The boundaries between human biology and intelligent technology are beginning to blur. Nowhere is this convergence more evident than in the emerging field of sensory augmentation, where artificial intelligence (AI), neuroscience, and wearable computing are coming together to redefine our fundamental senses.

For decades, the primary goal of this field was restoration, or developing technologies to help people regain lost sight or hearing. However, experts say that balance is now shifting. The field is moving beyond simple repair and increasingly is focused on augmentation: not just restoring, but actively enhancing and extending human capabilities.

This shift from restoration to enhancement recently was highlighted in a remarkable study. Researchers at Fudan University in Shanghai created an artificial retina that not only restored sight to blind animals but also gave them “super vision,” or the extraordinary ability to see infrared light, a spectrum invisible to them naturally.

This advance builds on decades of foundational work in sensory restoration. The most commercially successful and widespread example, the cochlear implant, already has restored functional hearing to hundreds of thousands of individuals by translating sound into electrical signals that are sent directly to the auditory nerve.

In the visual domain, devices like the pioneering Argus II retinal prosthesis (or “bionic eye”) proved that surgically implanted electrodes could restore a basic, low-resolution sense of sight to people with specific forms of blindness, establishing the viability of the “invasive” path of sensory augmentation that researchers are still pursuing. (The Argus II was discontinued after it failed to generate enough revenue and the company behind it, Second Sight, ran into financial issues.)

This acceleration is being driven by more than just implant technology. The true catalyst for the augmentation era is the processing power of artificial intelligence (AI).

AI is becoming essential for interpreting complex, real-world data before it is delivered to a human user. This is already common in advanced hearing aids, which use deep neural networks to isolate a single voice in a crowded room. This same computational power also is making non-invasive “sensory substitution” devices, which might translate a visual scene into complex soundscapes or tactile patterns, more sophisticated and learnable than ever before.

“While the motivation for restorative technologies remains high, we are entering an era increasingly focused on the augmentation (such as enhancement and extension) of human capabilities,” said Amber Maimon, a researcher specializing in sensory augmentation at Haifa University in Israel.

The Dual Frontiers of Sensory Augmentation

Sensory augmentation work is advancing along two major, parallel paths, according to Maimon.

The first is the invasive approach, which involves neural prosthetics and other interventions that directly interface with or modify sensory pathways. This is the category that includes the bionic eye, which aims to restore vision by electrically stimulating the remaining neural pathways.

The second is the non-invasive approach, such as sensory substitution systems that translate information from one sense to another. A key example is the EyeCane, a device that translates distance information into sound or vibration for the user.

“Both paths have been explored for decades, yet recent progress in neuroscience, materials, and computation is transforming what is possible,” said Maimon. Invasive neurotechnologies, for instance, are becoming increasingly sophisticated.

“Early visual prostheses relied on direct stimulation of retinal or cortical regions, producing only basic light or shape percepts,” she said. “Current approaches move toward higher-resolution, wireless, and biocompatible implants that combine advanced image processing with materials designed to adapt to neural tissue over time.”

Maimon said dense, flexible electrode arrays now enable more stable and long-term integration with the brain, as well as “improvements in wireless data and power transmission” that enhance both safety and reliability.

“At the same time,” she added, “emerging bidirectional brain-computer interfaces are forming closed-loop systems in which artificial sensors and neural circuits continuously interact,” bringing prosthetic vision closer to natural perception, or even extending its capabilities.

Non-invasive systems are evolving in parallel. Instead of surgically interfacing with the nervous system, these methods rely on training and active exploration to enable users to build new perceptual mappings. Research has shown, Maimon said, that the brain can reorganize itself to interpret these inputs as perceptual rather than symbolic. As a result, this path can yield impressive results while avoiding surgical risks and allowing rapid iteration, all in a way that can be tailored to individual users.

In fact, while high-tech implants capture the imagination, James Negen, a researcher at Liverpool John Moores University in the U.K., said the most exciting and effective technology in this space requires no electronics at all.

“Human echolocation is the objectively correct answer,” Negen said. “It is the only sensory augmentation approach to have a sustained community of real-world everyday users.” Negen pointed out that echolocation, a form of sensory substitution where a person uses clicks and their echoes to perceive their environment, has been “refined for decades” by scientists like Lore Thaler and practitioners like Daniel Kish.

Yet the most promising direction moving forward, said Maimon, is not pursuing one path over the other, but in developing hybrid systems that combine invasive and non-invasive approaches. By bringing together the precision and fidelity of invasive interfaces with the adaptability and accessibility of non-invasive ones, she said, new systems could “enable richer, more natural, and more adaptive forms of perception.”

“For example, a visual implant could convey the overall spatial structure of a scene directly to the visual cortex, while a non-invasive sensory substitution device provides complementary information about texture, contrast, or fine details,” she explained.

The Perception Problem

Despite the rapid progress on both frontiers, the dream of a truly functional artificial eye remains distant. According to Negen, when it comes to that specific goal, “We are so far away that we don’t yet know what we don’t know.”

“It is like asking Aristotle to lay out the path to the atom bomb,” he said. “We have only the most shallow understanding of basic questions like what physical properties give rise to conscious experience, what makes an experience perceptual, and how we come to feel ownership over a sensory stream.”

This philosophical hurdle points to a core technical issue.

“The greatest challenge is that perception is not simply electrical,” said Maimon. “It’s not enough to activate the right neurons or reproduce the correct signal patterns. The information must be organized in a way the brain can interpret and integrate, matching the timing, dynamics, and structure of natural sensory input.”

In other words, stimulation alone does not create perception; the brain has to make sense of what it receives.

Compounding this problem is the brain’s constant plasticity. Neural pathways reorganize as we learn and adapt, meaning an artificial implant “must be dynamic rather than static,” Maimon said. It must be “capable of adjusting not only to long-term neural reorganization, but also to the brain’s moment-to-moment variability across states of attention, context, and experience.”

Achin Bhowmik, CTO at Eden Prairie, MN-based Starkey Hearing, agreed, noting the technical challenges remain “formidable.” He cited key hurdles such as developing energy-efficient, real-time AI algorithms that can operate at the edge, creating biocompatible interfaces that communicate seamlessly with the nervous system and, perhaps most importantly, understanding how to harmonize artificial inputs with the brain’s perceptual and cognitive processes.

This challenge of creating a feeling of perception is also a challenge for non-invasive systems.

“It is not enough for users to decode patterns. The feedback must evoke a sensory quality that feels immediate and grounded in perception,” Maimon said, adding that it must do so without requiring extensive training or cognitive effort.

From Restoration to Enhancement

For much of its history, sensory augmentation was driven by a restorative goal. But as the technology matures, that distinction is fading.

“Looking ahead, I believe the distinction between restoration and enhancement will fade,” said Bhowmik. “Once we learn to emulate and extend our natural senses, the same technologies that restore hearing or vision will also enable people to perceive the world in entirely new ways.”

This new focus on augmentation generally falls into two categories. The first is enhancement, which aims to refine existing senses, such as sharpening vision, or improving touch or hearing. The second is extension, which moves further by granting access to sensory information humans do not naturally perceive, such as thermal, ultrasonic, or magnetic information.

This shift opens the door to a much larger potential user base.

“We have become much better at preventing sensory loss over the last century,” Negen said, explaining that “the potential user base for sensory substitution is much smaller than the potential user base for sensory augmentation.”

As new sensory skills combine with existing perception, they can enhance precision, making them a “true augmentation, not a fallback, even for someone without any sensory loss,” according to Negen. This creates an opportunity for niche but useful applications in areas like aging, working in hazardous environments, and perhaps even sports.

The final frontier, however, is the creation of truly novel senses that translate symbolic or informational content into sensory form.

“Imagine feeling air quality, perceiving electromagnetic fields, or intuitively sensing data streams,” said Maimon. “Such possibilities challenge our assumptions about perception, cognition, and the boundaries of human experience. They suggest that sensing is not merely biological, but something that can be designed, extended, and reimagined.”

This future is arriving in subtle ways. Bhowmik pointed to modern hearing aids, which have evolved far beyond simple sound amplification. “Modern hearing aids, for instance, use deep neural networks to separate speech from background noise, monitor physiological signals, and even detect falls,” he said. “They effectively function as miniature AI computers for the ear.”