Hangzhou Diagens Biotechnology The Next Decade of Medical AI: How Humans and AI Collaborate – A Conversation between Kevin Kelly and Dr. Ning Song of Diagens Tech (2526. HK)

On 25 September, during the 5th Global Digital Trade Expo held in China, the 2026 Festival of Medical AI hosted by Diagens Technology Co., Ltd. (2526. HK, Diagens‑B, “Diagens Tech”) , themed “the Next Decade of Medical AI”, brought together Kevin Kelly, futurist and founding editor-in-chief of Wired magazine, and Dr. Ning Song, Founder and Chairman of the Board of Diagens Tech. The two discussed the industrialisation of medical AI, the future relationship between clinicians and AI, technological innovation and next-generation capabilities.

The conversation centred on one core question: as AI’s adoption in healthcare becomes an established trend, the next phase hinges not merely on model performance, but also on sustainably producing medical AI, reshaping clinician-AI relations, and embedding technology into broader clinical practice. The discussion spanned the flourishing landscape of medical AI, “AI does medicine, humans do doctoring”, innovation, choice and learning in the AI era, and ultimately explored how humans, technology and industry could collaborate.

I. The Next Phase of Medical AI: From Owning a Single Model to Owning Capacity for Continuous AI Creation

Kevin Kelly described the integration of AI and healthcare as inevitable. In his view, AI’s entry into medicine was an irreversible trend, and Diagens Tech was among the first pursuing a holistic approach to medical diagnosis. Tasks that AI could perform faster, more reliably and more accurately would gradually be delegated to the AI.

Given that AI adoption in healthcare is inevitable, the next question arose: will medical AI eventually converge into one unified “best answer”?

Dr Song held the opposite view. He noted that even for the same disease, patient and medical image, interpretations might vary across hospitals, clinicians and regions. Because of this inherent diversity in medicine, Dr Song argued that the future of medical AI should not be dominated by a single solution, but instead feature a diverse ecosystem. In his vision, every specialist radiologist could train their own AI Copilot based on Diagens Tech’s self-developed iMedImage® foundational model for medical imaging. The AI assistant would absorb the clinician’s unique clinical expertise, rather than forcing everyone to rely on one generic AI system pre‑built by third parties.

Kevin Kelly responded promptly that customised medical imaging and diagnostic capabilities built for individual hospitals and clinicians would be of great significance. Clinicians naturally held slightly different observational perspectives, and these nuances could be reflected in their respective AI Copilot.

In Kelly’s view, software requires constant rewriting, algorithms keep evolving, and AI needs retraining in the AI era. A company’s true long-term asset, therefore, might not be a single advanced software platform or algorithm, but the process that continuously generates new software, algorithms and AI systems.

This very “process” was embodied in Diagens Tech’s iMedLoop™ platform, an accelerator for medical imaging AI research and production. Centred on the iMedImage® foundation model, it links iMedStudio™, the multi-modal intelligent annotation and data generation workspace; iMedMaaS®, the model training and deployment platform; and DoctorBench®, the unified evaluation framework. It spans the full pipeline: data generation, model training, standardised evaluation, release, deployment and application feedback.

This creates a medical imaging data flywheel: broader real-world deployment generates more high-quality real-world data and feedback to continuously enhance model capability. More capable models, in turn, support more specialties, tasks and clinical scenarios, generating fresh data and feedback to create a self-reinforcing cycle – the more the models are deployed, the more capable they become, and greater capability in turn fuels wider adoption.

II. When AI Handles “Medicine”: Redefining Clinicians’ Value and the Boundaries of Healthcare

As AI assumes a growing share of medical information processing and diagnostic tasks, how will the role of clinicians evolve?

Kevin Kelly offered his verdict: “AI does medicine, humans do doctoring.”

He observed that people already consult AI before seeing a clinician. Even patients with trusted doctors often seek information from AI first for its instant responses – this is not a future scenario, but an existing reality.

Yet accessing medical information is never equivalent to seeing a clinician. Beyond diagnostics and knowledge, humans need empathy, wisdom, counselling, support and judgement. Patients sometimes need someone to hold their hand, guide them through treatment journeys, and encourage them to adhere to medication and health-critical behaviours.

Therefore, as AI takes on more of the medicine work, clinicians gain more time to focus on doctoring.

Dr Song then shifted the discussion to the next generation. For adults today, AI is an emerging technology. But the younger generation grows up in a world where AI has always been present. Will AI be seen as a natural companion? If they place greater trust in AI-generated health advice, what expectations will they hold for the humanistic care delivered by clinicians?

Kelly answered with the analogy of the calculator. Today, people rarely doubt the output when using a calculator. He believed that once medical AI achieves sufficient accuracy and reliability, people might develop similar trust, and would no longer consciously question whether they can trust AI.

However, he quickly added: healthcare requires more than mere accuracy. People want inspiration. We want someone to help us be better. We need psychological support to guide us through. The younger generation might trust results from medical AI, yet human needs stretch far beyond information and diagnosis.

When the audience asked who stood to benefit most in the age of medical AI, Kevin Kelly replied first: “It’s ultimately the patients.”

Dr Song offered a different perspective. He stated that when looking ahead over the next five to ten years, it is better to focus on the broader population rather than segment people by their roles, such as clinicians, patients or researchers. The wealthiest groups already have access to scarce, high-quality care even without AI. Where AI can deliver real impact is in boosting the efficiency and reach of limited healthcare resources, improving population-wide health outcomes.

“The greatest value of this industry,” said Dr Song, “lies not in supporting a narrow group, but in benefiting a far larger, broader population – and this is what we have consistently strived to achieve.”

This gave deeper context to the notion of “AI does medicine, humans do doctoring”. For Diagens Tech, foundation models and industrialised systems are ultimately about more than producing more AI systems. They aim to make scarce high-quality medical capabilities accessible to more people with greater efficiency, advancing inclusive medical AI.

III. When Answers Are Easily Accessible: Judgement, Learning and Creation as Core Skills in the AI Era

As AI becomes increasingly adept at generating answers, will it transform how innovation happens? Kevin Kelly’s belief is that AI will bring substantial innovation across the entire healthcare sector, particularly in medical imaging, and this shift is taking place at great speed. As AI becomes cheaper and more accessible, it will be integrated into a wider range of medical devices and applications.

Yet he drew a distinction between two types of innovation. The first consists of routine innovation and incremental improvements, work that AI already performs very well. The second is breakthrough innovation, which seldom emerges from obvious answers but from unexpected, even disruptive ideas. Innovation of this kind remains challenging and still relies on humans.

Kevin Kelly noted that outputs generated by AI often reflect a form of “consensus”. True innovation, however, sometimes demands that people explore ideas beyond consensus. Clinicians may acknowledge that AI’s assessment aligns with prevailing consensus, while forming alternative hypotheses and judgements based on their understanding of the individual patient.

In his view, experimentation beyond consensus is where innovation can emerge. Dr Song fully agreed.

When the conversation turned to the capabilities required in the AI era, Dr Song outlined three pillars: first, asking good questions; second, making decisions; third, perseverance.

He emphasised the third point in particular: the capacity to persist and stay committed amid hardships and setbacks. This applies not only to young people but to everyone living in the AI age.

The themes of questioning, decision-making and perseverance anchored the ensuing discussion.

As barriers to knowledge fall rapidly and people gain access to ever greater volumes of information and possibilities, does choice itself become a burden? Can AI help resolve this dilemma?

Kevin Kelly called this the “paradox of choice”. When options multiply beyond a certain threshold, decision-making becomes harder. Sometimes, distilling dozens of possibilities down to four makes action simpler. People already use AI for this purpose: overwhelmed by a long list of attractions, they ask AI “What should I see and do?”, and AI helps narrow down the many possibilities into a shortlist.

While AI expands access to information, it also helps people streamline choices. This does not mean humans hand over the act of choosing to AI. On the contrary, as machines grow more adept at generating answers and options, humanity’s ability to ask meaningful questions, judge what is worth selecting, and take accountability for decisions becomes ever more critical.

The conversation on the next decade of medical AI circled back to learning.

Kevin Kelly argued that one of the most vital capabilities for the future is “learning how to learn”. Technology evolves at such a rapid pace that many skills required in the future have no established curricula as yet, making the ability to learn how to learn increasingly important.

As a result, learning will increasingly take place at work. Organisations such as Diagens Tech will become more like “schools” in the future. Many skills for working alongside new AI systems are not yet covered by established university programmes, yet enterprises already need talent with these capabilities. Companies therefore take part in talent development and education.

Dr Song concurred. He commented that companies constantly identify a need for such talent yet struggle to find enough graduates equipped with these skills. Hence, “Why not train them ourselves?” In September 2026, Diagens Tech and The Hong Kong Polytechnic University launched their Joint Laboratory for Artificial General Intelligence and Medical Applications, built precisely to nurture talent for the AI era.

Kevin Kelly then recalled his experience at Wired during the early days of the Internet. Back then, the Internet was still nascent; there were no formal training programmes, and almost no one possessed a full set of “Internet skills”. They recruited young people and learned alongside them to build new capabilities, effectively “inventing the internet together”. Rather than prioritising existing technical skills, they valued aptitude, attitude and character.

The future will bring more than a continuous stream of new models. A new generation of clinicians, engineers and young people will learn new knowledge, raise new questions, and co-create capabilities that do not exist today.

AI has repeatedly proven capable of tackling increasingly sophisticated tasks. In the decade ahead, the key question may shift away from how intelligent AI can become to how humans will work with this power and what direction we choose to take it.

Healthcare serves as a vital testbed. It demands technical accuracy while requiring humans to retain judgement, accountability, trust and creativity. It seeks to extend scarce specialist expertise to broader populations while preserving diversity across clinicians, patients and clinical scenarios.

This is the core value of the conversation. Kevin Kelly explored the emerging new order of the AI age, while Dr Song grounded these discussions in real clinical settings and the day-to-day decisions faced by clinicians, patients and companies.

For Diagens Tech, the future extends far beyond new product launches. Diagens Tech will help shape new frameworks for medical AI development and collaboration, advancing technology while engaging clinicians, engineers and young practitioners to frame new questions and build new capabilities. Where previous technological revolutions redefined how humanity accesses information and connects globally, medical AI now explores how technology can weave itself more deeply into human life, health and decision-making. Diagens Tech aims to be an active participant in this future as it takes form.