Reality is that which, when you stop believing in it, doesn’t go away. Philip K. Dick

There’s a persistent, bewildering notion in some circles in Mauritius that “AI can replace animal experiments right now.”

That claim strikes me with the same disbelief I’d feel if someone tried to sell me beachfront property at Trou aux Cerfs. It’s as disconnected from the reality I live in as a physician-scientist. Let me explain with a recent paper by Matsumoto et al. (Nature 629:91–99, 2025), which offers new insights into how the eye sees. But first, let’s talk about blindness.

Visual science is a frontier of modern medicine, from stem cell therapies to visual prosthetics. Nearly 40 million people worldwide live with blindness. A cure cannot afford to wait for AI to catch up. Solving this will take years of scientific rigor, including animal research. Let me explain why.

Peering into the eye of a vertebrate, an animal with a spine, like us, one can only admire what nature has crafted. This light-capturing wonder is a perfectly curved globe, layered with specialized tissues that work in harmony to capture, focus, and interpret light from the outside world. At the front, the cornea and lens bend light toward the retina, much like a camera lens focuses an image onto film.

The retina is a remarkably delicate sheet, lining the inside of the eye atop its tough outer shell, the sclera. It’s composed of ten precisely arranged layers of cells, all packed into a structure no thicker than a sheet of paper. When light reaches it, a cascade begins: photoreceptors detect the signal, and relay cells and fine-tuning circuits adjust for brightness, contrast, and motion, all before the message even leaves the eye. This entire ballet unfolds in milliseconds, effortlessly and without our awareness.

And perhaps most beautiful of all, at least to me, is this: the retina isn’t just part of the eye. It’s an extension of the brain, my area of expertise, a window not only into the world, but into the nervous system itself.

In 1888, the father of modern neuroscience, Santiago Ramón y Cajal, published his landmark study La rétine des vertébrés, becoming the first to systematically describe the vertebrate retina’s ten-layered architecture. Working with animals like rabbits, pigeons, frogs, and cats, he revealed not only how intricate the retina is, but how deeply conserved its structure remains across vertebrate species, a discovery that astonished the scientific world. Now, 137 years later, we’re still unlocking the retina’s secrets.

Among the retina’s most fascinating populations are the amacrine cells, interneurons that fine-tune visual signals before they exit the eye. As a Parkinson’s specialist, I have a particular interest in the starburst subtype, whose dysfunction is believed to contribute to visual hallucinations in PD.

That’s why I was thrilled to read the new study by Matsumoto and colleagues, a tour de force that likely took four years and crossed three countries: Japan, Denmark, and the United States. I estimate they used over 100 mice to complete the work. So, what did they accomplish that’s worth your attention?

They set out to systematically chart the staggering diversity of amacrine cells in the mouse retina. Using tools such as transcriptomics, functional imaging, viral tracing, and even serial electron microscopy, they built a comprehensive atlas of over 60 genetically distinct amacrine subtypes. Each was defined not just by gene expression, but by wiring patterns and functional behavior. In essence, they revealed that the retina isn’t just a light detector, it’s a miniature computer, parsing and shaping the visual world in real time before the brain perceives it.

So why do I say AI can’t replace animal experiments in medical research right now? Because Matsumoto and colleagues built on generations of scientists, and their work still needs to be vetted. We need to integrate findings across all retinal cell types, work that takes time, rigor, and live models. AI can’t generate these data. What it can do is help us analyze them and suggest new hypotheses for scientists to test. That’s how we’ll get to a cure for blindness, one step at a time.

For the final movement, what does all of this have to do with terminal provisions in the 2017 Mauritius Animal Welfare Act Amendments (MAWAA), the proposal for an Independent Life Sciences Authority (ILSA) for enlightened animal experiment regulations and monitoring, and the vision of Mauritius as the Switzerland of life sciences in Africa? Everything.

Let us begin by comparing the mouse, non-human primate (NHP), and human retina. While all three share the same ten-layer architecture, there is a clear progression in complexity and specialization. The eye volumes of the mouse, NHP, and human are approximately 20, 4,000, and 7,000 cubic millimeters, respectively. Their retinal surface areas follow a similar pattern: about 20 mm² in the mouse, 600 mm² in the NHP, and 1,200 mm² in humans. In both structure and function, NHP and human eyes are far more alike than either is to the mouse.

This has important implications for visual neuroscience and therapeutics. Currently, Mauritius is the world’s best source for lab-grade NHPs. So why shouldn’t we become the place where advanced vision research is done? Take, for example, stem cell therapies for restoring starburst amacrine cells affected in Parkinson’s disease. Before any such treatment can be tested in humans, it must first be validated in NHPs. With the right ethical and legal frameworks, there is no reason those crucial experiments cannot happen here. I believe this possibility is both realistic and within reach.

My vision is for Mauritius to become the Switzerland of life sciences in Africa, a contributor to global medical progress in curing blindness and a host of other diseases that touch our families and communities. But to make that vision real, we must pass MAWAA and establish ILSA without delay. If we hesitate, the window will close, and more daring, prepared countries will seize the opportunity we let slip.

AI will do many great things, but it does not stand for All Inclusive. And if anyone still believes that AI can replace ethically conducted animal research in critical fields like vision science immediately, I have beachfront property to sell you at Trou aux Cerfs.

Dr. Shivraj Sohur, originally from Espérance Trébuchet, is a neurologist at Mass General Brigham, part-time faculty at Harvard Medical School, and a leader in the global life sciences industry. The views expressed are his own.