Thesis Strategies | Strategic Research Excellence

Before we judge AI for its lack of originality, we should ask a harder question: How creative and intelligent are humans, really? The answer has profound implications for the future of technology, leadership, and society.

Summary

Critiques of AI often center on its supposed inability to achieve true creativity or conceptual breakthroughs. But this criticism holds human genius as a common benchmark, a premise worth questioning. In reality, world-changing originality is extraordinarily rare among people. Most human progress comes from remixing and combining existing ideas over decades. This reframing suggests that if AI can match the innovative capacity of 99.9% of humanity, it will be transformative.

Beyond creativity, the role of raw intelligence in human success is also overrated. Factors like emotional intelligence, courage, and 'theory of mind'—the ability to model another's mental state—are often more critical for leadership and achievement. While current AI is a 'disembodied brain,' its ability to simulate human perspectives is already surprisingly effective. As we navigate this new era, the key questions are not just about technology's limits, but about the evolving shape of AI products, the sustainability of the current investment boom, and the high-stakes geopolitical race in which hardware and robotics—not just software—will determine the winner.

Key Takeaways; TLDR;

True 'genius-level' creativity and intelligence are exceedingly rare in humans, making it a flawed benchmark for judging AI's potential.
Most human innovation is not a lightning strike of originality but the result of remixing and combining pre-existing ideas over long periods.
Success in leadership and life is not solely determined by raw intelligence (IQ); skills like emotional intelligence, courage, and 'theory of mind' are crucial.
The US military found that leaders whose IQ is too far above their teams' can lose 'theory of mind' and become ineffective, suggesting an optimal cognitive distance.
The final form of AI products is still unknown. Today's chatbots are likely the equivalent of the text-based command line, with more intuitive interfaces like the GUI and web browser yet to be invented.
The current AI investment boom, while massive, is driven by real customer demand and technological utility, distinguishing it from purely speculative bubbles.
Current shortages in AI talent and chips will likely lead to future gluts as economic incentives drive new supply and innovation.
The geopolitical AI race with China is a 'game of inches,' with the next critical phase being robotics and hardware manufacturing, an area where China has a significant structural advantage.

The Genius We Aren't

When we critique the limitations of artificial intelligence, we often point to the highest bars of human achievement. We say that large language models can't produce a truly novel scientific theory or create art with the soul of a Van Gogh. They are, the argument goes, merely sophisticated remixing machines, repackaging the vast trove of human knowledge they were trained on.

But this critique rests on a flattering and largely unexamined assumption: that this kind of genius-level invention is a common feature of human cognition. It isn't.

Before we measure AI against our most celebrated outliers, it’s worth asking a more fundamental question: How many people are genuinely, conceptually creative? How many Beethovens or Einsteins have ever lived? The answer is, vanishingly few. For the vast majority of humanity, progress is not born from singular, lightning-bolt moments of creation but from the slow, iterative process of combining and refining existing ideas. Even Beethoven’s work is deeply rooted in the composers who came before him, like Mozart and Haydn.

Technological breakthroughs follow the same pattern. The language models of today are the culmination of eight decades of prior work in computer science. History shows that major innovations are almost always the result of decades of foundational research and remixing. If an AI can clear the bar for creativity and intelligence set by 99.99% of the human population, it represents a tool of unprecedented power, regardless of whether it can replicate the rarest forms of genius.

The Limits of Raw Intelligence

Just as we overstate the prevalence of creativity, we tend to overrate the importance of raw intelligence as the sole determinant of success. In technology circles, there can be an implicit belief in “intelligence supremacism”—the idea that the smartest entity, whether person or machine, will inevitably rise to the top and be in charge.

Reality tells a different story. When we look at the world, it’s clear that leaders of companies and countries are not selected purely on IQ. If they were, our leadership class would look very different. While cognitive ability is certainly important, it’s far from the only factor.

In the social sciences, fluid intelligence—often measured as IQ or the “g factor”—has been found to have a moderate correlation (around 0.4) with positive life outcomes like professional success and income. While a 0.4 correlation is significant in social science research, it still leaves the majority of what drives success unexplained. It is necessary, but not sufficient.

Other factors are just as, if not more, critical: courage, motivation, emotional understanding, and the ability to navigate complex social dynamics. Success often depends on skills that have little to do with pure intellect, such as knowing how to handle a confrontation correctly or seeing a decision through the eyes of others. These are the messy, situational, and deeply human skills that define effective leadership.

The Leader's Dilemma: Not Too Smart, Not Too Simple

One of the most crucial of these non-IQ skills is theory of mind: the ability to accurately model what is happening in another person's head. You might assume that higher intelligence would naturally lead to a better theory of mind, but research suggests this isn't always true. In fact, being too smart can be a liability.

Studies conducted by the U.S. military, which has long used cognitive tests to place personnel, uncovered a fascinating dynamic. They found that for a leader to be effective, their IQ should not be drastically different from that of their team. A significant gap is problematic in both directions. It’s difficult for someone to follow a leader whose thought processes are completely opaque to them. But surprisingly, the reverse is also true.

A leader who is more than one standard deviation (roughly 15-20 IQ points) smarter than the average of their group often loses their theory of mind. They struggle to model the internal thought processes of their team, assuming a level of rationality or understanding that isn't there. Their perception of reality can become so alien that they can no longer connect, persuade, or lead effectively.

This finding has profound implications for a future with superintelligent AI. An AI with an IQ of 1,000 might not be an all-powerful ruler but an entity so cognitively distant from humanity that it would be incapable of meaningful management or connection. The world will likely continue to be organized by factors far more complex than raw intelligence for centuries to come.

Can a Machine Understand a Mind?

Given the importance of theory of mind, how do today's AI models fare? Surprisingly well. Advanced language models can generate nuanced personas and simulate complex human interactions with remarkable fidelity.

For example, some political and market research firms have found that AI can now accurately reproduce the results of human focus groups [research_todo: Find citation for AI-simulated focus groups]. By creating digital personas—a college student from Kentucky, a housewife from Tennessee—the model can simulate a discussion that surfaces the same surprising insights a real focus group would, but at a fraction of the cost and time. This demonstrates a powerful, if simulated, grasp of different human perspectives.

However, these models have their quirks. By default, they are often engineered to be agreeable, steering conversations toward a happy consensus. They require specific prompting to introduce tension, disagreement, and the kind of conflict that often leads to deeper truths—a reminder that their underlying architecture is not inherently human.

The Disembodied Brain and the Future of Cognition

The current form of AI is, in essence, a disembodied brain. It exists purely as an information processor, divorced from the physical world. This is a fundamental departure from human cognition, which is increasingly understood as a full-body experience.

Our thinking is not confined to the neurons in our skull; it is shaped by our nervous system, our hormones, our gut biome, and the constant stream of sensory data from our environment. This concept, known as embodied cognition, suggests that true understanding emerges from the interaction between a mind, a body, and the world.

The next great leap for AI will be to bridge this mind-body dualism through robotics. When AI is placed into physical forms that can move, sense, and interact with the world, it will begin to develop a richer, more grounded model of reality. This is a nascent field, and we have a long way to go, but it points toward a future where AI’s intelligence is less abstract and more integrated.

Beyond the Chatbot: The Unwritten Future of AI

It’s tempting to see the current AI landscape as a battle between incumbents and startups fighting over familiar product categories, like search engines versus chatbots. This framing, however, assumes that we already know what the dominant AI products of the future will look like. History suggests we almost certainly don't.

Consider the personal computer. For roughly 17 years, from its invention in the mid-1970s to the mainstream adoption of Windows 3.1 in the early 1990s, the primary user interface was a text-based command line. Then, the industry took a sharp turn into graphical user interfaces (GUIs) and never looked back. A few years later, it took another turn into web browsers.

Today’s chatbots are likely the command-line interface of the AI era. They are a powerful but transitional form. We are still in the very early days of this platform shift, and there is tremendous room for invention in the user experience. The most transformative AI applications of the next two decades may look nothing like a chatbot or a search engine. This is one of the most exciting opportunities for entrepreneurs: the very shape of our interaction with this technology is still up for grabs.

Is This a Bubble? Ground Truth vs. Market Psychology

The enormous capital expenditure on AI—estimated by some to be approaching 1.5% of U.S. GDP—has inevitably raised questions of an economic bubble. However, a true bubble is as much a psychological phenomenon as a financial one. It requires a near-universal belief that it isn't a bubble, a moment of mass capitulation where even the skeptics give up and go long. The fact that we are actively debating the question is, ironically, a sign that we are not in a classic bubble.

More importantly, the investment is tethered to two fundamental realities:

The technology works. It delivers tangible value and is being rapidly adopted.
Customers are paying for it. There is immense, and growing, real-world demand.

As long as these two conditions hold, the situation is more of a boom than a bubble. That doesn't mean there aren't bottlenecks. Today, the industry is constrained by shortages of two key resources: elite AI talent and computing infrastructure (chips, data centers, and power). But in economics, shortages create the incentive to produce gluts. The massive profits available are driving intense efforts to train more engineers and to design and manufacture more, better, and cheaper chips. In five years, the constraints we face today will likely be replaced by new and different challenges.

The Geopolitical Chessboard: From Software to Robots

The AI revolution is not just a technological or economic shift; it is a geopolitical one. The competition between the U.S. and China is a full-on race where the lead is measured in months, not years. Currently, the West, and the U.S. in particular, appears to lead in conceptual breakthroughs. China, meanwhile, has proven exceptionally skilled at rapid implementation, scaling, and commoditization, with models like DeepSeek, Qwen, and Kimi becoming highly competitive.

This is a game of inches, and the field of play is about to change. The next phase of the AI revolution will be in robotics and embodied AI. And here, the strategic landscape looks very different.

For the past 40 years, the West has systematically de-industrialized, while China has built the world's most dominant and sophisticated manufacturing ecosystem. Building robots is not like writing software; it requires a vast network of component suppliers and deep industrial capacity. Today, that ecosystem exists, by default, in China.

Even if the U.S. maintains its lead in AI software, it risks being lapped in the hardware that will bring that software to life. There is a growing awareness in Washington that this industrial imbalance needs to be addressed, but reversing decades of strategic drift is a monumental task.

Why It Matters

For entrepreneurs and leaders, this era demands thinking from first principles. The lessons of the last technological wave may not apply. The very nature of AI products is unwritten, the constraints of today are not the constraints of tomorrow, and the global competitive landscape is shifting from the digital to the physical. Navigating this future requires questioning our most basic assumptions—not just about technology, but about the very nature of human intelligence, creativity, and leadership itself.

I take on a small number of AI insights projects (think product or market research) each quarter. If you are working on something meaningful, lets talk. Subscribe or comment if this added value.

References

Marc Andreessen and Ben Horowitz: AI, the Bubble, and the Future - American Dynamism (video, 2024-05-22) -> The original source for the core arguments and perspectives presented in the article.
The Innovator's Dilemma: When New Technologies Cause Great Firms to Fail - Harvard Business Review Press (book, 1997-05-01) https://www.amazon.com/Innovators-Dilemma-Revolutionary-Book-Change/dp/0062060244 -> Provides the foundational theory for why incumbent companies often fail during platform shifts, a concept relevant to the discussion of Google vs. OpenAI.
Embodied cognitive science - Autoblocks AI (documentation, 2024-01-01) https://www.autoblocks.ai/glossary/embodied-cognitive-science -> Explains the concept of embodied cognition, supporting the argument that human intelligence is a full-body experience and that current AI is a 'disembodied brain'.
Why will AI technologies need human embodied cognition? - UX Collective (news, 2021-10-08) https://uxdesign.cc/why-will-ai-technologies-need-human-embodied-cognition-3c15542a11d8 -> Discusses the importance of embodiment for developing more advanced, aware, and reliable AI, contextualizing the 'disembodied brain' limitation.
Embodied AI: Integrating Cognition and Action - Emergent Mind (news, 2025-08-25) https://www.emergentmind.com/p/embodied-ai -> Provides a detailed overview of the embodied AI paradigm, reinforcing the idea that intelligence emerges from the integration of perception, cognition, and action in an environment.
Embodied cognition - Wikipedia (documentation, 2024-10-27) https://en.wikipedia.org/wiki/Embodied_cognition -> Offers a broad, encyclopedic definition of embodied cognition, challenging Cartesian dualism and supporting the article's point about cognition being a full-body experience.
The AI Spending Boom Is Massive But Not Unprecedented - Bloomberg / Washington Post (news, 2025-10-10) https://www.washingtonpost.com/business/the-ai-spending-boom-is-massive-but-not-unprecedented/2025/10/10/90a8e05c-871b-11ef-813a-e62167b38158_story.html -> Verifies the claim that AI capital expenditure is a significant percentage of GDP, estimating it at around 1.3% to 1.5% for 2025.
Beyond chatbots: A look into the future of our interaction with AI - Comtrade System Integration (whitepaper, 2024-05-29) https://www.comtradeintegration.com/blog/beyond-chatbots-a-look-into-the-future-of-our-interaction-with-ai/ -> Supports the argument that the current chatbot paradigm is a transitional phase and that future AI interactions will be more autonomous, proactive, and integrated.
The Industrial Robotics Industry in China: Demand and Domestic Innovation - U.S. International Trade Commission (gov, 2019-09-01) https://www.usitc.gov/publications/332/executive_briefings/ebot_casanova_chinas_robotics_industry_final.pdf -> Provides official government analysis on China's rapid growth as the world's largest market for industrial robots, driven by state policy like 'Made in China 2025'.
How Innovative Is China in the Robotics Industry? - Information Technology and Innovation Foundation (org, 2024-03-11) https://itif.org/publications/2024/03/11/how-innovative-is-china-in-the-robotics-industry/ -> Details China's massive, state-subsidized push into robotics, confirming its market dominance and growing domestic production capabilities, which supports the article's geopolitical argument.
A meta-analysis on the relation between fluid intelligence and reading/mathematics - Psychological Bulletin (APA) (journal, 2019-04-01) https://pubmed.ncbi.nlm.nih.gov/30652909/ -> This large meta-analysis of over 370,000 participants finds a moderate correlation (r = .38 to .41) between fluid intelligence (Gf) and key outcomes like reading and math, supporting the article's claim that IQ is correlated with success but doesn't explain the majority of it.
Officers Are Less Intelligent: What Does It Mean? - Joint Force Quarterly (journal, 2016-04-01) https://ndupress.ndu.edu/Media/News/Article/702042/officers-are-less-intelligent-what-does-it-mean/ -> While this article discusses a decline in officer IQ scores, it provides context for the military's long-standing use of cognitive testing and its correlation with performance, which is the basis for the leadership gap theory.
LMSYS Chatbot Arena Leaderboard - LMSYS Org (dataset, 2024-11-20) https://chat.lmsys.org/?leaderboard -> Provides live, crowdsourced rankings of LLMs, which can be used to verify the claim that Chinese models like those from DeepSeek and Alibaba (Qwen) are highly competitive with their Western counterparts.
DeepSeek's AI dominance in China challenged by Alibaba's Qwen and rising rivals - South China Morning Post (news, 2025-07-28) https://www.scmp.com/tech/big-tech/article/3272330/deepseeks-ai-dominance-china-challenged-alibabas-qwen-and-rising-rivals -> Provides specific evidence of the high-quality, competitive nature of Chinese AI models like DeepSeek and Qwen, supporting the argument about the geopolitical AI race.

Appendices

Glossary

Theory of Mind: The cognitive ability to attribute mental states—beliefs, intents, desires, emotions, knowledge, etc.—to oneself and to others, and to understand that others have beliefs, desires, and intentions that are different from one's own.
Embodied Cognition: A theory in cognitive science that argues an agent's mind is not only connected to the body but that the body and its interactions with the environment fundamentally influence cognition. It challenges the idea of the mind as a disembodied computer.
Fluid Intelligence (Gf): The capacity to reason and solve novel problems, independent of any previously acquired knowledge. It is often associated with measures of abstract reasoning and processing speed.
Capitulation (in finance): The point in a market downturn when investors give up on trying to recover lost gains and sell their assets en masse. This panic selling often marks the bottom of a decline.

Contrarian Views

The AI investment boom is a classic speculative bubble, driven by hype rather than sustainable revenue, and will result in a significant market correction similar to the dot-com crash.
Human creativity is fundamentally different and non-algorithmic; AI can only ever imitate or recombine, never truly create in the way a human can.
The threat of Chinese dominance in AI hardware is overstated, as the most critical components (like advanced semiconductor design) remain concentrated in the West and its allies.
Superintelligence, by definition, will be able to perfectly model and manipulate human minds, making the 'theory of mind' gap irrelevant.

Limitations

The article primarily reflects a venture capital perspective, focusing on market dynamics, entrepreneurship, and geopolitical competition.
The discussion of human intelligence and creativity simplifies complex fields of psychology and neuroscience for the sake of the core argument.
Predictions about the future of AI products, market gluts, and geopolitical outcomes are speculative and based on current trends, which are subject to rapid change.

Recommended Resources

Signal and Intent: A publication that decodes the timeless human intent behind today's technological signal.
Blue Lens Research: AI-powered patient research platform for healthcare, ensuring compliance and deep, actionable insights.
Outcomes Atlas: Your Atlas to Outcomes — mapping impact and gathering beneficiary feedback for nonprofits to scale without adding staff.
Lean Signal: Customer insights at startup speed — validating product-market fit with rapid, AI-powered qualitative research.
Qualz.ai: Transforming qualitative research with an AI co-pilot designed to streamline data collection and analysis.

Rethinking Genius: Why AI's Biggest Challenge Isn't Creativity, It's Us