Artificial intelligence is entering one of science’s most ambitious frontiers, learning to model the human body. Yet an extraordinary imbalance threatens to slow that ambition. While AI systems have absorbed vast libraries of human knowledge, the biological data needed to understand how people actually function remains incomplete, fragmented, and difficult to interpret.
The investment is arriving at scale. Yet the scientific challenge is considerable. A model can only make meaningful predictions from the quality and breadth of the evidence used to train it. In biology, much of that evidence remains unavailable or disconnected.
The problem extends into the foundations of medical research. Scientific studies have traditionally examined individual conditions, biological processes, or treatment responses within defined parameters. This has generated valuable discoveries, but it has also left researchers with an incomplete picture of how different systems influence one another. The gap becomes significant when two people with similar genetic characteristics experience different health outcomes or respond differently to the same medication.
Aphrodite Yao, founder of Aethernova, believes this is one of the most consequential questions facing modern healthcare. Working across creative technology, research and immersive innovation, Yao is developing a venture focused on collecting and understanding human functional data, with an initial interest in autism and the wider science of human health.
Her interest in the problem comes from a fundamental question about biological individuality. “How can identical twins share the exact same DNA, yet turn out so differently just because of how they live?” she asks, pointing to the influence of trauma, physical activity, habits, and psychological experiences on human biology.
According to Yao, the question exposes a limitation in how health is studied. A diagnosis identifies an observable condition, but understanding the mechanisms that produce it requires a much wider examination of the individual. Even established clinical assessments can introduce bias when identical frameworks are applied across patients with different ages, genders, cultural backgrounds, or environmental circumstances.
The implications for AI are significant. Language models can process enormous volumes of written information, but human experience cannot be fully understood through words alone. Physiological responses, cognitive patterns, and emotional states provide additional information about how an individual functions. Yao points to heart-rate variability and skin conductance as examples of measurements that can reveal internal responses that facial expressions or verbal accounts might fail to capture.
“All of that matters when you come to understand a human being, what they’re experiencing internally and how they function,” she argues.
The challenge also concerns the quality of the information being collected. Conventional video-based assessments, for example, can rely on estimated movements and visual interpretations that may lack sufficient precision. Yao argues that interpreting biological information requires careful consideration of the circumstances in which it was recorded. The same measurement can carry different meanings depending on the individual and the conditions surrounding it.
Her thinking is also shaped by Aethernova’s wider research interests in neurodiversity. The company’s mission centres on interdisciplinary research intended to improve quality of life. Yao brings experience in game development and immersive technology, alongside personal and research interests in autism. The venture remains in development, with Yao exploring how a more comprehensive understanding of human function could support future scientific research.
A particularly important question concerns the relationship between cellular research and the functioning of the whole person. Scientists are making advances in understanding cellular mechanisms, including mitochondrial activity, but Yao believes more work is needed to connect those discoveries with measurable changes across the human body.
She is drawn to the idea of a “science of healing,” a phrase she encountered through emerging research into the relationship between mitochondria, consciousness, and human experience. For her, the underlying question is why recovery rates differ, why certain treatments work for some patients, and how biological responses might eventually be understood with greater precision.
Such ambitions demand greater collaboration between researchers, clinicians, and technology developers. The Stanford AI Index 2026 highlights the continuing difficulty of collecting biological data, noting that experimental measurements remain expensive and time-consuming. The UK’s own AI adoption plan for life sciences identifies fragmented research records and inconsistent data standards as significant barriers to progress.
The next phase of healthcare AI will depend on whether these shortcomings are addressed. Researchers and investors will need to examine how data is collected, whose experiences are represented, and whether biological measurements retain the context necessary for meaningful interpretation.
Yao’s argument challenges a deeply established assumption in medicine: that health can be understood through individual diagnoses and isolated biological indicators. “We cannot separate the disciplines,” she remarks. “The connections between biology, psychology and other fields deserve greater scientific attention.”
If AI is to contribute meaningfully to the future of medicine, its greatest achievement may depend on what the field chooses to measure, and on whether it treats a person as a system rather than a diagnosis.
