Intel boffins usher era of 'I know who you are' TV
AI team forecasts remotes that control your viewing
Now this is scary: a TV remote control that knows who you are and can present what your prefer - or limit what you can watch - accordingly.
It might also tell the telly to tailor ads to whoever's handling the remote.
The gadget is the brainchild of a band of boffins at Intel Labs. While the group hasn't actually made one of these things, it has come up with an algorithm that allows a remote to guess who it's being used by.
Key to the scheme is to attempt "non-invasive" identification - working out who is holding the remote by means other than, say, photographing the user or demanding he or she log in to the device.
To make this work, you need an algorithm that lets the remote 'learn' the traits of its users - what channels they select, how they select them, how they move the remote and so on - and use these data to maintain a mental image of users which, in turn, allows the remote to predict who it's being used by at any given time.
For the system to work, the predictions have to be reasonably accurate, and that's what the team has managed to show.
The team captured user data using remotes modified with three-axis accelerometers which were supplied to five households for three weeks. A receiver by the TV tracked what buttons users pressed, and all the data was logged in real time.
The results show the team's system can track users with a 70-90 per cent accuracy, though in general it proved to be closer to the lower end of that range.
Formally identifying individuals - which allows usage patterns to be mapped to them - would help the prediction accuracy. While this might have once required "invasive" identification, the team say, the rise of internet-connected TVs, which present online content that individuals have to log in to access, could give their system the extra cues it needs to be more certain that, say, it's dad using the remote rather than junior.
If the remote is sufficiently certain that that's the case it might, for example, allow the user to switch to an adult channel and ignore the kids' channels, or ensure that the next ad break presents selling messages from Gilette and co.
You can read the team's paper here (PDF). ®
Sponsored: Hyper-scale data management