In a Leipzig classroom filled with MBA students last week, I posed a question that bridges biology and technology: Are there any similarities between the evolutionary milestones of human cognition and AIs? This exploration of intelligence’s origins will unfold across several posts, using the brain’s evolutionary journey as a lens to examine the mathematical and computational foundations of today’s AI and robotics—revealing both how far we’ve come and the vast territory still unmapped.
The nematode C. elegans, the most extensively studied worm on our planet, possesses a mere 302 neurons, compared to the human brain’s 86 billion. Despite this neural simplicity, the worm remains constantly in motion, navigating its world not through sight but through chemical detection. As it moves toward food sources, it follows increasing concentration gradients of specific molecules—essentially “smelling” its way forward with remarkable precision. Beyond olfaction, these tiny organisms employ specialised sensory cells that respond to light, temperature, and touch. They retreat from illuminated areas to evade predators and navigate away from temperature extremes that might compromise their survival.
Nematodes are a neuroscientist’s dream subject. Their evolution marked an enabling movement resulting from the “decisions” on what to approach and what to avoid. Scientists refer to this quality as “valence” – the inherent goodness or badness assigned to environmental stimuli. But this system created a new problem: what happens when conflicting signals arrive simultaneously? Imagine a nematode detecting both food and danger. Nature’s solution was elegant – a central processing hub where competing inputs could be weighed against each other to produce a single steering decision. This integration centre, this first primitive brain, became the command centre where life’s earliest either/or choices were made.
When a nematode detects something positive, specialised neurons fire and activate a “go forward” command. Conversely, when it senses something negative, a different neural pathway triggers a “change direction” response.
This valence system isn’t limited to just the outside environment. It adapts based on the worm’s internal conditions. A hungry nematode might brave dangerous terrain for food, while a well-fed one avoids the same risk.
The brain’s ability to reinterpret stimuli based on internal needs relies on sophisticated biochemistry. When well-nourished, cells secrete “satiety signals” in the form of insulin, indicating an abundance of energy. During energy deficits, they instead release “hunger signals,” effectively rewiring how the organism responds to the same environmental cues.
Steering isn’t just a feature of a worm. It is the very first evolutionary breakthrough in building more advanced intelligence.
Steering requires at least four things:
- a bilateral body plan for turning,
- valence neurons for detecting and categorising stimuli into good and bad,
- a brain for integrating input into a single steering decision, and
- the ability to modulate valence based on internal states.
The first artificial intelligences bear striking resemblance to our humble – but steering-capable – nematodes.
In the late 1940s, neurobiologist Grey Walter of the University of Bristol—already renowned for pioneering the EEG (electroencephalograph) machine—ventured beyond merely studying brains to actually attempting to create them.
Using salvaged alarm clock components and military surplus parts, he constructed two autonomous, battery-powered robots shaped like tortoises: Elmer and Elsie. Each featured a rotating light sensor atop its shell that detected and moved toward bright sources, plus touch sensors that triggered retreat when obstacles were encountered. Walter’s design connected just two sensory “neurons”—one for light detection, one for touch—to the robots’ primitive “brains.” Yet despite this simplicity, Elmer and Elsie exhibited surprisingly sophisticated behaviours, wandering their environments and even finding their way back to charging “hutches” when needed.
The parallels to nematodes are remarkable.
Just as the nematode’s sensory neurons detect environmental cues and adjust muscle contractions accordingly, Walter’s tortoises directly link sensory input to motor response. Both systems made steering “decisions” without complex central processing, achieving adaptive navigation with minimal neural architecture. The worm and the robots shared the same fundamental approach: accomplishing flexible, responsive behaviour with astonishingly few neurons.
While Walter’s tortoise robots demonstrated remarkable principles, they remained laboratory curiosities. The commercial breakthrough came decades later when MIT scientist Rodney Brooks cofounded iRobot in 1990. Brooks championed a revolutionary approach: instead of trying to replicate human-level cognition, he focused on building intelligence from simpler components. His philosophy materialised in 2002 with the Roomba vacuum robot, which quickly conquered living rooms across America.
The humble Roomba and the microscopic C. elegans share striking parallels in their architecture. Both navigate with remarkably limited sensory arrays—the Roomba detecting wall collisions and charging station proximity. Both operate with minimal processing power and bilateral symmetry, with the robot’s wheel configuration limiting it to forward and backwards movement with turns. Most significantly, both classify their worlds through a primitive valence system: in Roomba, walls register as negative stimuli to be avoided, while charging stations become positive targets when batteries run low. Upon detecting dirt, Roomba engages its Dirt Detect system and begins methodically circling the soiled area.
This capacity for directional movement represents evolution’s first major cognitive leap. Approximately 550 million years ago, life transformed from radially symmetric organisms like corals, which lack centralised brains, to bilaterally symmetric creatures resembling today’s nematodes. This revolutionary development—the ability to deliberately navigate toward resources and away from threats—became the cornerstone upon which nature would build increasingly complex neural systems.
Despite their vast differences, the humble nematode and household Roomba reveal a shared foundation of intelligence: both navigate their worlds through valence-based steering systems. Having established these parallels in rudimentary decision-making, our next exploration will examine how early brains developed the crucial ability to form associations and strengthen neural pathways through reinforcement—the evolutionary stepping stones toward more complex cognition.