Artificial intelligence (AI) influences which films you watch on Netflix. It warns your bank that your credit card number might be in the hands of a fraudster and makes alarmingly accurate predictions about what you are going to type into your browser next. AI will eventually take the danger (and the fun) out of driving. Some say it will help humans live forever, others say it will destroy humankind.
Hype and hyperbole surround all things AI, so as these technologies reach into just about every aspect of our lives, including healthcare, it is important to come to terms with what is possible, and where the dangers lie.
Reality vs. fiction
AI techniques and theories have been around since post-World War II, but in the last several years, the perfect storm of big-data generation, cloud-based storage and processing, and refinement of algorithms and techniques, has come together to create an AI boom.
“AI is not one technology,” wrote Nathan Benaich of the Berlin-based venture capital firm Point Nine Capital in a post on Medium. “It is in fact a broad field constituted of many disciplines, ranging from robotics to machine learning. The ultimate goal of AI, most of us affirm, is to build machines capable of performing tasks and cognitive functions that are otherwise only within the scope of human intelligence. In order to get there, machines must be able to learn these capabilities automatically instead of having each of them explicitly programmed end-to-end.”
Doomsayers predict that machines capable of learning will eventually acquire the power to dominate us, but the AI that exists today is limited to well-defined tasks.
Current AI systems are basically prediction machines. Increasingly sophisticated algorithms and machine learning techniques parse reams of data on a particular issue and generate insights, predictions and diagnoses with more efficiency than teams of humans ever could.
AI for better health outcomes
In healthcare, where better predictions clearly lead to better outcomes, AI technologies are already producing benefits and optimizing processes.
Take drug discovery. Major pharmaceutical companies are forming partnerships with AI specialists and developing techniques in-house so that they can speed up research and better determine which therapies and treatments are most likely to improve patients’ lives. The most advanced methods are being designed to take genetic information into account.
“If the proponents of these techniques are right, AI and machine learning will usher in an era of quicker, cheaper and more-effective drug discovery,” a recent article in Nature points out.
AI will also have a major impact on drug manufacturing. The factory of the future, which Sanofi is currently implementing, will include connected and intelligent equipment, with sensors capable of taking thousands of measurements throughout the production process and generating billions of data points used to monitor, analyze and control the manufacturing process. State-of-the-art analytical techniques will predict and prevent variations and ensure the quality of biological medicines.
AI and machine learning are also contributing to the development of next-generation vaccines, accelerating the development of medicines for conditions where there are no viable options today.
Wearable technology, combined with machine learning and AI, also have the potential to revolutionize how we offer solutions to those with health problems, including sleep disorders.
AI systems are also helping with basic patient care in parts of the world like rural China and Africa, where there are shortages of healthcare professionals. Using data and an advanced technology, machines can help healthcare professionals put patients in touch with the right kind of doctor, or help doctors make diagnoses and determine treatment remotely.
“In Algeria, we’ve been using digital internet technology to reach remote patients and remote healthcare professionals,” explained Jon Fairest, Head of Africa Region, Sanofi. “This way, innovation and digital technology is helping improve early disease recognition, disease management programs, and also empowering patients to look after themselves better when they have a chronic disease such as diabetes.”
Around the world, healthcare organizations are using robots that assist and help train nurses, machines that help doctors operate on patients remotely and even electronic “pets” that improve palliative care and provide companionship, along with monitoring.
Humans + AI = progress
As in other fields, the rapid progress that AI is fostering in healthcare raises concerns. Health data must be handled carefully, transparently and fairly, so that efficiency does not come at the expense of trust. Some AI systems reach decisions without the ability to audit or explain the process through which the conclusion was reached. And remote or robotic patient care, taken to extremes, would be isolating and damaging to patient morale.
The question of jobs inevitably arises as well: Will radiologists be replaced by the latest image recognition technology? Will machines outperform less experienced researchers, undermining their career development?
Fortunately, the robot apocalypse is not upon us. Technology is still a tool, created and controlled by humans. Jobs will be created to address the concerns around data collection and processing. People will be needed to retrain researchers and other professionals so that they can use the new AI- and data-assisted techniques. Freeing healthcare professionals from some of the repetitive manual tasks will enable them to apply their intelligence to more complex problems.
And there are human qualities that machines will probably never be able to match, such as the ability to see how patients’ personal circumstances - and their inherent strengths and weaknesses - will affect different treatment paths. Personal warmth, empathy and bedside manner are not yet in the skillset of even the most intelligent machines.
But that is not to say that it will never happen. So-called “strong” AI would give machines the ability to learn, adapt to different environments, and apply intelligence to any problem or set of problems. They don’t exist, but most experts agree that humankind will have to prepare for such an eventuality.
Getting one step closer to humanizing AI are startups such as CareAngel, an AI and voice-powered virtual nurse assistant and one of the winners at our startup competition at the international tech hub, VivaTech, this year. Founder and Chief Angel, Wolf Shlagman from California described his innovation as “using technology to extend the human touch. A lot of things we do every day are robotic and mundane, so CareAngel is just taking the robot out of the human.” His motivation to work with AI was his own mother. “I needed to be more proactive in taking care of her and this is the next stage in telemedicine, it’s not just technology but how you talk to your loved ones,” he said.
For now, though, humans still decide how to collect and structure the datasets that feed the machines.
“There’s more and more data available, and the power of AI to recognize patterns helps us understand things that we otherwise couldn’t have done,” said Ameet Nathwani, MD, Executive Vice President, Chief Medical Officer at Sanofi. “And we’re only beginning to learn how to apply it usefully to many parts of life. In health, where there’s so much information - genetic information, proteomic information, or the study of proteins, clinical data, social data - it’s allowing us to see patterns and gain insights into outcomes for patients that we couldn’t have dreamt about. AI will fundamentally change how we look at disease and health. The medical future in 10 years’ time will be totally different because of AI.”