The next frontier: the Brain and AI working in tandem

My father taught me to program a computer in the early seventies. I was awed by the fact that by writing “highly constrained prose” you could instruct the machine to do exactly what you’ve written. The essence of most computer code is executing rules: “if…then..else”. For example if “temperature > 40 then sound alarm”. In large programs there can be millions of lines of code, with many interdependent rules and complex data manipulations. Once a program is tested and ready to run, bugs will be fixed and functional changes will be made. The latter is a necessity because businesses and technologies constantly change. This can become cumbersome as code complexity and vulnerability tends to increase over time. Many systems have become so complex that no single person can understand the inner workings. This can lead to disaster as we have recently seen with the Boeing 737 Max flight control. So instead of getting better, traditional code has a tendency to get worse over time.

AI takes a fundamentally different approach. Many AI’s essentially detect patterns and probabilities in large data-sets that can be used for insights about what drives certain outcomes and it can become predictive on future outcomes. The more (quality) data you feed the AI, the better it tends to become over time. It has cognitive abilities. This is why AI is so impactful: it’s the first self-learning technology that can provide insight in complex, emerging systems at scale.

AI is increasingly used to predict behavior of machines, such as MRI scanners, plane engines or cars that have so many moving parts and instruction-sets that physical models no longer suffice. As health drivers are interdependent -clinical, socio-economic and behavioral factors at play and more than 70% of health data is unstructured (images, notes)- this holds a great promise for the application of AI. Not only to support more precise diagnosis and personalized treatment, but also to optimize the highly dynamic acute care.

As AI will have such a dramatic impact on society, the economy and even humanity, I thought it would be interesting to “pick the brains” of the next generation of AI scientists at Philips. These are the young professionals that will largely shape where we take the technology in the next decennia. Following our conversations it became apparent to me just how important it is to engage deeply with our next generation of tech leaders. Our discussion zoomed in on the connection between neural networks found in our brain and the neural networks that underpin AI. The dialogue then turned to how this relates to healthcare.

For many AI researchers, the ultimate goal is to emulate the capabilities of the brain at scale, to achieve true cognitive autonomy in a sense. More and more AI research and implementation algorithms inspired by the brain are being brought to the fore, and with solid results. In 2017, the DeepMind team published a paper about neuroscience-inspired AI that explores influence AI and neuroscience research. The creation of neural networks is fascinating: researchers are quite literally using the brain to create blueprints for AI innovation. Neuroscience has two key benefits for AI. Firstly, neuroscience provides a rich source of inspiration for new types of AI methods. Secondly, neuroscience can provide validation of AI techniques that already exist. So, for example, if a known algorithm is also found to be coded into the brain, then that is strong support that it plays an integral part in the composition of a cognitive system.

AI’s are already capable of performing expert tasks and augmenting humans. You only have to think about intelligent systems’ capacity to find cancerous lesions on an image, detect biomarkers in blood, interpret a person’s risk for disease in DNA, analyze and quantify physician notes, and optimize patient flow in emergency care. However in some basic areas AI still requires further development: understanding time, space and causality.

There’s an interesting next step in the evolution: AI and the brain working in tandem. We discussed an article that looks at how brain signals are being translated into speech using artificial intelligence. The amazing progress in biomechtronics, where AI is used to control and literally feel artificial limbs. Developments like this could potentially provide a voice for people who are unable to speak or limbs for people who can not walk, blurring the lines between man and machine.

AI is increasingly augmenting humans in their daily work. Advances in AI technology can be credited to myriad, converging technologies. But it’s also the combination of design, science, engineering and business expertise, working in multi-disciplinary teams that creates a hotbed for innovation.

Not only AI is benefitting from neuroscience’s foundations. It’s becoming symbiotic and reciprocal. For example, AI, and predominately machine-learning, has totally revolutionized aspects of neuroscience. I’m proud that Philips is a leading player in this space, we’re at the forefront of transforming tools that analyze MRIs and EEG’s, using physiological, molecular and behavioral data that populate brain models, to make more effective and efficient diagnoses while at the same time developing new treatment pathways with better efficacy. Obviously the brain is the most complex organ and in our pursuit to create “digital twins”, we are building on the work we are doing for heart and lung models.

While there is still a long way to go in creating an accurate model of the brain and deploying truly artificial cognitive systems, major strides are being made. If I go on the expertise and ambition of our young scientists, I see steady progress on the horizon in the application of AI in diagnosis and treatment of cancer and the diseases of the heart and lungs, with neurological disorders and behavioral sciences as the next frontier.