Quantcast
Channel: Cadence Blogs
Viewing all articles
Browse latest Browse all 6698

There's More Voice in Your Future

$
0
0
Rich Kennewick gave the keynote on the second day of the Linley IoT Hardware Conference recently. He is the CEO of a company called Voicebox and his keynote was Enabling Voice in the IoT . I had not heard of Voicebox and that is because they don't sell voice-enabled devices themselves. They work with companies such as Toyota and Samsung to add voice capabilities to their products. The most well-known voice enabled systems are Google, Apple's Siri, Microsoft's Cortana. And the elephant of them all, Amazon's Alexa. At the Consumer Electronics Show in January, Alexa in particular was everywhere (see my post CES Highlights ). Voice Interfaces Despite that fact that I—and probably you—haven't heard of them, Voicebox has been around for a long time. They first started shipping in 2008 with Toyota. Not a bad logo to have as your first customer! They support 23 languages, have shipped 300M devices with their voice solution, of which 20M are cars. Rich believes that voice will be the future user interface and that in many areas the screen will go away. It has taken longer than he expected to get to where we are now but now voice is in 100s of millions of devices and tens of millions of cars, and in many languages. He reckons that user interfaces have gone through two eras, point and click on the PC, followed by touch and slide on the smartphone. The future belongs to voice. The trends are towards more rules and statistics, and using that to work out what is being talked about. For example, "Traffic" is a movie. It's a 70s rock band (quite a good one, imho). Plus it is obviously road conditions. But it can depend what is on your screen or what you just talked about as to which context is the most likely. If you say "turn it off" then obviously, just as when you are talking to another person, the meaning of "it" will depend on what you have been doing and talking about. Context has to fade with time, though, just as it does in a real conversation. One prediction is that as the systems get better and people get more confident in using them, they don't want to hear so much, especially if it is not something irreversible. I can confirm this with my experience with Alexa, which I have set up to control several of my lights. "Alexa, turn off the kitchen light." Light goes out. About a second later, Alexa say "Okay." The light went off, I don't think I need verbal confirmation that you got the message and did what I asked. Another thing to be managed is that context needs to fade over time. In fact, verbosity is something that needs to adapt, with different levels for beginners. The system should keep track of how many mistakes you, the user, are making and adapt. Far field voice recognition is hard. That's why even the little Amazon Echo Dot has seven microphones. The noise context changes all the time with doors being open or closed, refrigerators and air-conditioners being on or off. They actually go to noisy environments like shopping malls and collect noise, since they need to aim the acoustic models they build at the environment in which the device will be used. One challenge is how big a model to use. Smaller models, such as song titles, are more accurate. Another context is who the user is. It is one of the weaknesses of Alexa that they don't distinguish the user and change the context. As Alexa gets more integrated with other applications in the background, this will become an important problem to fix. Your daughter probably doesn't want to share her text messages with you. It will become important for voice systems to distinguish between you, your daughter, your wife and strangers. Voice can be used for authentication, especially if other things like facial recognition are added. This is within the family, if it gets to things like ATMs you will need more secure biometrics. But the current Voicebox technology can get you to say three sentences and know who you are. Rich has three brothers, with similar voices/accents, and so it is a difficult test to tell them apart, so they use that for testing. Strategic Trends 2017 is the year of voice. Obviously Rich is a bit biased (he probably said 2015 and 2016 were too!) but Gartner has it as one of the top 10 strategic technology trends this year. Over 20% of Google searches are now by voice, with predictions that it will be 50% by 2020. It is already higher in China than the US, and they started later, perhaps because typing Chinese on a keyboard is a pain. Voice-first devices will increase dramatically. There are already over 1B handsets with voice—they are not voice first but voice is important. Voice is going to have a big impact on brands. The tech giants are destroying brand. If you Google some product, you don't get answers listed in order of brand awareness. Voice is even more restrictive. Rich demonstrated asking Alexa to order batteries, and it ordered (high-margin) Amazon own-brand batteries, and when further queried claimed that it couldn't find any others. Bad luck Duracell and Energizer (by the way, another brand mess-up is that when asked, more people associate the bunny that keeps going and going with Duracell than with Energizer!). Voice eliminates any advantage from packaging and design since you don't even get to see it. The challenge for a lot of companies going forward is that they will need to add voice (and natural language understanding) to their products, even though they know nothing about it. That's obviously Voicebox's business. One thing that they offer that the big guys do not is that the customer (Toyota, say) owns the data. If you build Alexa into your car, Amazon will know a lot of valuable things. Another interesting dynamic in the industry is that Apple's Siri is nowhere since it is only available on Apple's own products, whereas Alexa is not just available in the Amazon boxes but it is almost the voice system of choice in all sorts of other devices. It is probably not a winner-take-all market, but there probably aren't going to be dozens of winners in the space, just from an investment and scale point of view. But for sure, there is going to be more voice in your future. Sign up for Sunday Brunch, the weekly Breakfast Bytes email.

Viewing all articles
Browse latest Browse all 6698

Trending Articles