From the exorbitant costs of healthcare in the US to the recurring crisis in the NHS, the idea of effective and affordable healthcare can often feel impossible. This problem will only worsen as chronic conditions become more prevalent and new treatments for previously fatal diseases are discovered. These innovative treatments tend to be expensive, and introducing new approaches into healthcare systems that are resistant to change or overwhelmed is challenging. Additionally, the increasing demand for social care exacerbates funding pressures and complicates resource allocation.
Artificial intelligence (AI) is often proposed as a solution for services that already struggle to do more with less. However, the notion that intelligent computers can simply replace humans in medicine is nothing more than a fantasy. AI has proven to be ineffective in the real world, primarily due to its difficulty in handling complexity. Thus far, AI technologies have had limited impact on the complex, inherently human field of medicine. But what if AI tools were specifically designed to address the intricacies of real-world medicine, encompassing organizational, scientific, and economic complexities?
This “reality-centric” approach to AI is the focus of the lab I lead at Cambridge University. Working closely with clinicians and hospitals, we develop AI tools catered to researchers, doctors, nurses, and patients. While many believe AI’s primary opportunities in healthcare lie in analyzing medical images, such as MRI scans, or discovering new drug compounds, there are numerous other possibilities. One area of focus for our lab is personalized or precision medicine. Rather than employing a one-size-fits-all approach, we explore how treatments can be customized to align with an individual’s unique medical and lifestyle profile.
Utilizing AI-powered personalized medicine could lead to more effective treatment for common conditions like heart disease and cancer, as well as rare diseases such as cystic fibrosis. It could enable clinicians to optimize medication timing and dosage for each patient, as well as screen patients based on their individual health profiles instead of relying on blanket criteria of age and sex. This personalized approach may result in earlier diagnosis, prevention, and improved treatment outcomes, ultimately saving lives and optimizing resource utilization.
Many of these techniques can also be applied to clinical trials. Often, trials fail because the average response to a drug fails to meet the trial’s targets. However, if AI could identify specific groups within the trial data that responded well to treatment, researchers could refine their approach. By creating data models of individual patients, or “digital twins,” preliminary trials could be conducted before embarking on costly trials involving real people. This would reduce the time and investment required to develop a drug, making more life-enhancing interventions commercially viable while targeting those who would benefit the most.
Within complex organizations like the NHS, AI has the potential to efficiently allocate resources. Our lab developed a tool during the Covid-19 pandemic to help clinicians predict the usage of ventilators and ICU beds, which could be expanded across the healthcare service to allocate healthcare staff and equipment more effectively. AI technologies could also support doctors, nurses, and other healthcare professionals in enhancing their knowledge and combining their expertise. Additionally, AI could help address privacy concerns by generating “synthetic data” that reflects data patterns, allowing clinicians to gain insights while protecting identifiable information.
Clinicians and AI specialists are already exploring the potential of large language models like ChatGPT in healthcare. These tools could assist with paperwork, recommend drug trial protocols, or propose diagnoses. While these models offer immense potential, they also come with clear risks and challenges. We cannot rely on a system that regularly fabricates information or is trained on biased data. ChatGPT lacks the capability to understand complex conditions and nuances, which could lead to misinterpretations or inappropriate recommendations, especially within fields like mental health.
If AI is used for diagnosis and makes incorrect assessments, it is essential to establish responsibility. Are the AI developers or the healthcare professionals who utilize it accountable? Ethical guidelines and regulations have yet to catch up with these evolving technologies. Addressing the safety concerns related to using large language models with real patients and ensuring responsible development and deployment of AI is crucial. To achieve this, our lab collaborates closely with clinicians to ensure models are trained on accurate and unbiased data. We are developing new methods to validate AI systems for safety, reliability, and effectiveness, as well as techniques to explain AI-generated predictions and recommendations to clinicians and patients.
We must not overlook the transformative potential of AI in healthcare. Our focus should be on designing and building AI to assist healthcare professionals in enhancing their capabilities, rather than replacing them. This aligns with what I refer to as the human AI empowerment agenda – utilizing AI to empower humans instead of supplanting them. Our goal should not revolve around creating autonomous agents that mimic and replace humans but developing machine learning that enables humans to enhance their cognitive and introspective abilities, thus becoming better learners and decision-makers.
Mihaela van der Schaar, the John Humphrey Plummer professor for machine learning, AI, and medicine, and director of the Cambridge Centre for AI in Medicine at the University of Cambridge.
Denial of responsibility! VigourTimes is an automatic aggregator of Global media. In each content, the hyperlink to the primary source is specified. All trademarks belong to their rightful owners, and all materials to their authors. For any complaint, please reach us at – [email protected]. We will take necessary action within 24 hours.
