Redefining Intelligence: Human Mind and AI

The hallowed halls of Harvard’s Berkman Klein Center recently played host to a debate so fundamental it challenges our very understanding of intelligence itself.

Forget metaphors; a new, audacious proposition is gaining traction: what if human intelligence isn’t just like computational intelligence, but is computational intelligence?

This isn’t merely academic posturing; it’s a paradigm shift that could redefine our relationship with artificial intelligence, and indeed, ourselves.

On September 24, 2025, the BKC’s Fall Speaker Series kicked off with a compelling address from Blaise Agüera y Arcas, a Vice President, Fellow, and CTO of Technology & Society at Google.

His message, echoing the core thesis of his new book, “What Is Intelligence? Lessons from AI About Evolution, Computing, and Minds,” was unequivocal: the brain, far from simply resembling a computer, is a computer.

Not one of mechanical gears and silicon chips, perhaps, but a biochemical computational entity nonetheless.

It’s a declaration that ripples through the scientific community, forcing us to confront a possibility once relegated to science fiction.

The assertion, as radical as it sounds, draws parallels with the rapid advancements in large language models (LLMs) like ChatGPT, Claude, and Gemini.

When you type a prompt into an LLM, your words are encoded into numerical tokens.

These tokens then traverse a vast artificial neural network (ANN), trained on trillions of words, meticulously pattern-matching how humans communicate.

The AI’s task is to predict, token by token, the most probable next word, assembling a coherent, often astonishingly human-like response.

It’s a sophisticated act of word prediction, driven purely by computational processes.

Proponents of the “mind-is-computation” premise, often embracing the predictive brain hypothesis, suggest our own grey matter operates on a remarkably similar principle.

Incoming sensory data—sounds, sights, sensations—are converted into internal biochemical “tokens.”

These tokens flow through our biological neural networks, predicting suitable outputs, which then manifest as intelligible thoughts or spoken words.

The implication is profound: our consciousness, our unique capacity for thought, could be an intricate, biological algorithm, running on nature’s most complex hardware.

This “computationalism” offers a clean, elegant framework for understanding the mind, aligning it with the venerable Church-Turing thesis and suggesting our current pursuit of AI is, perhaps, fundamentally on the right track.

This seductive idea—that human and artificial intelligence are essentially two sides of the same computational coin—has led many to believe that the path to Artificial General Intelligence (AGI) lies primarily in scaling up existing AI models.

Richard Sutton, a revered pioneer in AI, famously articulated this “bitter lesson” in 2019, arguing that “general methods that leverage computation are ultimately the most effective,” and that “the only thing that matters in the long run is the leveraging of computation.”

For years, the industry has largely followed this dictum, pouring billions into faster GPUs and larger data sets, witnessing impressive gains in LLM capabilities.

The logic seems undeniable: more computation equals better AI.

Yet, a growing chorus of dissenters warns that this singular focus might be a “primrose path” to an unyielding wall.

They question whether the current architecture of LLMs and generative AI, despite its apparent prowess, is truly capable of achieving AGI.

Is predicting the next token, however sophisticated, a sufficient foundation for genuine intelligence?

These critics argue that diverting all resources into scaling existing models risks a catastrophic misallocation of talent and capital, potentially slamming into a dead-end with no viable alternatives explored.

Intriguingly, even Sutton himself appears to be recalibrating his position.

In a recent podcast aired on September 26, 2025, he suggested the need for entirely new AI architectures beyond current LLMs and computational transformers.

His updated view posits that LLMs “fail to encompass ground truths” and that token prediction is not the proper goal for AGI, anticipating the eventual obsolescence of today’s dominant AI paradigms.

This pivot from a foundational figure underscores the urgency of exploring diverse approaches, rather than placing all our eggs in one computational basket.

The debate over whether the brain is a computer, or merely like one, opens up tantalizing avenues for discovery.

We stand before two “mighty unknowns”: how do 86 billion neurons and 100 trillion synapses give rise to human thought and consciousness?

And, conversely, how do the vast artificial neural networks of LLMs generate seemingly human-like cognition?

The beauty of the computational premise is that it suggests a shared methodology for deciphering both.

What we learn about the interpretability and explainability of AI – a crucial, emerging field dedicated to understanding AI’s internal logic – could be directly applied to demystifying the human mind.

Conversely, insights into biological cognition could illuminate the path for next-generation AI.

This intellectual symbiosis is not new.

The fields of psychology and AI have historically enjoyed a co-collaborative bond, each informing the other.

The call to action, then, is clear: whether you fervently believe in the brain’s computational nature or adamantly reject it, the pursuit of understanding must continue with an open mind.

As John F. Kennedy wisely noted, “Change is the law of life. And those who look only to the past or present are certain to miss the future.”

In the quest to unlock the secrets of both human and artificial intelligence, an open mind is truly worth its weight in gold.

The future of understanding intelligence depends on our willingness to question, explore, and adapt, rather than clinging to current assumptions, however comforting they may be.