IFA 2018 Recap: It’s All On the Edge

Voice, AI, and the Future of Content

Last month, Snips’ VP of Sales Stephen Bauer spoke on a panel during the IFA technology conference in Berlin about how trends in content creation are changing the way consumers and businesses design their products and services.

Stephen was joined by a group of experts in a discussion lead by Nokia’s Global Head of Marketing, Paramita Bhattacharya. Below you’ll find some fascinating insights and key trends identified by the panel.

The conversation started with a look at the dramatic shift in storytelling that has taken shape during the last few years. Of course, Augmented Reality (AR) and Virtual Reality (VR) offer compelling use cases for entertainment, even just normal video has become far more engaging — content creators are using more and more microphones in content capture to develop richness in depth of audio. Despite increasing equipment availability, audio is still under-utilized considering how far it has advanced. For example, audio technology can now maintain a richness in quality despite noisy or windy environments. The panel pinpointed the trigger leading audio and video quality improvements as mobile — standard phones can now detect distinct speakers and provide noise cancelling benefits, along with their bigger screens and dual or triple cameras for depth perception.

A Move Away from Component-by-Component Product Design

At Snips, we do quite a bit of work with hardware platforms and audio- processing providers. As recently as a few years ago, all of these platforms for content creation functioned on a clunky component-by-component basis. While certain OEMs offered 4K video, audio still lagged; other companies focused on audio, but not necessarily video. But recently, these disparate use cases have begun to converge.

As a voice ecosystem, we are dependent on the quality of audio that is recognized and understood by our software. We look for white-noise reduction that matches the human brain, but with a machine filtering out the audio that isn’t directly in the context field. In other words, we need technology to be able to focus on individual speakers and sounds, using AI and the capabilities of multi-mics for depth of audio.

This convergence is driving an economy based on solutions as opposed to components. Many companies are taking note, offering bundled hardware+software+audio front end.

For Snips, this means that the “massive back-end effort” to drive solutions is intended to dramatically benefit user experience and functionality. By reducing the “component” nature of audio processing and filtering, companies can reduce friction with their products.

As Stephen mentioned on stage: “By getting rid of screens and menus and buttons, the intent is to make a more natural experience and to make the component aspect of both content creation and content consumption vanish. That way, we’re free (and the end user is free) to create the content they want without even thinking about connecting one app with another app to create a video.”

But within this solution framework rests the idea of edge computing.

Making Technology Invisible on the Edge

The steady progress towards immersive solution-based experiences means that technology is becoming more natural and less intrusive in our daily lives.

We are now also entering a new normal where the “invisibility” of technology is more common for end users. Examples of this “invisible” technology include software that enhances image clarity, audio quality, noise cancellation, bezel-less mobile devices, etc.

All this opens up the role for AI.

As Jyri Huopaniemi, Head of Product and Technology Strategy for Nokia, explained, “speech recognition and voice interfaces have dramatically improved over the past five to ten years with the advent of artificial intelligence.” [[Co-existence of edge and cloud solutions — many AI algorithms require vast amounts of data to run in real time. ]]

Snips is making major advances around voice computing power on the edge as costs diminish. As Stephen described, “Computing power on the edge is skyrocketing versus the cost of producing it, which allows us at Snips to run sophisticated, use-case specialized voice models directly on device, almost preloading the intelligence based on a user’s context.”

He went on to say that “we can now get to local command and control with natural language, thanks to the ability to digest copious amounts of data in the cloud, but then grab that collective intelligence and pre-load it onto a device. This results in providing low latency, offline capabilities, and privacy. This is the mix of what we at Snips to recognize the user.”

On this note, Jyri Huopaniemi chimed in to illustrate that “the device is almost disappearing, because you’re dealing with a large set of edge computing nodes. That allows for low latency and real-time computing.”

Paul Melin, VP of Digital Media at Nokia, added: “Most consumers don’t even want to hear about the complexity and technology inherent within their devices. You have to make it simple and just demonstrate that you get real ease-of-use and better output in a visible manner, regardless of 4K or other buzzwords.” Companies need to make the benefits clear and the technology invisible.

Due to the compute power on the edge, content generation is already exploding. And this compute power allows us to do extremely interesting things on-device, like embedding AI algorithms and neural networks that can perform fantastic context aware voice recognition.

Yet, as Stephen warned, “We need to continue the conversation around what happens to the massive amounts of data generated by end users on the devices and products they use. The amount of data that’s collected everywhere we go without us being conscious of it is mind boggling. Who is the custodian of that data? The OEM? The consumer who owns the OEM product? GDPR may not be perfect, with a clunky user experience, but we need to continue the public discussion about what happens to that data as our physical and digital lives melt into one another.”

If you liked this article and want to support Snips, please share it!

Follow us on Twitter: @Snips

If you want to work on AI + Privacy, check our Jobs Page