2 minute read

As the field of artificial intelligence advances, health care providers express concerns about whether computers will at some point outperform us. Although machine learning and particularly deep learning has produced some impressive achievements, the way that the human brain develops and acquires knowledge cannot at this point be replicated. This does not ignore the fact that artificial intelligence has already outperformed human experts in some areas. For example, properly trained convolutional neural networks can identify hard-to-see features in X-rays; recurrent neural networks can rapidly translate text from one language to another with reasonable accuracy; and reinforcement algorithms can teach cars to drive themselves, often more safely than their human equivalents (Aggarwal, 2018).

            However, machine learning requires vast amounts of data to train the network. Imagine a health care provider asking to see the records of 100,000 patients to diagnose a garden variety case of pneumonia. Yet 100,000 medical records provide too little data for some deep learning programs to train properly (Topol, 2019). Unlike computers, human brains can draw on evolution and personal experience to make decisions. As writer and neurobiologist, Peter Robin Hiesinger explains, we are our history (2021). As the brain develops throughout the lifetime, factors such as genetic coding, evolution, experience, and self-organization enable human beings to think both intuitively and analytically, making ‘good-enough’ decisions with minimal information and in some cases, emergent situations.

            Computers are not good multi-taskers. A convolutional neural network designed for image segmentation and feature extraction cannot teach a robot to walk. Although a machine learning program called AlphaGo beat the world champion at the Chinese game of go (a game like but more complex than chess), that AI system would be hard put to decide on the best treatment for a patient with autism, or the best rehabilitation strategy for an individual with chronic low back pain. In other words, machine learning algorithms have a narrow scope of practice.

            Finally, while chatbots such as Siri can engage in human-type conversation, computers lack the ability to connect with human beings on a deeper, emotional level. The unique way the human brain develops via a synthesis of evolution, environment and self-organization gives us the ability to intuitively sense the emotional states of others and respond appropriately. Empathy and compassion, both uniquely human attributes, are powerful medicine. While artificial intelligence can transform medicine with its diagnostic and predictive capabilities, the ability to heal, heart to heart and touch to touch, is something the human brain does best.

References

Aggarwal, C. (2018). Neural Networks and Deep Learning: A Textbook. Springer Nature. http://www.springer.com

Hiesinger, P. (2021). The Self-Assembling Brain: How Neural Networks Grow Smarter. Princeton University Press. http://www.princeton.edu

Topol, E. (2019). Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again. Basic Books (Hachette Book Group). http://www.basicbooks.com