In 2023, an AI-generated video of Will Smith eating spaghetti went viral. It was awkward. Fingers looked odd, and facial expressions were distorted. Last year, a new version appeared. This time, it was almost perfect. A similar pattern of progress can be observed in healthcare AI, moving from early systems full of hallucinations to tools that provide reliable, clinically practical guidance.
From Copernican to spaghetti moment
Will Smith Eating Spaghetti has become a benchmark for assessing the growing capabilities of generative AI and has even been featured on Wikipedia. Daniel Kraft, Founder of NextMed Health, used this comparison at the ICT&health World Conference in Maastricht to highlight the exponential development of AI in healthcare.
“Exponential change often looks slow until it suddenly becomes impossible to ignore,” according to Kraft. The Will Smith spaghetti example represents how AI systems quickly learn from data what’s different from other innovations that grow linearly.
We also see it in healthcare. In medical imaging, AI systems now match or exceed the performance of average specialists in areas such as skin cancer detection, diabetic retinopathy screening, and breast cancer analysis. In radiology, studies show double-digit reductions in false positives and false negatives when AI is used as a second reader. In cardiology, algorithms can predict atrial fibrillation or heart failure from a normal-looking ECG long before symptoms appear.
AI capabilities extend far beyond ChatGPT, which passes the United States Medical Licensing Examination (USMLE) with higher scores every month. New models mimic how human doctors make decisions. For example, MAI-DxO (Microsoft AI Diagnostic Orchestrator), designed for complex medical diagnostics, has achieved 85.5% diagnostic accuracy, outperforming the 20% accuracy of human doctors.
Precision improves, costs decline, and access and trust among doctors increase. Tasks that once required rare expertise available only at university medical centers can now be scaled globally, making them accessible even in remote, small villages with underdeveloped healthcare infrastructure.
I recall when, 6 years ago, Lucien Engelen, CEO of TransformHealth, referred to a “Copernican moment” in healthcare, in which “we (mind)shift from systems that revolve around the professional to systems that revolve around the patient.” And he was right. Now, it is “the spaghetti moment of medicine.” In 2022, we laughed at how ChatGPT lied, hallucinated, and was unable to diagnose. Today, according to the latest OpenAI report, 40 million people globally turn to ChatGPT daily for health-related questions (5% of all prompts are health-related).
AI is here. Use it or lose it
Before 2022 and the release of ChatGPT, AI had struggled in healthcare, except for medical imaging. Meanwhile, AI is increasingly used in health IT software and applications, including scheduling systems, triage tools, and virtual nurses. AI Scribes is the first in decades to enable quick, easy note-taking in electronic medical records without disrupting the patient’s visit. Pilot studies suggest that doctors and patients like it. It reduces administrative burden and saves a few hours per week while improving communication with patients, as healthcare professionals no longer have to stare at the computer screen.
Beyond the minor improvements, AI offers the opportunity to change the paradigm in medicine. We have long discussed healthcare that is reactive, episodic, too expensive, fragmented, and delivered in a “one-size-fits-all” model. These were theoretical debates because we lacked the tools to change it. With AI, care becomes continuous, proactive, and personalized, happening anywhere. These shifts could drive the transition that is the primary goal: from fee-for-service to fee-for-outcome systems.
AI is also a driver of consumer health, empowering individuals with tools previously available only in clinics and medical laboratories. Wearables track heart rate, sleep quality, glucose, and activity in real time; smartphones capture voice, movement, and behavior as health indicators; smart sensors collect data that decode unhealthy behavior. AI systems analyze these streams, flag risks early, and help us make informed decisions. Today, subtle changes in speech predict cognitive decline, while changes in typing patterns signal neurological disease.
Diagnosis becomes faster, and monitoring becomes real-time. Healthcare professionals’ capabilities get augmented: Instead of asking how you felt in the last six months, your doctor can see what happened yesterday or even an hour ago without the need to go through the gigabytes of data. AI can do it.
The human gap
Patients already check symptoms in ChatGPT and other tools. AI supports radiologists, pathologists, and dermatologists. Virtual care platforms triage patients. Algorithms optimize hospital operations. It is current practice, and we can only expect more AI in every aspect of our lives.
Soon, a personalized health agent will tell us how to live healthier. It will know your baseline, risks, goals, and preferences. It will nudge you, alert you when health indicators change, help you navigate care independently, and send you to the doctor when necessary. AI bots are expected to become constant companions between visits, particularly for patients with chronic conditions. This means that healthcare systems shift from treating disease to maintaining health. Hospitals will focus on acute and complex care, while most prevention and monitoring will happen at home or in the community. Care teams will intervene earlier and more precisely.
Listening to Daniel Kraft, the future gets bright. However, technology and AI can grow exponentially, but society grows linearly. Put simply: Technology changes fast; humans adapt slowly. We are already seeing technology fatigue, and many people are unable to keep up. The democratization of access to information through the development of the internet and smartphones has made knowledge at our fingertips, but also fake news.
This gap creates tension, fear, and sometimes backlash. Clinicians are beginning to worry about job loss and their role in a world where AI agents know everything, and data collected at home is becoming more relevant than data in the electronic medical record. Some patients welcome AI, while others fear it, which can exacerbate healthcare inequalities.
AI won’t change human nature and economic status
“We have to use AI as intelligence amplification, not artificial replacement,” according to Kraft. This goal is essential to preserving values such as empathy in human care and humility. This is also a call to experiment with AI in healthcare, with a focus on solutions that work in clinical settings. Because the most significant risk is not that AI moves too fast, but that healthcare systems move too slowly and widen the gap between what is possible and what is delivered.
Unfortunately, there is no equally strong example of AI progress in healthcare comparable to Will Smith eating spaghetti, which so vividly illustrates the rapid advances in generative video models. I disagree with Daniel Kraft that this example can be taken as evidence of exponential AI progress across all domains, including healthcare. Making a medical decision is fundamentally different from recognizing patterns in billions of pixels. While it certainly involves data analysis, it also requires context, responsibility, and human judgment.
AI will provide better insights into our health, but most recommendations will remain the same: eat well, move more, sleep enough, manage stress, and detect disease early. The real change is that we can intervene earlier and receive more personalized guidance. Still, people must implement these recommendations, which is often difficult given differences in backgrounds, life experiences, motivations, values, needs, living environments, and economic conditions. In our enthusiasm for what technology makes possible, we should not lose sight of this.
