In this day and age, AI and the understanding of how the mind works and has evolved over the ages is crucial. Getting how we went from unicellular organisms to the sophisticated creature that is building space ships and communicating over the internet permits to also steer how the AI could and should evolve over the next decade. Max Bennett does a great job walking us through this journey. Max explains through 5 breakthroughs what are the key milestones that brought us there.
Breakthrough #1 — steering
Steering emerged around six hundred million years ago in early organisms and allowed them to navigate their environment by categorizing stimuli as “good” or “bad.” This was a basic decision-making mechanism driven by chemical signals—early versions of what would become dopamine and serotonin. An organism could make simple choices: turn left or turn right? Is the potential reward of food worth enduring a painful obstacle to reach it? This breakthrough gave rise to fundamental experiences like pleasure, pain, satiation, and stress that still guide much of our behavior today.
Breakthrough #2 — reinforcing
Around five hundred million years ago, the first vertebrates—early fish—developed the ability to learn from experience and repeat behaviors that led to favorable outcomes while avoiding those that didn’t. Unlike steering, which was purely reactive, reinforcement learning allowed organisms to capitalize on past experiences and incorporate a more diverse set of parameters into their decision-making, including emotions like fear, curiosity, excitement, disappointment, and relief. This breakthrough also gave organisms a sense of time, allowing them to connect actions with their eventual consequences.
Breakthrough #3 — simulating
Around one hundred million years ago, early mammals developed the neocortex, which gave them the ability to mentally simulate actions before taking them. Instead of just reacting to their environment or learning from trial and error, these creatures could now imagine different scenarios in their minds. A squirrel, for instance, could visualize a path across tree branches and evaluate whether the route was safe and worth the effort before actually attempting it. This capacity for mental simulation enabled mammals to plan ahead, recall past experiences, and learn through imagination rather than just direct experience—a huge leap in cognitive efficiency.
Breakthrough #4 — mentalizing
Between ten and thirty million years ago, early primates developed an expanded neocortex that enabled mentalizing—the ability to understand intent and mental states in themselves and others. This wasn’t just about predicting what someone would do, but understanding why they were doing it. What is John trying to achieve? What’s his end goal? This capability allowed primates to anticipate not just their own future needs but also the motivations of others around them. Mentalizing was crucial for the emergence of complex social structures, as individuals could now live in larger groups, cooperate more effectively, and navigate social dynamics. It also transformed learning—primates could now learn by observation in a deeper way, not just mimicking actions but understanding the purpose behind them.
Breakthrough #5 — speech
Speech was the breakthrough that enabled the capitalization of knowledge across generations. Unlike the previous four breakthroughs, speech didn’t require a new brain structure—what’s unique to humans is the learning “program” that allows us to acquire language. The real transformation was that humans no longer needed to relearn everything from scratch in each generation. Knowledge could be efficiently transmitted and built upon, creating a sort of expanding global database of collective understanding. You didn’t need to personally witness someone getting poisoned by red berries to know they were dangerous—someone simply telling you was enough. This accumulation of knowledge also enabled specialization. Individuals could become masters of specific crafts, deepening humanity’s understanding of technology, manufacturing, and countless other domains, because they could build on the expertise of those who came before them.
Summary
Each breakthrough built upon the previous one, creating increasingly sophisticated forms of intelligence. Bennett argues that we may be witnessing the early signs of a potential sixth breakthrough with the emergence of large language models and artificial intelligence. Unlike biological intelligence, AI evolution won’t be constrained by the slow pace of biological change. Considering it took only a few billion years to go from single-cell organisms to human intellect, the question of what the next few billion years might bring—or even the next few decades—is both profound and exciting.
