Elon Musk recently announced that the next project for Neuralink will be a “Blindsight” cortical implant aimed at restoring vision. Musk mentioned that the resolution of the vision might start off low, similar to early Nintendo graphics, but could eventually surpass normal human vision. However, this claim is based on the misconception that neurons in the brain function like pixels on a screen.
Engineers often assume that more pixels equate to better vision, similar to how monitors and phone screens operate. To investigate this further, researchers created a computational model of human vision to simulate the potential vision provided by an extremely high-resolution cortical implant. When comparing a movie with a resolution of 45,000 pixels to a simulation using 45,000 cortical electrodes, it was evident that the details of the scene were lost in the latter.
The reason for the blurriness in the simulation generated by electrodes is due to the fact that neurons in the human visual cortex do not represent individual pixels. Instead, each neuron has a specific receptive field that determines the pattern and location of visual stimuli that trigger the neuron to fire. Therefore, electrically stimulating a single neuron results in a blob determined by that neuron’s receptive field.
When looking at a single star in the night sky, numerous neurons with overlapping receptive fields are activated, creating a complex firing pattern across these neurons. To replicate the experience of seeing a star through cortical stimulation, it is essential to reproduce a neural response pattern similar to natural vision.
Some scientists propose that generating the correct stimulation pattern could restore naturalistic vision. However, achieving this requires thousands of electrodes and precise knowledge of each neuron’s receptive field. Without understanding the orientation and size of receptive fields, the resulting image would be unclear.
While the potential of restoring natural vision with cortical implants is promising, challenges remain in determining the receptive fields of individual neurons in blind patients. Without this crucial information, achieving clear vision with cortical implants will remain a complex task.
In our work as computational neuroscientists, we developed simulations to predict the perceptual experience of patients aiming to restore their sight. By creating a virtual patient to simulate what cortical implant patients may see, we were able to predict the quality of vision that future cortical implants could offer.
Models like ours provide insights into the potential performance of bionic eyes, offering a realistic perspective on the vision they may provide. It is crucial to approach sight restoration technology with caution to minimize harm to patients in case of device failure.
While achieving grainy and imperfect vision with cortical implants would be a significant advancement for individuals with incurable blindness, it is essential to maintain a sense of cautious optimism moving forward. The complexity of the human brain poses challenges that require careful consideration in the development of sight recovery technologies.