“As the world becomes increasingly interconnected and technology continues to advance, the field of artificial intelligence (AI) is rapidly expanding” wrote ChatGPT when asked to write the first line of a news article, unprompted as to the topic. OpenAI’s invention ChatGPT is not only the most advanced publicly available AI, as demonstrated by its opening sentence, but also consumer app to date.
The software has been an internet sensation since its introduction in November 2022, yet it is not the only AI making waves, just the one making headlines.
Artificial Intelligence has been forecast to soon take over many everyday jobs. But, since being introduced to a machine that can predict cancer recurrence and a Tanzanian project to curb the prevalence of in the country, it is clear that healthcare is the area already benefiting from this advanced technology more than most.
AI in the Medical Practice
The National Institute for Health and Care Research (NIHR) has so far invested £123 million in 86 AI technologies. These include a system called Tommy’s App to assess risks in pregnancy and determine the chance of a premature birth, the cyberpunk-sounding software Mendelian, that can identify a patient’s risk of being diagnosed with a rare disease, and a new technology called Organ Quality Assessment (OrQA).
The latter, despite having the least jazzy name, is designed to streamline the transplant process by identifying suitable organs more accurately and efficiently than humans. Predictions state this will allow doctors to perform around 300 more transplants a year, good news for the 7,000 patients awaiting organ transplants in the UK.
The system uses similar technology to facial recognition software, trained using thousands of images to discern suitability more effectively than the human eye ever could. The speed of this process is vital seeing as organs only survive a short time once removed from the body.
The Future of AI in Healthcare
These advances were unthinkable in the not so distant past, so does it mean the scope for AI in healthcare will only widen? Mark Hoogendoorn, Professor of Artificial Intelligence at the Vrije Universiteit Amsterdam, seems to think so.
“If you look at the current penetration of AI techniques in health care this is still low” he tells Distilled Post, “I believe this will change rapidly in the near future. Currently, we see a lot of applications in radiology and a few other domains. In the future, I would expect this to widen to essentially all medical disciplines. Hereby, I wouldn’t expect the systems to take over in many domains, but rather act as support for medical professionals.”
When asked what forms of AI will thrive most, Hoogendoorn said, “I believe that forms where AI systems and medical staff collaborate are most fruitful as each have their own strengths. However, for some specific disciplines a high degree of AI-driven automation might show to produce best results.
“For example, when looking at generating treatment plans for tumor radiotherapy whereby the goals is to maximise the dosage for the cancer cells while minimizing the dosage for organs at risk is something AI systems can calculate more precisely than any human ever could, or think of precision surgery performed by a robot.”
Racist Robots? Biases Within Data
Though that last remark conjures up sci-fi-esque images of robotic arms for doctors, reminiscent of films such as Star Wars, the path to a futurist utopia has not been perfectly paved for us yet. There are still ethical and practical issues with the coming AI revolution.
The primary one is bias. “In the medical data that forms the basis of the development of these systems biases are often present, this can result in models that disadvantage certain groups” explained Professor Hoogendoorn, before adding that addressing this issue is of “utmost importance”.
Indeed, the effect of bias has been a salient issue for academics in the field practically since its inception. Researchers and governments have long raised concerns over racial and gender bias in AI databases, including search engines, risk algorithms, and evidently healthcare databases.
It is unclear how, and how soon, this issue will be resolved. As AI learns from past data that usually contains human biases, the AI machines essentially teach themselves various prejudices, see Amazon’s misogynistic recruitment tool for a perfect display of corporate foot-shooting. However, developers are confident that, by strategically deploying AI on carefully selected data, they can mitigate, or at least minimise, the effects of these biases.
Is ’Big Data’ Coming for Your Job?
Then there is the added anxiety of job security for doctors as the daunting prospect of ‘Big Data’ culling medical employment opportunities seems to be on the horizon. However, this issue is more divisive than the previous, which is agreed upon almost unanimously.
Lack of training costs, uncapped labour, and an ability to constantly improve make AI an attractive financial alternative to human doctors. In fact, this was raised as genuine fear by over a third of experts asked in 2019, who suggest “White-collar” jobs, such as medical personnel, would perhaps be more automatable than their “Blue-collar” alternatives like those in manufacturing.
Despite this, many physicians themselves appear to be embracing AI for its assistance rather than fearing it. They emphasise the need for collaboration and human input, grateful for AI, not suspicious of it.
When asked by Distilled Post, Professor Hoogendoorn also maintained this position. “Most AI systems on the short term will support healthcare professionals, not replace them” he answered, and then brought up the relevant point that, “in many Western countries there are enormous challenges to keep health care at our current level due to the aging population and staff shortages where we will need all professionals to address those challenges. AI can hopefully make a valuable contribution there too.”
