Date:5 October 2017
Amid new phones and a suite of smart speakers, Google announced a clever little camera at its hardware event today. The Google Clips is designed to sit on your counter, or your end table, or your desk, and points in a general direction where something interesting might happen. When its AI senses a picture, it takes the shot. But it’s best feature is what it doesn’t do—it doesn’t run your data over the internet automatically.
By Eric Limer
The advent of advanced machine learning opens up all kinds of technological superpowers. Google Photos can recognise faces. Smart speakers everywhere can recognise voices.
Typically, these tricks require the internet. Because machine learning takes a lot of computing horsepower, the heavy lifting is done by a powerful server off in a data centre located who knows where. This is supercharged hardware that is too big and power-hungry and expensive to put in every phone.
Keeping Google Clips secure
This necessarily creates some security risks. Everyone who who handles your data in transit (like your ISP) will have a chance to look at it. Even if the data is encrypted and unintelligible—as the recordings from all major smart speakers are—it can be tracked or saved for later.
Perhaps worst of all, the servers where this work is done make a juicy target for hackers. They are accordingly well-protected, but their security is out of your hands. That’s to say nothing of government pressure on tech companies to hand over information, often without a warrant. These are risks you run when you upload a picture to Google Photos for automatic optimisation, for example.
The Google Clips camera does its AI work within the camera itself. It’s an impressive technical feat made possible by new AI-processing chips, and which offers a few handy bonuses.
First, it can work even if you don’t have an internet connection. Second, anything you do not actively choose to expose to the internet will never leave the camera. It’s a distinct and personal sort of security. You don’t have to take Google’s word that its infrastructure is secure, or your ISP’s word that it is not snooping on your data, or the state’s word that it is not slurping up all the data it can find—you know, just in case it needs that information later. Instead, you can know that the data never left your house. It’s security that doesn’t rely on trust.
This refreshingly analog sort of security is more vital than ever in an increasingly online world. No doubt it is showing up in large part because small, cheap, powerful AI-processing chips enable it.
Always listening, always watching
But it’s hard to deny that the always-listening, always-watching vision of the future has been appearing more dystopian by the day, especially in the face of government surveillance and seemingly constant data breaches. It’s getting harder to sell a gadget that ignores these real security concerns.
Hopefully features like this—as well as things like physical camera shutters and mute buttons that electronically disconnect the microphones they control—will become more prevalent as time goes on. Sure, companies like Google will always thrive on your data in a way that can still be uncomfortable, using your photos to fine-tune its own technology or your searches as a means of targeting advertising. But security that acknowledges the fact that you might not trust everyone out there in cyberspace and accommodates that very fair suspicion is a welcome bit of good news.
From: PM USA