> A team at the University of Tokyo created a special device that can deliver 10 diverse odors in a way that is accurate and timely
I've always wondered how far off we are from being able to "3D print" tastes and smells. I had a machine learning professor tell me that chemoreception is the only sense that can't be decomposed into "eigensmells" or "eigentastes" - not sure if that still holds true today.
They don't talk about the mechanics of this machine, but my guess is it just emits ten unique chemical compounds, rather than doing any kind of mixing.
> I've always wondered how far off we are from being able to "3D print" tastes and smells. I had a machine learning professor tell me that chemoreception is the only sense that can't be decomposed into "eigensmells" or "eigentastes" - not sure if that still holds true today.
This gets into an interesting and controversial (in that scientists will argue about it) domain.
Smell may be detecting the quantum vibrations of electrons.
There's some work on using ML methods to map from the structure of a molecule to an odor. Interesting to see where it goes, current results aren't that mind blowing:
1. B. Sanchez-Lengeling, J. N. Wei, B. K. Lee, R. C. Gerkin, A. Aspuru-Guzik, A. B. Wiltschko, Machine Learning for Scent: Learning Generalizable Perceptual Representations of Small Molecules. arXiv [stat.ML] (2019), (available at http://arxiv.org/abs/1910.10685).
2. J. Kowalewski, B. Huynh, A. Ray, A System-Wide Understanding of the Human Olfactory Percept Chemical Space. Chem. Senses. 46 (2021), doi:10.1093/chemse/bjab007.
3. L. Shang, C. Liu, F. Tang, B. Chen, L. Liu, K. Hayashi, Odorant molecular feature mining by diverse deep neural networks for prediction of odor perception categories. bioRxiv (2022), , doi:10.1101/2022.04.20.488977.
I suspect that might be harder than it sounds. See this (circa 2014) Aussie chemistry teacher's crowdsourced organic compound perceived-smell infographic -- it seems pretty random.
My favorite part of the vibration theory... from the Wikipedia article:
> Carvone presented a perplexing situation to vibration theory. Carvone has two isomers, which have identical vibrations, yet one smells like mint and the other like caraway (for which the compound is named).
> An experiment by Turin filmed by the 1995 BBC Horizon documentary "A Code in the Nose" consisted of mixing the mint isomer with butanone, on the theory that the shape of the G-protein-coupled receptor prevented the carbonyl group in the mint isomer from being detected by the "biological spectroscope". The experiment succeeded with the trained perfumers used as subjects, who perceived that a mixture of 60% butanone and 40% mint carvone smelled like caraway.
I read an article about making perfumes (it was mostly about industrial scents like laundry detergent) that profiled someone who does that, and said the same thing as you re scents can't be decomposed. I'll try and think of where I saw it.
Aside: what is the "speed of smell"? I assumed it would slowly diffuse across the room, but when one of my kids has a dirty nappy/diaper it is like a lightning bolt in my nostrils from the other side of the room.
It's governed by diffusion. There is an equation ( Einstein eq. I think) that says something about the average distance diffused being proportional to sqrt(time) and a diffusion coefficient which depends on what is diffusing and in what medium. the units of the diffusion coefficient are in m^2/sec because of the square root dependence but it's sort of an analog to the speed.
Delta x = Sqrt(6Dt) in 3d
It's based on brownian motion leading to a gaussian distribution of distances travelled
I've wondered something similar, in particular about what happens when I open the window of a closed room. How fast do the gas compositions reach equilibrium again?
Haven't found an answer yet even for very simple assumptions, e.g. 1m³ cube room, 0.01m² window.
This is a great question i'd love the answer to as well.
My intuition tells me the speed particles passively would move through the air would be limited roughly by the speed of sound? That feels like a hard upper bound, though?
I've always wondered how far off we are from being able to "3D print" tastes and smells. I had a machine learning professor tell me that chemoreception is the only sense that can't be decomposed into "eigensmells" or "eigentastes" - not sure if that still holds true today.
They don't talk about the mechanics of this machine, but my guess is it just emits ten unique chemical compounds, rather than doing any kind of mixing.