Evolution of Human Intelligence
Human Cognitive Breakthroughs and Their Parallels in AI Advancements
I recently enjoyed A Brief History of Intelligence: Evolution, AI, and the Five Breakthroughs That Made Our Brains by Max Bennett, an AI entrepreneur with a deep interest in neuroscience. The book explores the evolution of human intelligence and its connection to the advancements in artificial intelligence (AI). The book claims that in order to advance AI we must deeply understand the evolutionary history of our own brains.
To highlight this point, the book opens with the question “Why is it that AI can crush any human on earth in a game of chess but can’t load a dishwasher better than a six-year-old?”.
The book is structured around five major evolutionary breakthroughs, each representing a critical milestone in the development of intelligence.
Steering: Early brain intelligence evolved to enable animals to navigate their environment.
Reinforcement Learning: Learning through trial, error, and reward feedback.
Simulation: The neocortex enabled mental simulation that was used for planning and completing complex tasks.
Mentalizing: The ability to understand others' thoughts and intentions was key to social behaviors and learning.
Language: Language transformed human intelligence by enabling knowledge accumulation and transmission across generations.
Even though I suspect some of the connection between brain evolution and anticipated AI evolution may be overly simplified, I still found the scientific review of the evolution of intelligence interesting. I’ve shared some more detail on each breakthrough below.
The Five Breakthroughs
1. Steering
The first breakthrough was the evolution of the primitive brain — an adaptation that enabled early animals to navigate their environment. Before the development of brains, early life forms relied on basic reflexes and chemical signals to interact with the environment. However, as organisms evolved, so did more complex nervous systems, eventually leading to the emergence of centralized brains.
These early brains enabled organisms to process sensory information and make more deliberate, controlled movements — a concept referred to as “steering.” Steering is more than just physical movement; it represents a fundamental cognitive function that allows organisms to guide their actions based on both internal and external cues.
The ability to steer was particularly significant in animals with bilaterally symmetrical bodies, as it allowed them to navigate their surroundings with greater precision. This evolution was not merely mechanical but intellectual, driving the development of more complex neural structures. As these animals evolved, so did their brains, which played a critical role in enabling more sophisticated forms of navigation and goal-directed behavior.
This first breakthrough laid the foundation for more advanced forms of intelligence, moving from simple reactive behaviors to complex higher level intelligence.
2. Reinforcement
The evolution of reinforcement mechanisms marked another significant leap in intelligence. Through neurotransmitters like dopamine, animals learned to adapt their behaviors based on the outcomes of their actions. “Dopamine is not a signal for pleasure itself; it is a signal for the anticipation of future pleasure.” This distinction is crucial for understanding reinforcement in both biological and artificial intelligence.
In biological systems, reinforcement is a powerful driver of behavior. Animals, including humans, learn by first performing random exploratory actions and then adjusting future actions based on the outcomes — positive outcomes reinforce behaviors, while negative outcomes discourage them.
Reinforcement learning is further broken down into two components: the actor and the critic. The actor makes decisions, while the critic evaluates those decisions based on the likelihood of success, not actual rewards. This method, known as temporal difference learning, mirrors the way dopamine functions in the brain.
For reinforcement learning to work, reinforcement and reward must be decoupled in this manner. This decoupling allows animals — and AI systems — to adapt and learn over time.
The basal ganglia play a crucial role in this process by learning to repeat actions that maximize dopamine release, while the hypothalamus determines actual rewards. Over time, the basal ganglia learn to predict rewards, adjusting behaviors accordingly. This system of reinforcement learning is not just about pursuing rewards but about navigating the complexities of the environment with increasing sophistication. Notably, dopamine is not a signal for reward but for reinforcement.
3. Simulation
Simulation, enabled by the neocortex, was the third breakthrough in the evolution of intelligence. Early mammals, equipped with the ability to simulate future possibilities and remember past events, gained a significant cognitive advantage. This ability allowed them to perform vicarious trial and error, mentally exploring different scenarios before taking action.
The neocortex plays a role in enabling imagination: the ability to consider future possibilities and relive past events. This capability was a game-changer, as it allowed mammals to plan, perform fine motor skills, and execute complex tasks that required foresight and flexibility.
The neocortex’s power to simulate also ties into the development of pattern recognition, attention, and working memory. These cognitive abilities, while still challenging for AI to replicate, are foundational to human intelligence and have significant implications for the development of more sophisticated AI systems.
“The secret to dishwashing robots lives somewhere in the motor cortex and the broader motor system of mammals. Just as we do not yet understand how the neocortical microcircuit renders an accurate simulation of sensory input, we also do not yet understand how the motor cortex simulates and plans fine body movements with such flexibility and accuracy and how it continuously learns as it goes.”
4. Mentalizing
Mentalizing, or the ability to infer the thoughts and intentions of others, was the fourth breakthrough. This capability allowed early primates to engage in complex social behaviors, such as politics and imitation learning. The book discusses how mentalizing enabled the development of theory of mind, a crucial skill for social interaction and learning.
“Teaching is possible only with theory of mind. Teaching requires understanding what another mind does not know and what demonstrations would help manipulate another mind’s knowledge in the correct way.”. This breakthrough had profound implications for the evolution of social structures and cooperation, setting humans apart from other species in their ability to learn and transmit knowledge.
The ability to reason about your own mind and reason about other minds is the same process. And this metacognition — the ability to think about thinking — plays a crucial role in learning new skills.
“Acquiring novel skills through observation required theory of mind, while selecting known skills through observation did not.”
The theory of mind enables novices to identify the intent of a complex skill, which can motivate them to keep trying to adopt it.
The key claim here is that for AI to achieve human like intelligence — it will need to have theory of mind.
5. Language
The final breakthrough is the emergence of language. Language transformed the human brain from an organ for individual survival into a medium for accumulating and transmitting knowledge across generations.
“The real reason why humans are unique is that we accumulate our shared simulations across generations. We are the hive-brain apes.”
“The emergence of language was as monumental an event as the emergence of the first self-replicating DNA molecules. Language transformed the human brain from an ephemeral organ to an eternal medium of accumulating inventions.”
Language enabled humans to build complex societies, share ideas, and develop technologies far beyond what any other species could achieve. It allowed for the creation of a collective intelligence, where knowledge could be stored, refined, and passed down, driving human progress over millennia. “The incredible ascent of humankind during the past few thousand years had nothing to do with better genes and everything to do with the accumulation of better and more sophisticated ideas.”
Its noted that the evolution of language was unique compared to the other breakthroughs that were primarily evolutions that benefit the individual organism — whereas language is only valuable to a group — a much more nuanced evolutionary mechanism that depending on the emergency of complex social structure to enable its proliferation.
