“A good application of artificial intelligence is a bit like a very good pasta,” said Ryan den Rooijen, head of the global data and analytics team at technology company Dyson. “The secret isn’t in the complexity of the recipe. It’s in the quality of the ingredients.”
He compared fine ingredients to high quality data as a way of saying, "good stuff in means good stuff out. It’s the same with good analytics. Somebody puts in the groundwork to ensure that the data quality is as high as possible, ensuring all the relevant training data exists, making sure the labels are as accurate and descriptive as possible,” he said.
“Artificial intelligence is a very natural evolution of big data analytics.”
Den Rooijen is excited by artificial intelligence because, for him, it is “a very natural evolution of big data analytics.” The mission of his team is to enable Dyson to use data to provide a better experience to its customers, but also to ensure that the organisation makes more sensible decisions. He said that big data and analytics supports that mission as it “allow humans to make better and more intelligent decisions.”
For him, there are several advantages of AI. One is the fact that AI techniques can draw upon a much “broader space,” which means he and his team can capture much greater nuances and have a much higher degree of accuracy than humans alone.
Another advantage of AI is that it can easily draw on data from a variety of sources, a task that would be exceedingly difficult for a human being. “If you’re trying to map out and make decisions, based on disparate sources of information or completely different data types, it’s a little bit tricky,” he said.
Using the medical field as an example, den Rooijen explained that doctors are asked to make incredibly difficult choices based on what they have learnt from experience, case notes, academic papers, text books, imaging data and, perhaps, audio recordings. However, for a machine with the appropriate algorithm, it’s relatively simple.
"Humans are not so good at non-linearity."
Another advantage is that humans bring cognitive biases into their decision-making processes. “Humans are not so good at non-linearity,” he said. “Our brains simply aren’t evolutionarily wired to be really good at this kind of stuff. The nice thing is we know this from years of cognitive research, so we can offset these biases in the types of solutions we implement.”
One of the lessons that he has learnt from working with AI is that one has to be very specific, so to others intending to implement AI solutions, he advised asking the right questions at the start of the process.
Den Rooijen remembered being told by executives to “just deploy some machine learning” and “make it better”. Den Rooijen responded by asking what was the specific question they were trying to answer. “If you’re training a neural network and you don’t have a very good idea of what you want to get out of it, it’s very easy to get lost and it’s much harder then to interrogate the data,” he said.
“Giving teams permission to innovate and to improve themselves is critical.”
Culture is also of high importance. “Giving teams permission to innovate and to improve themselves is critical to ensuring that you make money off your initiatives because it means that the quality of the initiatives will be much higher.”
Finally, den Rooijen said it is imperative to consider context and communication when dealing with bots and applications. When dealing with AI, one must make sure that one understands how messages or communications will be received. He gave an example of a Strava bot timestamping his 8am gym workout - pulled from his activity tracker - with Pacific Standard Time. This pushed out a notification to his dad and five other Europe-based Strava followers that den Rooijen engaged in 50 minutes of strenuous activity at a few minutes past midnight. To this, his dad responded: “I didn’t need to know that..”
Ryan den Rooijen was speaking at AI Congress London.