductor manufacturer Infineon. “The bandwidth and response time of the Cloud is but very limited, if any device in the industry have their data there to process.”

much faster it will go if you pay at the counter, and the machines would be played – especially if the Cloud Internet connection is fast enough or incomplete. About Edge devices could be data bundles and anonymous, before you go in the Cloud. Infineon supplies the Hardware.

Will replace Edge Computing to the Cloud so soon? No, says Seitz from the technical University of Munich. “Edge Computing is an extension of the Cloud, both worlds are connected to each other.” Especially in the case of Artificial intelligence or machine Learning, you need huge computing power for processing huge amounts of data. “These things are in an Edge to perform and to learn, is unrealistic,” says Seitz.

information filtering, pooling, and only the bare Essentials to the Cloud-send – the can Edge devices, nevertheless, afford, the experts agree. For privacy, it would be a progress.

dpa