Ever since the arrival of the Smart Speaker and its Voice Activated friends, there have been those voicing concern about the perceived security of such devices.
The idea of an electronic stranger listening and monitoring everything that is said in the home is, understandably, disconcerting for some. It is also perfectly natural that with a new technology comes with a certain amount of mistrust from its adopters. When it comes to Voice Assistant’s however, it could be that attempts to allay consumer fears over security could result in Smart Speakers being less… well… smart.
In the past few weeks, the state of California put forward legislation that would require the likes of Amazon and Google to get a users permission, in writing, before their Smart Speaker devices can store voice recordings. It would also prevent manufacturers from sharing any recordings they DO make with 3rd parties (in Amazon’s case the developers who are creating their Voice Apps or Skills).
Jordan Cunningham, who authored the “Anti-Evesdropping Act” said:
“Today, the State Assembly sent a strong message to the tech giants who have spent years recording and retaining private conversations in the home via smart devices.”
If the bill passes into law whilst it may well provide some reassurances to those worrying that tech giants are spying on their everyday conversations, at the same time it could greatly hinder the development of this technology.
Google, Amazon and Apple all rely heavily on the data, utterances and conversations that they record and process to improve the way their technology interacts with the human voice. Without this data, progression in this area could slow or even stall.
Amazon has already taken their own initiative to combat these concerns, and maybe placate the lawmakers, by announcing that it “soon” intends to allow users of the Alexa platform to delete everything recorded by their device on a daily basis (nicknamed the ‘Amnesia Command’). Although, it remains to be seen if that step will be far enough, especially when you consider that Alexa listens, records and communicates with the “Cloud” every few seconds in order to identify, receive and process commands (although only legitimate commands are stored). This step is in addition to the existing layers of security and privacy protections.
Security IS important for ANY device that is part of the IOT. But, being overly cautious in this area could seriously stifle the development of the platform and create huge barriers to its useful application.
These type of Voice Recognition systems rely heavily on machine learning. The more you communicate with them the more they can recognise speech patterns, vocabulary, nuance, slang. This happens not just in general terms but in some cases on a user-by-user basis.
Processing that huge volume of data doesn’t happen in the little tube on your desk either. Recorded audio must be uploaded and analysed elsewhere, by machines and humans. If we limit the data that we are sending to be processed we are also limiting the information from which the machine can learn and thus limiting the machines ability to understand, and even mimic, human communication.
One of the concerns raised by Cunningham was that such devices can “wake” with erroneous commands with users accidentally prompting a response from their device. The irony is that such early bugs will be impossible to fix without the very data that the act wishes to deny access too.
Voice recognition is already showing it’s benefits to the user. From streamlining daily activities to communicating for those who struggle to communicate the applications for the technology are forever expanding.
It’s going to come down to a simple choice. Embrace the new technology and potential invasion of privacy that comes with it (as we have done with previously Social Media and Smart Phones) or reject the idea of a future full of frictionless voice activation and interconnected homes.
The two may not be compatible.