AI is coming - we are ready
I argue that AI will do some of the work of a doctor in the near future. But I don't yet know what those jobs will be exactly.
For ten years, I have been developing scalable solutions to the mismatch between supply and demand in healthcare. I am particularly interested in the interface between healthcare and the customer, where interaction plays an important role. For example, when building chatbots, we have relied on manual algorithms and the interaction has been structured and clunky. The ability to respond to open text has been difficult to implement.
That's why I'm so excited about ChatGPT, which can understand and produce conversational text. I can see that now the word AI is gradually starting to live up to the expectations that have been placed on it. And I would argue that, in line with Gartner's hype curve, within a decade, AI like ChatGPT will surpass even current expectations - although a small death spiral of hype is expected in the next few years.
Humans are needed because AI cannot learn empathy - or can it?
One of the key controversies in healthcare has been that humans are needed because AI cannot learn empathy. In response, I have argued that if empathy can be defined, what is stopping us from creating empathic AI?
In the past, I myself thought that the human level of empathy will not be reached. Building algorithm-based rules for empathy has been laborious, and the hit rate for generating empathy has not been very good.
Study finds AI can already beat doctors in empathy
How wrong I was. AI can already outperform doctors in empathy in a single task. This was shown in a recent North American study (Ayers et al. 2023) comparing AI and telemedicine responses to patients' questions.
In the study, doctors and AI answered the same anonymous online questions without any background information and the quality of the answers was assessed. The questions were not straightforward, but related to common issues that arise in general practitioner consultations: for example, accidents, abnormal laboratory tests, or potentially serious symptoms such as arrhythmia.
The most surprising finding, in my opinion, was related to empathy. The number of messages rated as empathetic by AI was 45%, compared to only 4.6% by doctors. ChatGPT responses were also more often rated as medically benign.
The reasons why AI was perceived as more empathetic in the study may be related to the higher number of characters. The AI wrote four times as many words in responses as humans. This also fits with my own thoughts that the scalability of AI compensates for its other human shortcomings.
Will artificial intelligence replace doctors?
Will AI partially replace me in the future, should I be worried? No, I'm not. The oversupply in healthcare is so deep that I am happy to delegate tasks to AI where it is better than me. I, for my part, can deliver effectiveness in more demanding tasks that AI cannot yet do for me. The authors of the research article therefore suggest that further research should explore the use of chatbots to improve doctors' responses - not to compete, but to collaborate with AI.
We are also - at least for now - poor at tolerating mistakes made by AI. Despite the fact that, statistically, AI solutions would be safer than human intervention. The same challenge with another action: accidents caused by Tesla's autopilot are being made news, even though statistically the technology is already safer than human drivers.
Instead of artificial intelligence, we build supporting intelligence
On the other hand, the fact that we cannot tolerate mistakes made by AI pushes us to develop deployable solutions that are significantly safer than current solutions. As a result, the quality of healthcare will improve alongside productivity gains.
Indeed, the introduction of AI directly at the customer interface is still on the horizon for the future. But we are already well on the way to deploying AI: developing solutions that support the work of Terveystalo's professionals. We are not competing against AI, but working with AI - together we are stronger than alone. AI is coming, we are ready.
* Ayers JW, Poliak A, Dredze M, et al. Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum. JAMA Intern Med. 2023;183(6):589–596. doi:10.1001/jamainternmed.2023.1838
Tuomo Oikarainen's role as Terveystalo's Digital health doctor combines three of his interests: he works at the interface with medical operations, digital development and business.
"Generative artificial intelligence has developed at an astonishing pace over the past year, and we can't even imagine how much it will affect the healthcare sector over the next decade."
Read more blogs
Here are the digital trends in healthcare 2024
Artificial intelligence, health assistants, preventive care and personalised medicine - the digital trends in healthcare in 2024 will focus on how we can modernise healthcare and tackle its deepest problems.
What will occupational health look like in 2030?
Working life is in transition. Many forces challenge both employers and the workforce at the same time. Equally, occupational health must be actively developed in the direction of the future we wish for, so that we don't just drift along the flow.
The best occupational health care in the world is a €24 billion issue for Finland
Finnish occupational health care has been built up over the years by employers and employees, who also finance it almost entirely (employers about 80%, employees 20%). Occupational health care covers more than 1.9 million Finns of working age, whose health and fitness for work are crucial to the success of Finland as a whole, to its competitiveness and to the maintenance of a welfare society. Occupational health is an absolutely central part of our health care system; in addition to its specific role in balancing the growing burden of public health care, it does so in an internationally unique way and with excellent results.
Musculoskeletal disorders absenteeism down 13% in 2022 - What finally made the ship turn around?
I remember well a drawing I made in my notebook during my specialisation. I drew a stick figure running along the line; a stick occupational physiotherapist. The line I drew depicted the transformation of occupational health and working life over the last couple of hundred years; from the industrial revolution, the early days of occupational health and factory doctors, to the Occupational Health Act of the 21st century and the large teams of occupational health specialists. While occupational health had developed by leaps and bounds along the way, our working lives and environments had changed even more rapidly. Despite the leaps and bounds, the stick occupational physiotherapist was still constantly one step behind, often only involved when the problem had already arisen. I wrote the question under my drawing; when will occupational health get to the point where we walk alongside the workplace and see what is coming, so we can prevent it? My stick occupational physiotherapist had to wait a good decade for that answer...
Successful labour migration requires everyone's contribution
The Finnish government has set a target to double the amount of labour immigration by 2030. This means that the aim is to attract around 50 000 workers from abroad to Finland over the next seven years. After that, it is hoped that around 10 000 people a year will move to Finland for work. As immigration for work increases, it is important to ask what this means for everyday work and what changes the diversity of the workforce will require of managers.
Health information is needed for successful work ability management, but what is health information and how should it be protected in an organization?
The most important tasks of occupational health are the prevention of illnesses and maintaining work ability. This can be achieved when information about the health of the staff is available. However, information always goes hand in hand with data protection.