Vision wasn’t made by any single creator. It evolved independently dozens of times across the animal kingdom over hundreds of millions of years, driven by the same basic molecular ingredients. Along the way, scientists uncovered how eyes develop, how the brain interprets what they capture, and eventually how to give machines a version of sight. Here’s the full story of who and what “made” vision.
Life’s First Light Detectors
The earliest forms of light sensitivity appeared billions of years before anything resembling an eye existed. Simple single-celled organisms like cyanobacteria, which have been around for roughly 3.7 billion years, could already respond to sunlight for energy. But responding to light isn’t the same as seeing. True vision required a specific molecular partnership: a protein called opsin bonded to a light-absorbing molecule derived from vitamin A (the same reason carrots are linked to eye health).
Interestingly, the very first opsin protein probably couldn’t detect light at all. Research published in the Proceedings of the National Academy of Sciences reconstructed the ancestor of all opsins and found it likely lacked the chemical hook needed to grab onto its light-sensitive partner. That ability evolved later, possibly in the common ancestor of all animals with true tissues, through a kind of molecular repurposing. A protein that originally had nothing to do with sight was gradually modified until it could capture photons and convert them into a cellular signal.
From Eyespots to Camera Eyes
Once that light-detecting molecule existed, evolution moved surprisingly fast. The progression went from a simple patch of light-sensitive cells (an eyespot) to a cupped shape that could sense the direction of light, to a fully enclosed chamber with a lens. Calculations suggest this entire sequence, from flat eyespot to camera-style eye, could happen in as few as 364,000 generations, potentially taking only half a million years.
Some of the earliest steps played out in tiny single-celled organisms. A group of protists called dinoflagellates includes species with a structure that looks eerily like a camera eye, complete with a lens-like component, packed into a body only 50 to 70 micrometers long. Jellyfish, which appeared around 635 to 543 million years ago, independently developed their own camera-style eyes. At least two species of box jellyfish have them, despite lacking anything we’d call a brain.
The oldest confirmed eye in the fossil record belonged to a trilobite called Olenellus fowleri, dating to the early or middle Cambrian period, roughly 540 million years ago. Trilobite eyes were unlike anything alive today. They used lenses made of crystalline calcite, a mineral, rather than the organic proteins that make up every other animal lens on Earth. Calcite has a higher refractive index than the chitin used by other arthropods, which gave trilobites a sharper focus underwater. Some species in the suborder Phacopina had lenses up to 2 millimeters across with an internal structure that corrected for optical distortion, essentially a built-in fix for blurry edges that human engineers solve with multi-element lens designs.
Calcite does produce double images because light splits into two rays as it passes through, but trilobites solved this too. Their lenses were oriented so the crystal axis ran parallel to the optical axis, minimizing the effect. And the tiny gap between the two images was smaller than the width of individual receptor cells, making the doubling invisible.
Eyes Evolved at Least 40 Times
One of the most remarkable facts about vision is that it wasn’t invented once. A landmark survey by von Salvini-Plawen and Mayr concluded that eyes evolved independently on at least 40, and possibly up to 65, separate occasions across the animal kingdom. Insects, vertebrates, mollusks, and jellyfish all arrived at their own solutions to the same problem.
Yet despite this diversity, nearly all seeing animals share a single master gene that kickstarts eye development. This gene, called Pax6, controls eye formation in organisms as different as fruit flies and humans. In vertebrates it directs lens and retina formation. In flies it does essentially the same thing for compound eyes. Experiments have shown that activating Pax6 in the wrong part of a fly’s body can trigger eye tissue to grow on a leg or wing. The gene is so deeply conserved that a mouse version of Pax6 can drive eye development in a fly. This shared genetic blueprint suggests that while different animal lineages built very different eyes, they all started from the same ancient toolkit.
How Light Becomes Sight
Building an eye is only half the story. Vision also requires converting light into electrical signals the brain can read. This conversion happens through a chain reaction inside photoreceptor cells in the retina. In darkness, a molecule called cyclic GMP keeps ion channels on the cell surface open, allowing a steady current to flow. When a photon hits the light-sensitive pigment rhodopsin, it triggers a shape change in the molecule’s vitamin A component, flipping it from one configuration to another. That shape change activates a signaling protein, which in turn activates an enzyme that breaks down cyclic GMP. As cyclic GMP levels drop, the channels close, the electrical current changes, and the cell sends a different signal to the brain. A single photon can trigger this entire cascade.
After the signal leaves the eye, it travels to the visual processing area at the back of the brain. From there, information splits into two parallel streams. One pathway runs upward toward the top of the brain and tracks where objects are and how they’re moving. The other runs downward toward the sides of the brain and handles what objects are, recognizing faces, reading words, identifying colors.
The Scientists Who Decoded Vision
Understanding how the brain actually processes visual information required decades of painstaking experiments. The biggest breakthrough came from David Hubel and Torsten Wiesel, two neuroscientists who began recording from individual brain cells in cats in 1959. Previous work had shown that cells in the retina respond to small spots of light. Hubel and Wiesel discovered that cells in the brain’s visual cortex respond to something completely different: oriented lines and edges.
A neuron that ignored a dot of light would fire vigorously when shown a bar tilted at a specific angle. They also found a hierarchy of processing. “Simple” cells responded to an oriented line in one precise location. “Complex” cells responded to the same orientation but tolerated the line being in slightly different positions. The properties of complex cells could be explained by combining input from multiple simple cells tuned to the same angle. This revealed a fundamental principle: the brain builds increasingly abstract representations of the visual world, step by step, transforming the pointillist dot pattern captured by the retina into the edges, shapes, and objects we consciously perceive. The discovery earned Hubel and Wiesel the Nobel Prize in 1981.
Teaching Machines to See
The idea of giving computers vision is almost as old as the field of artificial intelligence itself. In 1963, Larry Roberts at MIT introduced the concept of “block world,” a simplified visual environment made of basic 3D geometric shapes, and tried to get a computer to interpret their three-dimensional structure from flat images. Three years later, in 1966, Seymour Papert at MIT wrote a now-famous proposal to build a complete vision system as a summer project, with the goal of getting students to construct “a significant part of a visual system” in a few months.
That summer project was never finished. It turned out that vision, which feels effortless to any animal with eyes, is one of the hardest problems in computing. The challenge of recovering three-dimensional information from a two-dimensional image, recognizing objects, and interpreting scenes has driven decades of research and ultimately became one of the core applications of modern deep learning. Today’s computer vision systems owe their architecture, at least in part, to Hubel and Wiesel’s discovery that biological vision works through hierarchical layers of increasing abstraction, the same principle behind the layered structure of neural networks.