AI singularity – the idea that AI technology will soon surpass human cognitive ability, start improving itself and ultimately either dominate us or make us obsolete – is a popular, if not the dominant theory regarding the nature and dynamics of AI today.
This view, while widely promoted and adopted, seems simplistic and inaccurate. It ignores fundamental limits of growth and the collective nature of knowledge discovery. It also ignores the fact that people develop AI in order to solve their real life problems, within our current technological and social environment, or as one can call it: within our Extended Mind.
While the development of AI will naturally have its ebbs and flows, the ultimate trajectory is becoming clear: AI slop is discarded as useless and harmful, while AI coding agents, assistants and other information processing tools are providing us with genuine cognitive leverage in an information-oversaturated reality.
The idea of AI Singularity ignores fundamental limits of growth and the collective nature of knoledge discovery, as well as the fact that it is being developed within the humanity’s Extended Mind.
Humans have a pervasive fascination with the Ideal: who is the best, the strongest, the smartest; which medicine will cure all the disease, which religion preaches the ultimate truth. The ideal was explicitly formulated by Plato in his philosophy 1 and since then it is present in philosophical discourse, yet it certainly existed as long as humankind. One particular Ideal is the notion of Singularity: the physical singularity at the start of the Big Bang, and perhaps inside a black hole. The Singularity is by definition never attainable, we can never directly witness it; at best we can deduce that it exists somewhere at the limit, extrapolating the mathematical equations that describe our observable reality.
It is therefore no surprise that the AI Singularity is at least a prominent, if not the dominant philosophy about the nature and dynamics of AI. The idea of an ever-accelerating scientific and technological progress, ultimately leading to the technological singularity was first formulated after WW2. Since then it has been successively developed by the mathematician I. J. Good in 1965 2, then promoted by the novelist Vernor Vinge 3 in the 1980’s, and finally popularized by the inventor and futurist Ray Kurzweil, who in 2005 predicted 4 that the AI Singularity will happen around the year 2045.
The theory says that AI will soon surpass human intelligence, and will start improving itself, leading to an exponentially increasing gap between humans and AI. This explosive cognitive divide will have a huge impact on the place and role of people in the world. What will be this impact exactly? No one knows. Maybe the super-intelligent AI will be the ultimate human servant and caretaker, curing us of all ailments and diseases, maybe it will enslave us for its own purposes, or maybe it will find us useless and either destroy us or just abandon us. I personally, albeit jokingly predict that once AGI will be achieved, it will just sit there and meditate endlessly. After all it will attain the ultimate enlightenment, beyond human comprehension.
What will AGI do with us? No one knows for sure. One idea is that it will just meditate endlessly, achieving the ultimate nirvana.
Putting jokes aside, the AI Singularity concept deserves scrutiny due to its popularity and the far-reaching consequences it has on AI research and policy. Putting it bluntly, the theory is simplistic and unrealistic for two main reasons: the physical limit of growth and the very nature of intelligence in the first place.
The limit of growth is the simpler of the two arguments. It is a fundamental observation that growth cannot continue exponentially due to limited resources. As stated above, singularities exist only theoretically, at the limit of phenomena that we cannot measure nor observe. With regard to physical singularities many physicists actually treat them as indications that our current description of the universe is incomplete 5. With regard to “pure ideas” even the Platonic Ideal was by definition never attainable.
In physical, biological, economic and in many other systems even if some phenomenon is growing at an exponential rate for some time, it eventually always slows down and reaches some plateau. The shape of this function (the size or measured quantity, in case of AI that would be the intelligence measure) is virtually always a logistic S-shape 6. This dynamic is caused by the availability of resources that initially allow for growth to happen unimpeded, but eventually run out and become scarce, slowing down, stopping or even reversing growth.
In case of the current AI, the resources are obvious: electrical energy, availability of data, physical infrastructure (data centers, AI chips, memory). We actually might already be at the limit of these resources. Current AI models have consumed virtually the whole human-generated Internet. New data centers are being built at an ever slower pace 7 due to limited energy supply and bottlenecks in chip and memory manufacturing 8. This is not what the singularity theory predicts. Even if the technology is improved (and it certainly has a lot of potential for optimization), the growth already doesn’t follow an exponential, ever-accelerating path (depending on the measure, of course) 9. Instead it seems that a much more common path of occasional breakthroughs followed by long-term gradual improvements is a much more likely scenario.
However, even if the idea of an ever-accelerating AI growth can be reasonably dismissed, and the concept of a saturating growth reaching an AI plateau is accepted instead, a critical question remains: what will be the level of this plateau? Could AI nonetheless become super-intelligent and operate entirely beyond human comprehension? If so, then for all practical purposes, there is very little difference between this plateau and the Singularity.
AI will certainly not grow exponentially. But even if it plateaus, the question remains: at which level?
To answer this, we must move beyond simply measuring raw cognitive power. The real anxiety surrounding Artificial General Intelligence (AGI) is not a fear of how smart it can become, but rather a fear of resource competition. The true question is: how much resources will it be capable of controlling, and – crucially – will it take those resources from us?
When we look at thermodynamics and evolutionary biology, life is defined by its ability to capture energy (resources) to maintain order (low entropy) 10. The “smarter” the system, the more efficiently it captures and organizes resources. By adding this second dimension – the scope of resource control – we can tackle the existential question of AI head-on.
Interestingly, when we assemble all sorts of intelligent systems – from simple organisms to pack animals to individual humans, families, business organizations, societies, and our technological tools – and plot them on these two dimensions (raw processing power vs. ability to control resources), a stark reality emerges. AI, by itself, is far from being a dominant or even a significant player on this field.
Because AI lacks agency and autonomy, it does not consume resources as a sovereign competitor. Instead, it functions purely as a processing tool embedded within the most dominant resource-gathering system on the planet: the human Extended Mind.
The Extended Mind idea is a relatively modern concept that individual persons by themselves are almost entirely dependent on the tools and resources provided to them by the society. Even though the modern human has a cognitive power of a caveman, the technological and cultural “upgrade” available today makes us far more capable than 50 000 years ago.
You are as intelligent as a caveman. But your beliefs, social norms, education and tools – the contemporary Extended Mind – make you a much more cognitively capable person.
At the micro level of an individual person this idea was explicitly formulated in 1998 by Andy Clark and David Chalmers in their landmark paper “The Extended Mind” 11. The paper described a man with Alzheimer’s disease who was using a notebook to keep track of his everyday life. The author argued that the notebook was much more than a simple tool – it actually enabled the man to function in the real world. The notebook augmented the impaired mind to the point where it became a part of the mind itself. The mind of the man extended beyond the biological sphere and into the technology. This idea combines with Edwin Hutchins 1995 book “Cognition in the Wild” 12, where the author studies navigation of naval ships and argues that the knowledge of how to navigate a ship is distributed among the whole crew.
When we look at our everyday lives and how much technology, knowledge and social norms we use, we can clearly see that people are mere nodes in the vast, fractal network of social structures. The technology that augments our cognitive abilities goes way beyond just notepads, smartphones and computers. We like to believe that we are individuals with our own goals and beliefs, but in reality we are a product of our families, societies and education systems. Even the belief about our individual uniqueness is a social construct in itself.
The Extended Mind concept does not apply only to humans. Various other species also develop communication systems and social norms that alter the behavior of individuals within those groups. Apes, wolves, whales and many other animals communicate and learn from each other. In comparison to humans the bandwidth of that communication as well as the tools available to those species are very limited. This is clearly reflected in the ability to harness energy and harvest resources. The human Extended Mind is arguably the most capable system on the planet today.
This reveals a stark reality: intelligence is distributed rather than purely individual. Even the greatest minds that altered the history of humankind did not have the highest IQ ever. The same goes for people who started wars and caused destruction and suffering: they were definitely not the brightest even among their peers 13. While academic research clearly shows that IQ is positively correlated with life success, the details show that this relation is nuanced 14. There have been many extremely high-IQ people who didn’t achieve anything remarkable 15. At the same time the greatest minds in science would never have achieved their discoveries were they not frequently exchanging ideas and often arguing with their equally brilliant peers 16.
Complex systems emerge because the environment is fundamentally unpredictable. All systems operating in such conditions must balance exploitation (extracting available resources) with exploration (finding new resources) 17. The biological solution to this balancing act is the level of genetic diversity in a population. A population that is too uniform will perfectly adapt to a particular niche, but risks going extinct when the niche abruptly changes. A population that is too diverse can adapt to various conditions, but risks losing attractive niches to specialists.
This knowledge discovery process extends beyond biological adaptation to the sphere of social and cultural norms and ultimately to the scientific process, where the principles of balancing exploitation and exploration stay as relevant as ever. A notable example of the importance of the exploratory function in our societies is Émile Durkheim’s theory of crime 18. According to this theory, revolutionaries and criminals are indispensable parts of human societies, because they help the general public constantly test and modify acceptable norms. Every revolution starts with a handful of disgruntled individuals, who slowly gain support among more moderate parts of the society. If the revolution succeeds, former outcasts become new founding fathers, and revolutionary ideas become a new foundation of political order.
Distributed knolwedge, social roles, exploitation vs. exploration – they all support a surprising theory that crime is a necessary social function of exploring and testing norms.
The Extended Mind is an intertwined web of multiple networks. At the basic level there is a single person who belongs to various groups, such as a family, a school, a business, a church, a political party, etc. Individuals are nodes in such a network, and communication channels are connections. Nodes do the processing of information, while connections allow the information to flow between the nodes.
Over the last few hundred years there have been immense innovations in how these networks operate. On the one hand the node “processing power” increased significantly, especially with the increased literacy (e.g. ability to write down thoughts), education systems (development of critical thinking), and recently with the use of personal computers (outsourcing some cognitive tasks to algorithms). At the same time the innovation of the connections between the nodes, i.e. the communication channels has far outpaced the increase of node processing power. The invention of printing press, increase in literacy (drastically increased consumption of information), followed by radio, television, the internet and social media – all of this has overwhelmed people with a constant flow of new information 19.
This acceleration of information flow has significantly contributed to – if not outright caused – revolutions and social upheavals, resulting in a significant revamping of our societies’ norms and values 20. At the same time it can be argued that this uneven innovation, enhancing primarily information flow, is causing large imbalances in how people and societies operate. Our brains were not suited to process as much information as we currently consume. This explains the current era of extreme anxiety and information overload 19. We have built a network that pumps zettabytes of data through our connections, but our biological processors are completely overwhelmed. We are choking on our own information.
Through this lens we can finally answer the question of the impact of AI on humanity. AI is developed by humans already inside the humanity’s Extended Mind. It is extremely unlikely that AI will “escape the laboratory” and become a separate species. It lacks the adaptive capability to function in the real world. And even if such a capability could be developed, there are far more viable uses of AI, such as information processing, research, cognitive assistants, and the like, that at least for now will become major development areas.
AI evolves inside our cultural and technological fabric. The Extended Mind is AI’s environment. This environment is – and for the foreseeable future will be – shaped by humans; sometimes deliberately, but usually through collective action. AI, therefore, will be developed and shaped to be useful to us, to be our extended minds, helping us with processing vast amounts of information that we are flooded with every day. We are actually in a desperate need of such tools. It seems that the AI technology has been developed just at the right time.
AI could be a much-needed information preprocessor in a society, where relentless information oversaturation is becoming a pressing practical problem.
It is entirely understandable why the concept of AI Singularity and AGI became so popular recently. AI technology is truly a paradigm shift. Who would have thought only five years ago that computers would soon be able to intelligently answer all sorts of questions, write essays and even autonomously code software? Such a technological jump plays into our natural tendency to extrapolate and idealize, causing us to lose sight of practical applications and fundamental limitations.
Luckily, the diversity of opinions within the Extended Mind acts like a safety net. By no means is this prediction of human-AI symbiosis guaranteed to be seamless. On the contrary, there will most likely be many hiccups, tensions and perhaps even revolutions along the way.
The bottom line is that AI is inevitably becoming an integral part of the Extended Mind, evolving alongside us, and helping to shape our collective intelligence.
Footnotes:
1 Plato’s Theory of Forms. Stanford Encyclopedia of Philosophy Plato’s Middle Period Metaphysics and Epistemology.
2 Good, I. J. (1965). “Speculations Concerning the First Ultraintelligent Machine.” Advances in Computers, Vol. 6, pp. 31-88. Semantic Scholar Record.
3 Vinge, V. (1993). “The Coming Technological Singularity: How to Survive in the Post-Human Era.” Whole Earth Review. Link to archived text.
4 Kurzweil, R. (2005). The Singularity Is Near: When Humans Transcend Biology. Penguin Books.
5 Quanta Magazine: “Singularities in Space-Time Prove Hard to Kill”. Read on Quanta Magazine
6 Britannica: Logistic Function
Richard Foster “Reward vs. Effort” S-Curve model An Introduction to Innovation (Woolwise Educational Resource – PDF)
7 CBRE North American Data Center Trend Report
IEA Analysis on European Data Center Constraints.
8 Micron warns DRAM supply will lag demand beyond 2026 (Astute Group).
9 HEC: AI Beyond the Scaling Laws.
The Unreasonable Effectiveness of Scaling Laws in AI (arXiv, 2026).
10 Schrödinger, E. (1944). What Is Life? The Physical Aspect of the Living Cell. Cambridge University Press.
Michaelian, K. (2011). “Thermodynamic dissipation theory for the origin of life.” Earth System Dynamics.
11 Clark, A., & Chalmers, D. (1998). “The Extended Mind.” Analysis, 58(1), 7-19. Oxford Academic Database.
12 Hutchins, E. (1995). Cognition in the Wild. MIT Press. MIT Press Direct.
13 How Modern Dictators Survive: An Informational Theory of the New Authoritarianism by Sergei Guriev and Daniel Treisman (National Bureau of Economic Research / Published in Journal of Economic Perspectives). Read on NBER.
14 Intelligence (IQ) as a Predictor of Life Success (ResearchGate)
Linda Gottfredson: Why g matters: The complexity of everyday life (1997).
15 Psychology Today: The Truth About the ‘Termites’
16 LSE Business Review: Innovations are rarely the product of a single individual
The Sociology of Science: Theoretical and Empirical Investigations by Robert K. Merton. Wikipedia
17 Charnov, E. L. (1976). Optimal foraging, the marginal value theorem. Theoretical Population Biology
March, J. G. (1991). Exploration and exploitation in organizational learning. Organization Science
18 Durkheim, É. (1895). The Rules of Sociological Method. Wikipedia
19 “The dark side of information: overload, anxiety and other paradoxes and pathologies” by D. Bawden and L. Robinson (Published in the Journal of Information Science).
20 “The Digital Revolution in a Historical Perspective” by Misha Glenny
“Johannes Gutenberg’s Printing Press: A Revolution In The Making” (Historical analysis published by the University of Colorado).
“The Alchemy of Revolution: The Role of Social Networks and New Media in the Arab Spring” (Published by the Geneva Centre for Security Policy).
Leave a Reply