The global scientific research system is trapped in an unprecedented paradox: funding grows exponentially while the proportion of disruptive output continues to collapse. U.S. research productivity declines by an average of 5.3% per year, and a large-scale study published in Nature shows that the share of papers capable of redirecting their fields has dropped by over 90% in sixty years. Meanwhile, more than 70% of global research funding is captured by an “academic oligarchy” system dominated by administrative power and scholarly relationship networks, while researchers with genuine breakthrough capability — whether young scholars inside the system or “geeks” outside it — are systematically excluded from resources. Through cross-regional empirical data, this paper reveals the global face of this “high-input, low-disruption, intense involution” structural impasse, argues for AI as an inevitable game-changing variable, and re-examines the real innovative contributions of the “amateur scientist” community stigmatized by mainstream academia.
Sliced Research: The Dimensionality Reduction Trap of Specialized Silos
The core organizational logic of the modern research system is industrial assembly-line division of labor. Each researcher is assigned to an ever-narrower “slice,” focused on an ever-smaller problem. This model has its rationale at the efficiency level — it ensures continuous growth in paper output. But at the cognitive level, it produces a fatal consequence: no one sees the complete picture anymore.
This is what we define as “Sliced Research.” It is the further degradation of specialized “Silo Knowledge.” Silo Knowledge means you only understand your own well; Sliced Research means you only understand one layer of soil within your well. When all incentives for researchers point toward “producing papers on the minimum publishable unit,” the rational choice is to do the safest, most predictable, least risky research — that is, “scientific patching.”
Scientific Patch — incremental tuning and mending of existing frameworks, rather than questioning and rebuilding the frameworks themselves. What can be written into a grant application is, by definition, already no longer the most important question.
Looking back at pivotal breakthroughs in the history of science, virtually none could pass today’s funding review. Darwin took over twenty years before he dared publish On the Origin of Species; Einstein pondered the speed of light while working in a patent office; McClintock’s transposon research was neglected for decades. What these achievements share is that they all appeared to have “no practical application” at the time of their birth and could not match any modern research grant’s KPIs.
Massive Expenditure, Marginal Output: The Collapse of Global Research Efficiency
The landmark NBER (National Bureau of Economic Research) study reveals a striking fact: U.S. research productivity is declining at an average rate of 5.3% per year. Merely to sustain a constant rate of GDP growth, research investment must double every 13 years — not because new discoveries yield greater returns, but because under the current system, discovering new ideas has become increasingly difficult.
Meanwhile, U.S. R&D spending has reached approximately 3% of GDP, surpassing the peak of the Space Race era in the 1960s. Yet total factor productivity (TFP) — the core metric measuring innovation’s contribution to economic growth — has grown only 0.5% over the past decade, less than half the 1960s rate. This is an extremely clear signal: money is increasing, breakthroughs are shrinking.
The landmark 2023 study published in Nature (Park et al.), analyzing 45 million papers and 3.9 million patents, concluded that papers and patents are increasingly unlikely to push science and technology in new directions by breaking with the past. This pattern is universal across all academic fields and is not affected by changes in citation practices or publication quality.
Despite the exponential growth of new scientific and technological knowledge, the rate of disruptive discoveries is plummeting. The study attributes this trend in part to reliance on narrower slices of existing knowledge — which is precisely the direct consequence of “Sliced Research.”
Academic Oligarchy: The Closed-Loop Monopoly of Power, Funding, and Connections
The root cause of collapsing research efficiency is not that “science has gotten harder” — though that is one factor — but that the resource allocation system itself is manufacturing mediocrity. We call this “Academic Oligarchy.” It takes different forms globally but is identical in essence: the deep entanglement of power, funding, and interpersonal relationships, systematically excluding those with innovative capability from resources.
| Region | Core Form of Academic Oligarchy | Key Evidence |
|---|---|---|
| China | Administrative power dominates funding allocation | Over 70% of research funding distributed by “titles, honors, and positions”; “higher rank = larger grants” is a universal phenomenon |
| United States | Elite institution monopoly + politicization shocks | NIH funding highly concentrated in a few states; states with 15.7% of population receive only 6.4% of funding; Trump administration drastically cuts research budgets |
| Europe | Position scarcity + network dependence | Permanent positions extremely scarce with opaque hiring; in Italy, obtaining tenure depends heavily on connections within the university hierarchy |
| Japan | Institutional rigidity + academic clan traditions | National university base operating budgets cut 10% over a decade; highly cited paper rankings dropped to 13th globally; PhD students continue to decline |
| South Korea | Academic inbreeding + chaebol infiltration | 95.6% of Seoul National University faculty are its own alumni; SKY university graduates dominate government and chaebol management |
Regardless of institutional differences, all advanced economies’ academic systems exhibit the same structural deficiency: the channel between capability and resources is severely distorted by interpersonal dynamics. The forms differ — institutional concentration in the U.S., position scarcity in Europe, budget shrinkage in Japan, inbreeding in South Korea, administrative control in China — but the essence is the same.
75% of surveyed researchers reported that pressure to publish has continuously intensified over the past decade; nearly 60% believe that management’s demand for quick results — short-termism — has directly damaged research productivity.
The deeper problem is the failure of the talent selection mechanism itself. The latest research indicates that approximately half of the decline in research productivity is not because discovery has gotten harder, but because the system has failed to effectively screen for high-output researchers. Large numbers of lower-productivity individuals are admitted into the system, while those with genuine breakthrough potential are excluded for lack of “relationship capital” — the massive post-WWII expansion of education and research funding has produced no improvement whatsoever in productivity growth.
The Geek Counterattack: Innovators Excluded by the System
Mainstream academia habitually labels researchers outside the system as “pseudoscientists,” equating them with pseudoscience. But this stigmatization conceals an important fact: before science became professionalized, virtually all scientists were “amateurs.” And even in today’s highly specialized world, the “geek” community outside the system continues to produce astonishing results.
- Must produce “publishable” safe results
- Funding depends on relationship networks and academic titles
- Intellectual property belongs to the institution, not the individual
- Interdisciplinary work is blocked by institutional barriers
- Short-termism pressure prevents long-term exploration
- Driven by curiosity and the problem itself
- Unconstrained by disciplinary boundaries or administrative hierarchies
- Intellectual property belongs entirely to the creator
- Naturally interdisciplinary — because they are not “inside” any discipline
- Can tolerate failure; no need to publish papers
However, the mainstream academic establishment’s attitude toward these outside-the-system innovators is fundamentally instrumental — “you can help me collect data, but you can’t sit in my seat.” Citizen scientists are typically assigned only auxiliary work such as classification, counting, and labeling, while the real power to set scientific agendas and allocate resources remains firmly held by the institutional “Scientific Patch Masters.”
You don’t need a PhD to do biology. In community labs, graduate students can develop technology outside the university system, so they don’t have to hand over intellectual property to the university — they are the sole inventors. This directly challenges the institutional logic of “taking a cut” from innovation outputs.
Triple Mismatch: Why Resources Systematically Flow Toward Mediocrity
Synthesizing global data, the causes of collapsing research efficiency can be distilled into a triple mismatch — these three mismatches mutually reinforce one another, together constituting a self-sustaining mediocrity production machine.
This triple mismatch collectively explains a paradox: why global research budgets run into the trillions yet produce fewer and fewer disruptive results. The problem is not that humanity lacks innovative capability — Lorenzo’s Oil, Green Pea galaxies, the deciphering of Ice Age lunar calendars all prove that innovative capability is everywhere — the problem is that the current system systematically separates those with innovative capability from innovative resources.
AI: The Counterforce That Breaks the Dimensionality Reduction
If “Sliced Research” is the dimensionality reduction behavior of specialized silos, then AI is the counterforce that breaks this dimensionality reduction — not accidentally, but inevitably. AI’s structural characteristics precisely hedge against every deficiency of the academic oligarchy system.
| Structural Deficiency of Academic Oligarchy | AI’s Structural Hedge |
|---|---|
| Disciplinary barriers — knowledge carved into isolated silos | Naturally cross-domain — simultaneously draws on multi-disciplinary frameworks to examine the same problem |
| Relationship gatekeeping — resource access depends on connections | No social network — unconstrained by cliques, mentorship lineage, or alumni ties |
| Short-termism — funding cycles force safe choices | No funding pressure — no need for promotion; can explore “unfundable” questions |
| Information monopoly — knowledge locked behind commercial platforms | Information democratization — lowers knowledge access barriers, empowering all researchers |
| Inbreeding — intergenerational transmission of homogenized thinking | Cognitive diversity — training data covers the broad spectrum of human knowledge |
More importantly, AI has a dual empowerment effect on innovators both inside and outside the system. For suppressed young researchers within the system, AI can help them bypass “relationship gatekeeping” and directly access cross-disciplinary knowledge and methodological tools. For neglected geek communities outside the system, AI can provide professional-grade analytical capabilities previously beyond their reach — DNA sequencing interpretation, protein structure prediction, complex system modeling — capabilities once exclusive to top laboratories are now rapidly democratizing through AI.
The biohacker community has already begun combining AI with DIY experiments. AI algorithms are being used to predict gene editing outcomes, optimize bioengineering workflows, and even design novel antibiotic molecular structures. This “AI + Geek” combination may become the most effective pathway for bypassing the academic oligarchy system.
Of course, clarity is needed: AI is currently better at “lateral connection” (discovering overlooked links between disciplines) than “vertical breakthrough” (original insight that redefines the problem itself). The latter still relies on human intuition and judgment. Therefore, the true paradigm shift will come not from AI replacing humans but from the intersection of “AI’s breadth + human depth” — especially when that “human” is a geek unconstrained by the system.
Redefining “Who Is Qualified to Do Science”
The argument chain of this paper can be summarized as follows:
First, the “sliced” organizational logic of the modern research system has degraded knowledge production into the mass manufacture of “scientific patches” — paper volume explodes while the share of disruptive output collapses. The data are indisputable: total factor productivity stagnates, the disruption index continues to decline, and research efficiency decreases at 5.3% per year.
Second, this efficiency collapse is not due to the single cause of “discovery getting harder” but is the structural inevitability of the “academic oligarchy” system. The closed-loop monopoly of funding-power-connections creates a triple mismatch, causing resources to systematically flow toward mediocrity, safety, and connections rather than toward genuinely breakthrough problems and talent. This is a global problem — China, the U.S., Europe, Japan, and South Korea differ in form but are identical in essence.
Third, the outside-the-system geek community — stigmatized by mainstream academia as “pseudoscientists” — proves precisely that innovative capability is not the exclusive property of institutions. From Franklin to the garage computer revolution, from “Lorenzo’s Oil” to the deciphering of Ice Age lunar calendars, the people who change the world are often those not “inside” the system, not “recognized” by the system, but driven by the problem itself.
Fourth, AI is the counterforce that breaks this dimensionality reduction — not an accidental technological advance but an inevitable structural hedge. Its cross-domain nature, social-network-free nature, and funding-pressure-free nature form a precise mirror symmetry with every deficiency of the academic oligarchy system. The AI + Geek combination may be the most effective pathway for bypassing global academic bureaucracy and re-unleashing humanity’s innovative potential.
What truly impedes the progress of human research is not the complexity of the universe but the institutional walls humans have built themselves. When we hand the power to define “who is qualified to do science” to administrative bureaucrats and academic oligarchs, we have already pre-excluded those most likely to change the world — because they have never been “insiders.” Tearing down this wall is more important than any single scientific discovery.
References
- Bloom, N., Jones, C.I., Van Reenen, J. & Webb, M. “Are Ideas Getting Harder to Find?” American Economic Review, 110(4), 1104–1144 (2020).
- Park, M., Leahey, E. & Funk, R.J. “Papers and patents are becoming less disruptive over time.” Nature, 613, 138–144 (2023).
- Oxford Economics. “The State of Scientific Research Productivity.” (2021).
- Gold, E.R. “The fall of the innovation empire and its possible rise through open science.” Research Policy, 50(5), 104235 (2021).
- Ekerdt, L. & Wu, K.J. “Self-Selection and the Diminishing Returns of Research.” (2024).
- Economic Information Daily. “70% of Research Funding Allocated by Administrative Rank.” People’s Daily Online (2015).
- Springer. “Closing the door behind: metric-based research evaluation systems and gatekeeping towards young researchers.” Scientometrics (2025).
- Science. “South Korea Attacks Academic Inbreeding.” (1998).
- Initiative for Science in Europe. “Towards establishing funding schemes that promote academic job security for early career researchers.” (2024).
- Nature. “Japanese research is no longer world class — here’s why.” (2023).
- Tencent News. “How prevalent are academic ‘second-generation heirs’? New research reveals academic ‘inheritance’ dynamics.” (2024).
- CSDN Blog. “Exposing academic clan phenomena: what kind of ‘disciple wars’ are playing out in academic circles?” (2025).
- Zhang Wenjun. “A Critique of the Industrialization of Academia.” Aisixiang (2009).
- State Administration for Market Regulation (China). “Maintaining Fair Competition in Academic Literature Database Services.” (2022).
- Bruegel. “How much research talent could Europe grab from the US?” (2025).
- Delfanti, A. Biohackers: The Politics of Open Science. Pluto Press (2013).
- World Economic Forum. “3 incredible examples of citizen science in action.” (2023).
- Cambridge Archaeological Journal. Bacon, B. et al. “An Upper Palaeolithic Proto-writing System.” (2023).