Fernando Fischmann

Why Innovation Must Go Beyond Disruption

5 March, 2015 / Articles
Fernando Fischmann

Henry Ford famously quipped that if he’d asked what people wanted, they’d have said, “faster horses.” There are countless numbers of ideas being funded every day that are aimed at essentially building faster horses. The result is that we have available an enormous embarrassment of riches in technology, information and economy – but how many of them are truly groundbreaking or innovative?

The real breakthroughs happen when we venture outside of convention and learn to look at problems a different way. The real ideas come when we make links where no one else has, a theory put forth by James Webb Young in his little known yet seminal 1939 book, “A Technique for Producing Ideas.”

Young believed that all ideas are simply a new combination of old elements – that by putting together these old elements in a new way, we create new ideas. In other words, ideas don’t just pop into our head. They are the result of a process that, as Webb himself said, “is just as definite as the production of Fords.”

It goes beyond the concept of disruptive innovation made popular by Clayton Christensen. True innovation springs from a combination of a deep understanding of customers’ needs and a willingness to approach a problem from a different angle – connecting the dots where no one else has.

Steve Jobs understood this process intuitively – “creativity is just connecting things” was one of his most often quoted lines – and it’s why he’s frequently held up as the master of innovation.

Jobs and his team at Apple understood the needs and wants of the marketplace. The seamless integration of hardware and OS, the ecosystem of application developers, the emphasis on design and the “attainable exclusivity” and brand cachet (iconic white ear buds?) all served to carve a space for this innovator in a place that was already too full of competitors.

Reed Hastings at Netflix also understood this need. From its streaming media to its all-you-can-watch model, Netflix pioneered online digital delivery of content, enabling users to watch movies anytime, anywhere and on any device they could connect to the Internet. Netflix blurred the lines between content distribution and content creation by offering high quality original programming.

None of this fit into the competitive model that Blockbuster was offering at the time. Unlike that bygone company, Netflix figured out a way to create a richer, easier experience for consumers that consumers had no idea they wanted.

At Cypher, we followed this model to figure out a way to improve cell phone call quality. We have seen our smartphones make enormous advances – year after year, cell phone makers pile on the features, enhance processing time and update media capabilities – yet little has been done to improve sound quality, an important element of phone communication that has been left behind. Given our increasing reliance on our mobile phones and willingness to abandon our landlines, I believe better sound quality is a feature customers both want and need.

Traditional approaches to poor mobile phone sound quality are principally acoustic and focus on the noise – something we knew wasn’t working. We started with a small team of developers and an idea that the real way to address the need in this space was not noise suppression but voice isolation. We believed that the sad state of the art was a result of an obsession to hunt and kill an ever-expanding portfolio of noise types. Essentially, there were too many different noise inputs for anyone to catalog and counter. We needed a way to connect the dots, and thought tracking human speech might be a way to do it. So we set about using this alternative approach to guide our research.

There is a substantial body of pattern-matching and applied math from related fields that we used to examine the problem. We took a very large database of speech and built an artificial intelligence engine to analyze these recordings. By doing this, we developed a set of abstract descriptors to identify human speech. We used these characteristics in our algorithm to “tag” speech elements. We did not really care about anything else, such as background noise.

By using this knowledge, we imbued our software with the ability to recognize the primary speaker and sift it out of any other noise. It was like the old problem of looking for a needle in the haystack. Everyone else’s attempts to build faster hay-sorting machines had met with limited success. But by asking the fundamental question of what makes a needle different from the hay we designed a magnet that pulls the needle out quickly and cleanly.

After two years of turning theory into code, our first deployable software solution blocks out over 99 percent of the background noise while improving the quality of the speaker by an average of 20 percent and improving speech recognition by 25 percent using industry-standard tests.

As phones and people became free to make and take calls from virtually anywhere, the difficulty of controlling the background noise is literally impossible. It’s also impossible to effectively cancel an infinite variety of noise types. That left us with the option of isolating the voice rather than trying to cancel the noise. To paraphrase Sir Arthur Conan Doyle, by eliminating the impossible, the remaining, however improbable, was our path to success.

This was a great lesson to us – and should be to anyone else faced with the challenge of whether to build a faster horse or give consumers something they really need but may not have realized they were missing.




Te puede interesar