Decorte’s Mayor of London GROW Summit speech in full

For more pictures of the GROW summit, see below. For more info regarding the GROW summit, click here.

INTRODUCTION

Hello hello, thanks to the organisers for inviting me to speak today. When I received the brief, I was asked to speak on how AI will transform healthcare in the 21st century.

That’s a pretty big topic for 4 minutes.

But the good thing about that big a topic is that it gives me the opportunity to talk about a key concept I believe is vital to understanding how AI will impact the future of healthcare. One that is often overlooked, and which I call “natural data saturation”.

NATURAL DATA SATURATION

Grab any Machine Learning or AI engineer from the street and ask them “what do you need”, and before even thinking about food, drink or money, they’ll answer “data”. Any founder, CTO or techie here will be familiar with that constant need. Datasets, their size, quality, complexity: they are what makes AI powerful. In most industries, such data is generated artificially, through human action or provoked by research.

Now grab a doctor from their consultation room, and ask them what they need, and before anything else, they’ll answer “time”. If you try to offer them data – here’s thousands of datapoints detailing your patient’s physiology over the last week – chances are they’ll run away screaming.

Healthcare is one of the few areas where valuable data is not just abundant, it is overwhelming. Our bodies – of each and every one of us sitting in this room – are constantly outputting highly valuable datapoints, every single second. And despite their tremendous value, due simply to the complexity of them, none of these data points are currently being collected in any systematic way.

That is because humanity’s traditional healthcare systems, without AI, have always had only one way to make sense of this wealth of naturally generated data: by being extremely selective. Our current healthcare system is curative, rather than preventative (it focuses on problems after they present, rather than predict and intervene as they develop), and is punctuated, rather than continuous (we only collect data at specific moments when we enter the doctor’s office, and essentially ignore it at all other times).

WEAKNESS TO STRENGTH

That this system, being the best humanity could do for millennia, is deeply flawed, is the reason why I started my company four years ago. When I was 11, my father nearly died due to a congenital heart condition that wasn’t picked up on for 39 years of his life. Had anyone listened to his heart for longer than 30 seconds, over the period of 39 years, his and my life would have been tremendously different. But no one did. To this day he cannot work or function normally. Think about that for a moment: 39 years, and no one listened to his heart for longer than 30 seconds. And his story is one of hundreds of millions in the developed world, let alone in areas that don’t have easy access to healthcare.

In the age of AI, however, what used to be humanity’s greatest problem in healthcare – natural data saturation – will, I believe, become its greatest strength. The very nature of Machine Learning is that it enables extremely powerful pattern recognition when offered extremely large datasets.

THE BRIDGE

So why hasn’t AI already impacted the personal health of each and every one of us here?

The problem no one has been able to solve until now, is that, in order to capture the many different data-types our bodies are constantly emitting, current technology requires individual sensors and medical equipment for each metric: ECG sensors for our hearts, respiratory sensors for our breathing, imaging for abdominal health, and so on. Commercially available wearables, meanwhile, use the same single-sensor/single-metric technologies, but simply less accurate versions than those available in hospitals.

While we know AI will thrive on large datasets, and we know that this will unlock preventative and predictive medicine, short of sitting in a hospital connected to lots of machines our entire lives, there is currently no cost-effective or user-friendly way to connect the human body into the digital world.

Our company has solved this problem. To provide an accessible data link between our bodies and AI. And the answer lay in AI itself. We are a team of mostly Cambridge PhDs as well as engineers from around the world, and we went back to one of the oldest forms of medicine – sound-based analysis, listening to the sounds of the human body – using AI to automatise and massively enhance it (a process called automatic auscultation). Inspired by academic research, our ongoing clinical trials in the US, EU and India will be the first globally to prove in a commercial setting that we can extract complex medical data, including cardiovascular, respiratory, mental and physical health signs, with a level of accuracy that normally requires specialised equipment, all from a single sensor. A single sensor that, moreover, is sitting in each of our pockets right now: a microphone.

Over the last two years, we have been building this sound-based AI engine, and we have now partnered with hardware and software developers like Google and Microsoft, to scale this technology to billions of devices, and gather medical data non-invasively, remotely, continuously, and, due to the low cost of microphones, in an extremely accessible way. In this way, we are finally building this continuous link between the human body and digital world – enabling an internet of bodies. A first tool for AI to fully transform the health of each and every one of us.

CONCLUSION

Let’s end on a scary but also (for me at least) motivating note. I assume many in this room are over the age of 35. In our current system, 1 in 3 of those over 35 will die in the next few decades of a – most likely preventable – cardiovascular condition. But not if my company succeeds in its mission. That is what we do at Decorte. These are the ways in which AI can transform healthcare.

Thank you.