Transferring the Olfactory System to Silicon

Reading Time: 4 minutes

Issue 15, Volume 110

By Daisy Lin, Shriya Anand 

Most of us have heard of Artificial Intelligence (A.I.) at least once in our lives. A.I. refers to the usage of software to create a system able to perform tasks such as speech recognition, visual and odor perception, and decision-making. Because of its rising prevalence over the past few years, professionals in the field are developing algorithms to improve technology's ability to emulate biological processes. In fact, we see the effects of these algorithms in our daily lives, from Siri to Touch ID to facial recognition on our phones. Have you ever heard of odor detection, though? Another breakthrough in science—neuromorphic chips (computer chips that mimic the human brain)—is responsible for such a feat. Only three years ago, Intel, a company well known for manufacturing computer hardware like traditional motherboard chipsets, integrated circuits, and graphics chips, released a neuromorphic chip capable of perceiving different smells. Its invention is known as the Loihi chip and is recognized for its ability to distinguish between varying odors much like the olfactory system in humans.

Olfaction is a form of chemoreception by which organisms respond to chemical stimuli gathered from their environment. It refers to the perception of smell, beginning with the olfactory epithelium in the nasal cavity of humans. The olfactory epithelium contains millions of olfactory receptor cells, each linked to a single short branched extension of a nerve cell known as a dendrite with cilia, a microscopic hairlike structure on the surface of certain cells, that spread across the olfactory epithelium. When odors enter the nasal cavity either through inhalation or the rising of particles from the mouth, the cilia are stimulated, triggering the depolarization of olfactory cells and the sending of action potentials to the olfactory bulb through dendrites. These dendrites collect in small clusters known as glomeruli. Here, synaptic connections are formed with different types of olfactory bulb neurons, including mitral cells and tufted relay neurons, both of which are connected to the olfactory tract, a bundle of fibers that carry olfactory information to the olfactory cortex. The olfactory cortex is found in the temporal lobe of the brain involved with sensory input. After receiving the information, the olfactory cortex processes it, allowing humans to recognize different scents.

So, how exactly can the olfactory system be transferred to the silicon chip? Intel says that its chip’s arrangement is "derived from the architecture and dynamics of the brain's olfactory circuits.” The 128 neuromorphic cores, or processing units, work asynchronously using a spiking neural network (SNN), a method almost identical to the biological processes that transmit signals through the nervous system. SNN allows for any number of neurons to send out impulses to neighboring neurons through the synapses. This type of communication is enabled by the packaging of messages that are sent over impulses. Once enough spikes caused by these impulses accumulate at a certain position, the core sends the message out to other cores. Each core also consists of a “learning engine” that adapts according to the spike timings it receives, allowing for the Loihi chip to be more flexible in its environment.

In order to collect data for this chip, Intel’s neuromorphic computing team used a wind tunnel with 72 chemical sensors responsible for detecting and differentiating ten different scents as they circulated. The data received from these sensors was then transferred to the Loihi chip, which created neural representations of each odor via machine learning. This is analogous to the human brain, which assigns a certain pattern of electrical signals to different scents. The Loihi chip may even be considered more efficient than the human nose, despite being familiar with only 10 smells. It is capable of identifying the odors even with strong scent interferences.

Because of the Loihi chips’ unique ability to recognize scents, scientists have begun to utilize them in the physical world. After thorough testing, it was found that the chip could successfully detect specific odors even in the presence of other smells. As a result, the technology can be used to identify pheromones released when a person is carrying a certain illness or other smells connected to drugs, explosives, and environmental gases while blocking interfering stimuli. This would be hugely beneficial: sniffer dogs would be put out of harm's way and humans could still benefit from a heightened sense of smell provided by the Loihi chip. Nabil Imam, a senior research scientist working at Intel, also commented that the chip’s ability to mimic other neurological processes “demonstrates Loihi’s potential to provide important sensing capabilities that could benefit various industries.” For example, it could be implemented into different household appliances such as fire alarms and ovens. Currently, fire alarms work using either photoelectric sensors or ionization sensors that detect smoke. However, they are often inefficient and are triggered at unnecessary times. Therefore, it would be beneficial to use Loihi chip sensors that can detect the smell of fumes rather than the presence of smoke. Additionally, Loihi chips may be used in ovens because the chip can be programmed to identify the smell of charring food, as there is currently no way to combat the inconvenience. Considering the many uses of the Loihi chip, it is likely that it will soon be commercially manufactured and sold to perform difficult tasks with greater ease.

Through the development and possible uses for the brain-like neuromorphic Loihi chip, many new capabilities for A.I. have been introduced. Pieces of technology now possess the ability to interpret and respond to stimuli in a way never thought possible. Specifically, the Loihi chip features the ability to sense smell. Thanks to this breakthrough, complicated jobs can be made safer and simpler with these humanesque machines. With each passing day, there is continued development of new technologies, applications, and A.I. Though scientists have yet to create devices able to mimic all human processes to a tee, the dream may not be far from reality. For example, Intel is currently working on other programs capable of detecting touch and movement. Soon, advanced A.I. will no longer seem like a foreign concept—but rather systems incorporated into people’s daily lives.