We use dogs to sniff for drugs even when the smell is masked or inside of things, they can identify and track a person by their unique smell, and dogs have also been able to smell all sorts of illnesses on people so it’s obviously something that would be incredibly useful but what’s the barrier stopping us from reliably making a versatile odor detector artificially?
In: 2
Smell is very different from other senses like vision or hearing.
Vision works by light-sensitive cells that send an electric signal when light hits them. You only need a few different types of cell that have different sensitivities to different wavelengths (“colors”) of light, to be able to discriminate a bunch of different colors. Then, you just stick a bunch of copies of those same few cell types onto a grid, put the grid behind a light-focusing system (cornea, pupil and lens), and you’ve got yourself an “image sensor” (the retina).
Hearing works by detecting air pressure waves (sound) using a thin membrane (the ear drum), which is connected mechanically to a kind of multi-tone “pitchfork” (the *cochlea)*. This pitchfork resonates with different frequencies along its length. High frequencies make the base of the pitchfork vibrate in resonance, while the tip of the pitchfork resonates with low frequencies. So, this mechanical device separates sound waves into different frequencies. Then, you simply stick vibration-sensor all along the length of the pitchfork, which send electrical signals when they detect vibration. Note that these sensory cells don’t need to know anything about frequency. They just sense vibration intensity at their location, which due to the pitchfork happens to correspond to a certain frequency. So a single type of sensory cell is sufficient for hearing different frequencies.
So we have vision using 4 sensory cell types, and hearing using just one. What about smell? Well, smell is simultaneously more complicated and more simple. It’s more simple because there is no intermediate mechanism between the sensory cells and the thing they sense. No frequency-separation mechanism and no light-focusing or spatial grid. So how is it more complicated? Well, smell is nothing more than detecting the presence of certain molecules in the air. And to do this, you need a different receptor for every molecule. A receptor is like a keyhole that only certain “keys” (molecules) fit into. When a molecule fits the receptor, the cell sends an electrical signal. So, if you want to sense 1000 different molecules, you need 1000 different types of receptor.
And therein lies the rub. To make an “artificial ear” (e.g. a microphone), you just need something that turns air vibrations into electrical signals. You can do the frequency separation afterwards using software, but you don’t even need to do that if all you want to be able to do is play back the sound (which just requires taking the recorded electrical signal and turning it back into air vibrations). To make an “artificial eye” (e.g. a camera), you just need a light-focusing mechanism and a grid of light-sensitive sensors, with a few different color sensitivities (e.g. red, green and blue). But to make an artificial nose, you need to design a molecular detector for every possible molecule that you want to detect.
On top of that, we don’t necessarily know which molecules we want to detect. And, quite often, it’s not about the presence of a single molecule, but rather a pattern of intensities across a number of molecules.
Dogs come “pre-equipped” with a nose that has evolved (over millions of years) to pick up thousands of different smells that occur in nature. We can probably make a more sensitive detector for any one given molecule that the dog can smell – that’s not really the issue. The issue is replicating every single detector that’s in the dog’s nose, when we don’t even know all the molecules that the dog is able to smell (or that we might want to detect).
Latest Answers