Kevin Kelly Talks with WANG Shirui, CEO of Future Doctor: The Era of Medical AGI is Approaching
As the wave of general large models passes through the technological frenzy of the "Hundred Models Battle" and quietly enters the deep - water area of industrial implementation, the medical field, with its extremely high barriers and near - zero tolerance for errors, has undoubtedly become the "touchstone" to test the quality of AI.
In this deep sea closest to birth, old age, illness, and death, how will AI reshape the coordinates of human destiny?
Recently, under the spotlight of the Sina studio, Kevin Kelly, the "Technology Prophet" (hereinafter referred to as "KK") and Wang Shirui, the founder and CEO of Future Doctor had a peak dialogue across generations and technological cycles.
This dialogue not only revealed the evolution code of AI medical technology but also directly addressed a question concerning the common destiny of humanity: Can human lifespan really break through the limit of 120 years?
Facing this proposition, the two interlocutors gave tense answers based on their completely different industrial positions and perceptions of the era.
As a "prophet" who has witnessed the rise and fall of technology for more than half a century, KK always maintains awe of the complex life system. He believes that the fundamental breakthrough in life science is a long and arduous journey: "Maybe humans can eventually live to 120 years old, but it's hard to see it in our lifetime."
In response, Wang Shirui, who stands at the forefront of medical AI innovation and was the first to propose and apply the fast - slow dual - system to medical AI, expressed resonance and long - term optimism. He frankly stated that due to the irreversible accumulation of diseases, the expected healthy lifespan of our generation may be 100 years. However, for the next generation (those born in the 2020s and 2030s), if 'Medical AGI' can extremely compress or even eliminate the lifespan loss caused by diseases and follow the natural law of telomerase, they definitely have the opportunity to knock on the door of 120 years old in advance. This is exactly the core development direction of Future Doctor. In the next three years, Medical AGI will surely achieve a key breakthrough.
Behind this discussion, which ranges from the awe of the technology prophet to the generational outlook of the AI entrepreneur, what kind of technological logic and industrial truth are hidden?
01
Penetrating the Route Divergence, Serious Diagnosis and Treatment is the Solid Foundation of Medical AGI
To discuss how humans can cross the lifespan gap of 120 years, we first need to clarify the underlying logic of life extinction: accidents, aging, and diseases.
In the evolution of modern society, the accidental death rate has generally shown a downward trend; the aging mechanism remains a black box that life science has yet to fully conquer; only 'diseases' have become the variable most likely to be continuously compressed by technology today.
The logic behind this is very clear: if the medical system can give earlier warnings of pathological risks, make more accurate diagnosis and treatment decisions, and iterate the best treatment methods more quickly, the lifespan loss caused by diseases will decline precipitously, and the upper limit of human healthy lifespan will also be completely opened up.
However, in the current industrial ecosystem, attempting to infinitely "compress" diseases with technology will inevitably hit extremely strict compliance red lines and trust barriers. This has also directly led to the implementation of large models in the medical field taking two clearly - defined divergent paths.
The first route is the various "AI health assistants" that have emerged like mushrooms after rain in the market.
Peeling off the commercial packaging, most of these products operate in the shallow waters of "popular science, light consultation, and report interpretation". The end - game of their product logic invariably points to "recommending offline medical treatment". In essence, this is a kind of "edge exploration" that skillfully avoids clinical red lines and does not assume substantial medical legal responsibilities.
But this is by no means the vast ocean of medical AI.
The essential value of medical AI is decision - making. In this reshaping of human health, its moat and commercial ceiling lie in: whether it dares to penetrate into the core area of clinical practice and shoulder the ultimate responsibility of "curing the sick and saving lives".
"Our work is the last stop. We must cure the sick and save lives. 'Future Doctor' doesn't just explain and comfort but cures. Therefore, we have to take legal responsibility." In the dialogue, Wang Shirui pointed out the fundamental difference between the two routes incisively.
Compared with lightweight health consultations, this is a tough battle that allows no compromise.
Facing over 30,000 known diseases and complex real - world medical needs, the inherent "hallucination" of general large models has become a fatal weakness. After all, in the life - or - death medical scenario, the tolerance for errors must always be zero.
This also confirms the sharp warning issued by KK in the dialogue: "The current open - source intelligent agents are only about 90% reliable, but that's far from enough. What we need is not just one 9, but two 9s (99%), or even three 9s (99.9%) of reliability before we can trust AI to do things for us. Currently, it hasn't reached this level."
Therefore, adhering to "serious diagnosis and treatment" is an unshakable foundation for reaching the peak of medical AGI.
This means that AI must face the most hidden and complex pathological maze in the global medical community.
To cross this gate of life and death, Future Doctor has constructed a strict "triage system" of "AI preliminary screening and structured consultation - multi - disciplinary consultation (MDT) intervention - specialized intelligent matching - authoritative expert review and decision - making". Every medical order must be reviewed and signed by a real - life doctor, and the platform and the doctor jointly assume legal responsibility for the diagnosis and treatment results.
Future Doctor's Triage System
In Wang Shirui's vision, only by establishing such a strict mechanism of "human - machine collaboration with clear responsibilities" - AI as an auxiliary tool, and the platform and real - life doctors jointly taking responsibility for the final results can we ensure that every medical decision is well - founded and accountable, thus building an unshakable trust fortress in the deep - water area of medical large models.
KK's understanding of AI medicine coincides with this concept: "Currently, in diagnosis and even treatment, the best model is the partnership between humans and AI. AI has 'more profound knowledge'. They have read through all the materials and are very smart in this regard. But humans are better at reasoning, logic, judgment, intuition, and learning in practice. Combining the two, we can obtain the advantages of both humans and AI. This is exactly what Future Doctor is doing, and I think it is indeed the correct model."
02
Exploring the Ultimate Form of Medical AGI, Future Doctor is Committed to "Systematically Accelerating the Iteration of Clinical Guidelines"
If adhering to "serious diagnosis and treatment" is the ticket to enter the deep - water area, then on this extremely complex track, who can be the first to open the door to medical AGI?
For a long time, the global medical industry has been trapped in the "impossible triangle" of "high - quality, affordable, and accessible".
Aiming at this pain point, Future Doctor's current strategy is to first let AI take on the important task of "inheriting the lost knowledge" - by large - scale replicating the diagnosis and treatment capabilities of top doctors, making high - quality medical services accessible to the general public from a scarce resource.
However, in the complex clinical scenarios, the path of simply "feeding" large amounts of raw data to large models and expecting "miracles through brute force" has become ineffective.
On the technical foundation, Future Doctor has built a unique "three - layer architecture" that simulates the cognitive logic of the human brain for its core system, MedGPT: the "fast system" similar to the cerebral cortex is responsible for rapid response and semantic communication; the "slow system" similar to the prefrontal lobe is responsible for invoking the medical experience library for in - depth reasoning and verification.
The most crucial part is the "ACC layer", which, as the supreme judge, is responsible for reconciling the contradictions between the fast and slow systems, conducting real - time consistency checks, and performing risk classification and confidence weighting, firmly controlling the most fatal "hallucination" of AI.
The core of this architecture comes from the structured data of the "symptom - disease - medication" triad repeatedly cross - verified by more than 120 top experts from tertiary - level A hospitals and the vast 100,000 medical guidelines.
Future Doctor's Expert Wall
Matched with this is the "dual - flywheel engine": the small flywheel focuses on the clinical standardization path, firmly guaranteeing the absolute lower limit of diagnosis and treatment safety; the large flywheel continuously absorbs and spits out the 20,000 real - world diagnosis and treatment feedbacks precipitated by tens of thousands of doctors on the platform every week, continuously raising the upper limit of diagnosis and treatment capabilities.
The moat forged by real - world data, the original architecture, and medical expert resources makes it difficult for competitors to force - mature in the short term with just capital.
A greater breakthrough lies in the reshaping of the industry evaluation system.
Currently, many medical large models are keen on using "passing the licensing examination for practicing physicians" as a marketing gimmick and even scoring high in the HealthBench evaluation benchmark launched by OpenAI.
However, "factual accuracy and communication ability" are by no means equivalent to real "clinical effectiveness and safety".
To prevent medical AI from becoming a "test - taker", Future Doctor recently led 32 top domestic medical experts to publish the world's first "Clinical Safety - Effectiveness Dual - Track Benchmark (CSEDB)" for evaluating the clinical applicability of medical AI in the top - tier journal npj Digital Medicine under Nature.
The birth of this standard marks that the evaluation system of medical AI has shifted from the "standardized test" divorced from the real world to the "clinical decision - making" mode in real - world scenarios, filling the international gap in the serious diagnosis and treatment evaluation system.
However, large - scale replication of top - notch medical experts is only the first piece of the jigsaw puzzle in Future Doctor's medical vision. Its next - stage ambition is to "bring peace to the world" - to explore the ultimate form of medical AGI.
In the traditional history of medical evolution, the update of a clinical guideline often takes a long time, 3 to 5 years. From proposing a hypothesis, recruiting patients, conducting real - world research, to data collection and final verification, it is a slow and costly process.
The medical AGI defined by Future Doctor not only needs to horizontally replicate the capabilities of today's experts but also vertically conduct systematic iterations through the dynamic feedback of a large amount of real - world data, ultimately subverting the speed and quality of the evolution of clinical medicine.
In the picture painted by Wang Shirui, in the future, AI will take the lead in information collection, hypothesis - making, finding target patients for enrollment, and conducting follow - up analysis.
"AI will collect all the information, analyze all the information, and put forward all the hypotheses. AI will find doctors for approval. After approval, AI will find all the target patients who can be enrolled in the trial. Then AI will continuously track these patients, conduct follow - ups, and finally complete all the data collection and analysis."
Under this in - depth industrial reshaping, the clinical verification of new treatment methods, which originally took two or three years, will be extremely compressed. Wang Shirui predicted: "Within three years, the most important AGI can be realized. For some diseases, results can actually be obtained in 72 hours... The update of clinical guidelines is very likely to be shortened from three to five years to less than three months."
This will be a revolution in the medical paradigm - medical AGI will upgrade medicine from "mainly relying on the slow accumulation of expert experience" to "being able to be systematically accelerated and evolved by AI".
When the update speed of disease treatment methods explodes exponentially, the scientific vision of "letting humans live healthily to 120 years old" will have the most solid technical and data support.
03
The Ultimate Gift of Technology: Answering "Real and Important" Life Propositions
As this more - than - two - hour dialogue came to an end, the topic had gone beyond a discussion of the technological evolution route and evolved into a philosophical speculation about the future of human civilization.
When asked about his dream for the future, KK, who has seen the rise and fall cycles in the technology circle for decades, gave a monologue full of humanistic warmth, which also left the most important footnote for pathfinders like Future Doctor.
"It's absolutely impossible for me to be standing here alive today if it were 50 or 100 years ago. This must be attributed to the progress of medicine." KK emotionally admitted, "What we are discussing and doing now is an undertaking that touches the most fundamental and underlying aspects of civilization - the pursuit of health, longevity, inner peace, and satisfaction with life, rather than creating anxiety and fear."
In KK's view, neither the desire for longevity nor the leap of AI aims to let humans spend long and weak retirement years in boredom.
As he said: "The meaning of longevity is not having a few more years of retirement time but having more time to create, contribute, share, and bring beauty to the world."
Those who can truly drive humans to cross the limits and shape a better future are always the down - to - earth optimists.
In the current rapid development of technology, capital and technology are easily lost in the short - term and fast - paced flow of traffic games or commercial monetization. In contrast, medical AI is destined to be a long and arduous "difficult but correct" journey.
"This is not as simple as making a short video or generating traffic. This is a truly fundamental, essential, and life - related matter for everyone... Someone once said that the smartest brains in the world are thinking about how to make mobile phone advertisements more accurate. But now, the world's top minds are doing truly important and meaningful things." KK highly praised Future Doctor's industrial exploration in this way, "I think Future Doctor is at the center of this revolution (the evolution of the medical paradigm)."
The journey of medical AGI is a long - distance race for long - termists. It not only requires the courage to break through the medical "impossible triangle", consolidate a profound medical cognition and data barrier, but also firmly adhere to the bottom line of "serious diagnosis and treatment".
Looking into the future from