Tag: voice recognition

2469: OK Google

0469_001

With the courier work I’ve been doing for the past few days, I’ve been relying heavily on Google Maps for navigation around the area, and I’ve been discovering the benefits of voice controls — it’s much easier to simply say “take me to…” and Google work it out for you than to type in a postcode using Android’s cumbersome and clumsy keyboard.

I’ve actually been pretty impressed with the accuracy of the voice recognition, since it even recognises non-standard words such as street names without too much difficulty, and it uses your location to make an educated guess at which one of the many Alder Roads in the world you might have actually wanted to go to. I counted only two hiccups in an entire day’s work: one when it wanted to send me to Hedge End (which is the other side of the Southampton conurbation to where I was working) and one when it wanted to send me to Birmingham. Granted, one of those mistakes was pretty large, but given that it understood me on all the 50+ other occasions throughout the day, I think I can forgive it.

I find myself wondering if voice recognition will actually become particularly widespread or accepted. Apple now includes Siri with Mac OS as well as iOS, Microsoft has Cortana in more recent revisions of its operating systems, Google seems keen to bake voice recognition into Android and all its services and even my TV will let you talk to it. The technology is certainly there and seems to work reasonably well in most cases — certainly considerably better than it did even just a few short years ago — but it’s still painfully awkward to use, particularly if you’re in an environment where there are other people around you. And while I’ve seen the benefit of being able to shout at my phone while I’m in my car, I don’t see the same benefit from talking to my computer, TV or games console when its physical controls are right there and allow me to complete the task I want to complete just as quickly “manually”.

I think we’re still lacking a certain degree of artificial intelligence necessary to make voice activated technology truly useful, worthwhile and ingrained in society. The aim, presumably, is to have something along the lines of Computer in Star Trek, where you can say pretty much anything to the voice activated computer and it will successfully parse what you say (within reason) and perform any task from turning the lights on to inverting the phased magnetic resonance coils into a Gaussian feedback loop. Specify parameters.

I wonder whether that’s something that is truly desirable, though. Is it really more convenient to be able to vocalise something you want your computer to do? It probably is for those who aren’t as computer-literate, but then there’s still a chunk of the population who don’t use computers or mobile phones at all. A shrinking chunk, admittedly, but a chunk nonetheless, and I’m not sure fully voice-capable hardware — which will probably still be on the expensive end of the spectrum — will convert that sort of person into being a believer in technology.

Still. “OK Google” helped me find my way around today, and that, at least, impressed me. Perhaps I’ll discover more interesting uses of it in the future.

1847: Your TV Is Not Trying to Kill You

So another outlandish “privacy scandal” looked set to erupt on Twitter earlier. For the benefit of anyone who might be considering sharing anything regarding Samsung Smart TVs sending your personal information to third parties, allow me to clarify a few things.

Samsung Smart TVs have a voice recognition feature. I know this because I have one. (I also never use it, because voice recognition is, for the most part, stupid and pointless when you have a remote control right there. Assuming you have hands, it is pretty much always just as quick to use the remote as it is to remember exactly how you’re supposed to phrase a voice command.)

Anyway. The way this voice control works is very simple. You press a button on the “special” remote, not the “normal” one, and the microphone in the remote starts picking up your voice. When you’ve finished speaking, it sends what you said over the Internet to a speech recognition service (that more than likely converts the speech into computer-friendly text for more accurate processing) and then your TV receives an instruction based on what you said. The TV itself isn’t doing any real processing; that all happens remotely, and the TV simply receives the instruction to do something based on what the speech recognition service thinks you said.

Astute iPhone-owning readers will know that this is exactly how Siri on Apple devices works — it’s why you can’t use Siri when you don’t have an Internet connection, even to access information stored locally on your phone such as your address book and suchlike.

The reason these services work like this is to take some of the processing workload off the phone/TV/other device with voice recognition. It’s not an ideal solution, but it does mean that the devices in question can be less expensive because they don’t need hefty processing power or software to recognise voices pre-installed on them. One day we may have devices that can recognise our voices accurately without requiring an Internet connection — although chances are by the time we’ve perfected that, the Internet will be “everywhere”, rather than just in Wi-Fi hotspots and mobile coverage areas — but until then, this is how voice recognition tends to work.

As such, a necessary part of the entire process involves sending a recording of what you said to the third-party speech recognition service. This means that if you press the microphone button on your Smart TV remote and then decide that the appropriate thing to say at that moment would be “My credit card number is…”, a recording of you saying your credit card number will be sent to this speech recognition service. Chances are, nothing will happen with it, but as with any sort of unencrypted information transmitted across the Internet, there’s a slim risk of nefarious types intercepting the transmission and taking advantage of it.

Because of this slim risk of stupid people telling their TV remote what their credit card number is, Samsung have had to put a disclaimer in their Smart TV documentation that the TV may send your personal information to a third party, and of course, people have misinterpreted this as the TV always listening to what you’re saying, and it therefore being unsafe to share any personal information while within earshot of your TV. This is, of course, utter nonsense, because as I’ve outlined above, you have to specifically press a button in order to activate voice recognition mode, and the “third party” it’s being sent to is doing nothing more than converting your babblings into something the computer in the TV can recognise as an instruction to do something.

That is it. Nothing more. Nothing sinister. And if you’re still uneasy, you could 1) not buy a Smart TV, since technology clearly terrifies you, 2) not use the voice recognition function (which, in my experience, is patchy, slow and pointless anyway) or 3) not talk about credit card numbers or other personal information when you’ve pressed the button that specifically asks your TV to listen to you.

So there you go. This has been a public service announcement. I thank you.