Thursday, March 4, 2021

Artificial Pheromones and Biomimetic Mind Control


Breathing may change your mind about free will
Feb 2020, phy.org

This reminds us how interconnected the olfactory system is with our body. "Osmetic Ontogenesis" by Hosek and Freeman in 2001 is a great dive into this subject, with a title that's hard to forget.
Scientists at EPFL in Switzerland have shown that you are more likely to initiate a voluntary decision as you exhale. 
via by Ecole Polytechnique Federale de Lausanne: Hyeong-Dong Park et al. Breathing is coupled with voluntary action and the cortical readiness potential, Nature Communications (2020). DOI: 10.1038/s41467-019-13967-9



Gut bacteria may modify behavior in worms, influencing eating habits
Jun 2020, phys.org
Specific gut bacteria in the worm may modify the animal's behavior, directing its dining decisions.

"In this way, the bacteria can take control over the host animal's sensory decision-making process, which affects their responses to odors and may influence food choices" said Dr. Sengupta.
via the National Institutes of Health: MP O'Donnell et al. Modulation of olfactory behavior by a gut bacteria-produced neurotransmitter. Nature, 2020. DOI: 10.1038/s41586-020-2395-5


Non-invasive nerve stimulation boosts learning of foreign language sounds
Aug 2020, phys.org

Can this be used for olfactory nerves as well?
Researchers significantly improved the ability of native English speakers to distinguish between Mandarin tones by using precisely timed, non-invasive stimulation of the vagus nerve.
via University of Pittsburg: Non-invasive peripheral nerve stimulation selectively enhances speech category learning in adults. Fernando Llanos et al. npj [Nature Partner Journal] Science of Learning volume 5, Article number: 12 (2020). https://www.nature.com/articles/s41539-020-0070-0


A lab that reads—and writes—our dreams
Apr 2020, phys.org

Well-known olfactory researcher Judith Amoore dispenses odors as subjects slip into sleep. In follow-up interviews, subjects report memories associated with the smells.

The device is supposed to trigger scents with positive associations during nightmares to help trauma and PTSD sufferers without them even being awake.

via MIT and the Dormio device

Augmented Reality, EPFL École polytechnique fédérale de Lausanne

Pregnancy test for water' delivers fast, easy results on water quality
Jul 2020, phys.org

Frankenstein as f***
In cell-free synthetic biology, researchers take the molecular machinery—including DNA, RNA and proteins—out of cells, and then reprogram that machinery to perform new tasks. The idea is akin to opening the hood of the car and removing the engine, which allows researchers to use the engine for different purposes, free from the constraints of the car. In this case, Lucks' team used molecular machinery from bacterial cells.

"We found out how bacteria naturally taste things in their water," Lucks added. "They do so with little molecular-level 'taste buds'. Cell-free synthetic biology allows us to take those little molecular taste buds out and put them into a test tube. We can then 're-wire' them up to produce a visual signal. It glows to let the user quickly and easily see if there's a contaminant in their water."

These reprogramed "taste buds" are freeze-dried to become shelf-stable and put into test tubes. Adding a drop of water to the tube—and then flicking it—sets off a chemical reaction that causes the freeze-dried pellet to glow in the presence of a contaminant.

In Paradise, California, after the recent wildfire disaster there, ... their teams tested ROSALIND alongside gold-standard water tests and discovered that ROSALIND was able to identify the presence of elevated toxic metals in the water supply. It also provided much faster and less expensive results.
via Northwestern University: Cell-free biosensors for rapid detection of water contaminants, Nature Biotechnology (2020). DOI: 10.1038/s41587-020-0571-7


Exhaled biomarkers can reveal lung disease
Jul 2020, phys.org

Inhalable nanosensors:
"We envision that this technology would allow you to inhale a sensor and then breathe out a volatile gas in about 10 minutes that reports on the status of your lungs and whether the medicines you are taking are working," says Sangeeta Bhatia, the John and Dorothy Wilson Professor of Health Sciences and Technology and Electrical Engineering and Computer Science at MIT.
via Massachusetts Institute of Technology: Chan, L.W., Anahtar, M.N., Ong, T. et al. Engineering synthetic breath biomarkers for respiratory disease, Nature Nanotechnology (2020). DOI: 10.1038/s41565-020-0723-4


Researchers develop sensors that detect human biomarkers and toxic gas
Nov 2020, phys.org

VOCs are going to be the new big data treasure chest.

via Penn State: Ning Yi et al. Stretchable gas sensors for detecting biomarkers from humans and exposed environments, TrAC Trends in Analytical Chemistry (2020). DOI: 10.1016/j.trac.2020.116085


AI-powered 'electronic nose' to sniff out meat freshness
Nov 2020, phys.org

Go figure it has to see it to smell it...
The e-nose developed by NTU scientists and their collaborators comprises two elements: a colored barcode that reacts with gasses produced by decaying meat; and a barcode reader that uses AI to interpret the combination of colors on the barcode. To make the e-nose portable, the scientists integrated it into a smartphone app that can yield results in 30 seconds.
via Nanyang Technological University: Lingling Guo et al. Portable Food‐Freshness Prediction Platform Based on Colorimetric Barcode Combinatorics and Deep Convolutional Neural Networks, Advanced Materials (2020). DOI: 10.1002/adma.202004805

The Cat Copter, for real.

A system for swarm robotics applications inspired by pheromone communication in insects
Jul 2020, phys.org

Biomimetic artificial pheromone signaling in robotic swarms:
One of the most promising systems developed so far is COSΦ, a system that uses light to emulate pheromone release in humans and animals.

So far, the researchers evaluated their artificial pheromone system in a series of experiments in which a swarm of small mobile robots moved around and adapted to different environmental factors. Their results were highly promising, as their system enabled effective communication and prompted the desired group behaviors among members of the swarm.
via the University of Manchester: Seongin Na et al. Bio-inspired artificial pheromone system for swarm robotics applications, Adaptive Behavior (2020). DOI: 10.1177/1059712320918936


Researchers find a chemical that makes locusts swarm
Aug 2020, Ars Technica

Distant future, we will also have pheromone programming or engineering, so that we don't "swarm" and so we can be more easily controlled

via the Chinese Academy of Sciences: 4-Vinylanisole is an aggregation pheromone in locusts
Xiaojiao Guo. Nature volume 584, pages584–588(2020). 12 Aug 2020. DOI: 10.1038/s41586-020-2610-4
https://www.nature.com/articles/s41586-020-2610-4


Researchers one step closer to bomb-sniffing cyborg locusts
Aug 2020, phys.org

An irresistible scent makes locusts swarm, study finds
Aug 2020, phys.org

'The Smellicopter,' an obstacle-avoiding drone that uses a live moth antenna to seek out smells
Dec 2020, phys.org

Awesome. Made me think of the catcopter though. Ah, old times.
"Nature really blows our human-made odor sensors out of the water," said lead author Melanie Anderson, a UW doctoral student in mechanical engineering. "By using an actual moth antenna ...
via the University of Washington: Melanie Joyce Anderson et al. A bio-hybrid odor-guided autonomous palm-sized air vehicle, Bioinspiration & Biomimetics (2020). DOI: 10.1088/1748-3190/abbd81
 
Posted by at No comments:
Labels: , , , , , , , , , , ,

Thursday, February 4, 2021

Smells Like Covid


It took an earthshattering global pandemic, but we're all of the sudden realizing that we, as humans, have a sense of smell, and that it's kind of important to us. Go figure.

But, this whole time, it never occurred to me that something would actually "smell like Covid," but alas: 
"Eggs physically repulse me and I'm unable to enjoy beer or wine as they have a flavour I simply call Covid."

-Parosmia: 'Since I had Covid, food makes me want to vomit'
BBC News, Feb 2021
Posted by at No comments:
Labels: , , , , ,

Saturday, January 9, 2021

Phantosmia

AKA Regeneration of Olfactory Neurons and Combinatorial Perception in People Recovering from the Novel Coronavirus of 2019

This comes from a post-scripted entry to another post about the Tree of Heaven, to which I am pseudo-anosmic.

After some traumatic disturbance to your olfactory neurons, like from being attacked by a virus, you may experience changes in smell or taste. This is also called anosmia, partial anosmia, paraosmia, or phantosmia, the last referring to not a loss of smell but a change in the way things smell.

Phantosmia, like all phenomena in olfactory science, is not understood enough to say much from an evidenced-based point of view. I'm making a broad speculation here, not to explain, but to make someone interested enough that they will investigate further for themselves, and maybe even initiate more research into the topic.

The change in smell that comes from phantosmia is common, but its origins are often overlooked. It likely signals a change in the structure of neurons used to smell. These are the only part of your brain that pass the blood-brain barrier and rest outside your skull, in the mucous atop the epithelium skin way up in the top of your nostril canal (right where they swab that sample for your PCR test by the way, and not a coincidence). These neurons are thus both very vulnerable to damage, and able to regenerate indefinitely.

Combine this with another fact about olfaction -- it is combinatorial. That means when you smell "apple," there are a bunch of different receptors all lighting up in a pattern that means "apple." There is no Apple receptor, and no gene that codes our nose for Apple. For some there are, but for the most part, no. Olfaction is all gestalt. Take one piece out, and the entire picture gets weird as hell. Something's wrong but you can't tell what. So your brain misfires, it says "cigarette smoke" when it's really something else entirely. 

But after damage to your system, it is re-learning how to smell. Your nose-brain is a deep learning neural network that requires countless iterations to "learn" what a smell is. And while it's relearning after an infection, it gets confused.

In a very mild manner, and for reasons I will attribute to having been infected myself, my Tom Ford Italian Cypress lost its depth and presented as cinnamon and bubblegum, for about three days. If you've ever smelled Italian Cypress (and if not good luck it's discontinued since 2014), you would know that it does NOT smell like cinnamon and bubble gum. That's happened to me once before, and I will now assume it was because I was then also mildly infected with a virus. (Twice before actually, because the first time I ever smelled it, this is what I got, and why I didn't buy it; after almost a week the blotter revealed its powerful basenote, so I eventually gave it another try, and eventually I could smell the "whole thing," and then went and bought a bottle). 

And another thing -- when faced with new smells, we tend to call them bad. After repeated exposure, we can start to see them as good, but it's more likely we call new smells bad at first. Or at least "weird." So if your system is relearning how to smell, then lots of typical odor exposures will present as "bad" to you, simply by their being new, that is, new to your newly developing system. And all of the sudden, anything that doesn't compute properly on your new system could become "cigarette smoke," for example. Rotten meat is another good one for this, but it could be anything really (and it could also be really debilitating, just imagine.)

We also really suck at naming and describing bad smells. What does the Tree of Heaven smell like? How about a stink bug? Or go ask the folks at the South Coast Air Quality Management District about this. There's a direct correlation then, about the mental propensity for linguistically encoding sensory data, and combinatorially processing that raw data in the first place. Not only is it hard to "find the name" of a bad smell, it's actually hard to even "smell" it in the first place. 

Eventually, the system will recalibrate and relearn how to smell, and you'll be back to normal. But that doesn't always happen. Blunt trauma can kill those neurons forever. Really bad infections too. Sometimes it doesn't come back because you're not using it; like therapy after a stroke, it only comes back if you try really hard to use it.

And some of us, well, we're just getting older. Things don't work like they used to. And not only that, changes in smell can predict all kinds of neurological diseases decades in advance (Parkinson's, Alzheimer's, etc.).

But don't lose hope, I have apparently learned how to smell something new at 40 years old -- something I never had smelled before -- the Tree of Heaven. This may sound outrageous (it still does to me), but this massive revelation in olfactory research puts a lot of the details into that story. It comes from two different studies, one from Hebrew University and one from the Harvard Medical School, both of which are saying that there's two layers to olfactory perception, with the first layer of coding based on physicochemical properties of the odorant, and a second layer coded by your previous interactions with it. 

I'll repaste their description:
The general profile of excitatory vs. inhibitory responses by mitral cells changed with learning and task demands. In naive animals, most responsive cells (71%) responded by excitation to the odours. Following the learning of the 5-decision boundary task, the ratio of excitatory/inhibitory responses reversed. After learning, the majority (71.4%) of neurons now responded by inhibitory calcium transients to the odours (Fig. 6C,E). The ratio of inhibitory vs. excitatory responses reverted back to normal after retraining the mice on the 1-decision boundary task. Specifically, 73.3% of responsive neurons were again excitatory on day 18.

Hebrew University - Flexible Representations of Odour Categories in the Mouse Olfactory Bulb. Elena Kudryavitskaya, Eran Marom, David Pash, Adi Mizrahi. Hebrew University of Jerusalem. Mar 24 2020. BioRxiv. doi: https://doi.org/10.1101/2020.03.21.002006

Harvard Medical School - Stan L. Pashkovski et al, Structure and flexibility in cortical representations of odor space, Nature (2020). DOI: 10.1038/s41586-020-2451-1

If you lost your sense of smell, or can't get it back, check out some of these links to get more info about it and what you can do:

How Covid-19 can damage the brain
June 2020, BBC Future

How COVID-19 causes smell loss
July 25, phys.org

Suggested mechanism for COVID-induced smell loss - David H. Brann et al. Non-neuronal expression of SARS-CoV-2 entry genes in the olfactory system suggests mechanisms underlying COVID-19-associated anosmia, Science Advances (2020). DOI: 10.1126/sciadv.abc5801

AbScent - UK Organization raising public awareness of smell loss

National Institute on Deafness and Other Communicable Disorders (NIDC) - Smell Disorders

ENT UK - Loss of Smell as Marker of Covid-19 Infection

What Makes a Better Smeller?
Asifa Majid, Laura Speed, Ilja Croijmans, et al. Sage Pub., Jan 2017, volume 46 issue pp 406-430

This paper covers olfactory deficiencies and factors that make someone a better smeller, is based on neurodiversity and odor environment, and reviews how ambient odor or culinary traditions can influence odor perception.

There are at least three factors to consider as foundations of variation: our biological infrastructure, the experiences we navigate during our lifetime, and our physical and social environment (biology, experience, environment). Note the Japanese masters of koh-doóan, the ancient Japanese tradition of incense appreciation.

A few more recent sources of info on this topic:
Eric Song et al. Neuroinvasion of SARS-CoV-2 in human and mouse brain, Journal of Experimental Medicine (2021). DOI: 10.1084/jem.20202135

David H. Brann et al. Non-neuronal expression of SARS-CoV-2 entry genes in the olfactory system suggests mechanisms underlying COVID-19-associated anosmia, Science Advances (2020). DOI: 10.1126/sciadv.abc5801

Trying to Make Sense of Long COVID Syndrome, Dr. Francis Collins. NIH Director's Blog, January 19th, 2021. https://directorsblog.nih.gov/
***
image source: Neural cells in a live mouse - Richard Roth and Richard Huganir

Posted by at No comments:
Labels: , , ,

Tuesday, October 20, 2020

Mummy Meat


People used to eat mummies.

Last week for the first time, I saw someone else writing about this, but in the spirit of the season, which is Halloween here in the US. Good time for a perennial favorite here at Limbic Signal:

A while back I discovered that people with lots of money and imagination were eating preserved human bodies to get high, about 200 years ago.

I thank Annick Le Guérer for this tidbit, she wrote about it in her book Scent, the Mysterious and Essential Powers of Smell, written in 1988, and translated from French in 1994.

There was a time, we must remember, when mummies were a new thing, never before imagined by the Westerners excavating these immortalized bodies. It's hard to conjure the pretense of shock at something that has been around since long before you were born.

For a moment if you will, try to imagine what it would have been like to learn that deep within the awe-striking pyramidal limestone masses were 3,000-year old physically intact human bodies. This at a time before we had refrigerators! We couldn't even keep a bowl of potato salad from going bad in a couple days, and here's an entire human body with its skin still intact, and older than the entire city in which you live.*

That's magic to a person of the 19th century. Today, our tupperware will probably last longer than our species itself, nevermind the bodies we leave behind. We have plastic flowers for goodness sake. But if you can transport yourself back to a time where everything was ephemeral, you can begin to understand the fascination.

And the exoticism. The preserving substances used on mummies were much less known to Europeans hundreds of years ago. Today we can fly from London to Cairo in four hours. Then, it could have taken up to a month. Today we can have in our pantry any spice produced in any place in the world, within a few days. Things were different then. Egypt in itself was pretty exotic, and mummies, forgetaboutit.

So if you can now picture yourself at an all-nighter in a regal estate, well after midnight, deep into the spirits, when your host spreads on the table these tiny morsels of dry-aged royalty from another era, and who might as well be from another planet, and tells you to dig in – you will be intoxicated. The meat doesn’t make you intoxicated, of course; the idea is enough to placebo the heck out of your dopamine receptors.

I get into the details of how smells are so good at tricking us in this older post. But if you're interested in throwing your own mummy-party, these folks from the University of York have decoded the ancient recipe:

Mummy
-a plant oil –  possibly sesame oil;
-a "balsam-type" plant or root extract that may have come from bullrushes;
-a plant-based gum - a natural sugar that may have been extracted from acacia;
-crucially, a conifer tree resin, which was probably pine resin
-source

*Note that this isn't entirely true, for we have known for a long time about preserving things. Many of the same substances used to preserve mummies also preserve our food. Also note, however, that roughly speaking the practice of using spices to preserve food decreases as you move from the equator, with those places tending to use fermentation as a means of preservation instead, which is the opposite of using spices – one keeps microbial activity at bay, and the other uses it on purpose to regulate the rate of decay. Fermenting mummies would not have worked as well. But that’s pretty tangential, and a transparent excuse to say fermented mummies.

** Know that Europeans are not entirely unfamiliar with mummies; they’re called relics, and they’re not nearly as old.

***Finally, preserving the dead is not the most uncommon thing ever; Japan has a long history of it.


Post Script:
Embalming was just one aspect of preservation. Other steps included:
-Removal of the brain - possibly using a "whisking" process to cause the brain to liquefy
-Removal of the internal organs
-Putting the body into a natural salt to dry it out
-Coating the body in the embalming recipe , to kill bacteria and to seal it
-Wrapping the body in linen


Notes:
Ancient Egyptian mummification 'recipe' revealed
Aug 2018, BBC

Mummy-Eaters
Nov 2016, Limbic Signal

Like the Stanley What
Feb 2016, Network Address

I Got Your Back
Apr 2017, Network Address

Sokushinbutsu
Japanese Mummies
Posted by at No comments:
Labels: , , , , ,

Thursday, October 15, 2020

Exobiotic Nose Patrol


Feb 2020, phys.org

"Romantic Partner" they call it.
155 people, sweaty t-shirts, you get the idea.
You don't even have to be there, just your shirt: "effect similar to taking oral melatonin supplements."

I'm going to chime in here and say this is something to do with security and familiarity. Knowing that another member of your tribe is with you while you sleep makes you less paranoid of the silent creepers waiting to attack you in the dead of night, be they lions, tigers or viruses. And your body knows who is a member of your tribe via both olfaction and your immune system (which are inter-related by their chemosensing inputs).

Your immune system knows who's deep inside your network because their microbes are deep inside your body crawling and growing and metabolizing all over your face, in your nostrils, under your fingernails, and throughout the entire length of your digestive system. Your immune system has one job, and that's to keep the xenobiotics at bay.

Yes, your immune system is hyper xenophobic. As you move through life, you pick up all kinds of microbes from your environment and especially from the people around you. All the while your immune system is working like crazy to detect, attack and remove all those microbiological trespassers.

If you spend enough time with certain people, then your immune system will need a way to identify those people and their biomes so that it doesn't waste time attacking things that aren't actually a threat. It's a probability game -- if you hang out with these people a lot, then they probably aren't a threat, and neither are their microbodies; you'd be better off saving your energy for another time, like when you're at the gym swimming in a sweaty soup of other people's effluence.

Point being, the more immunologically exposed you are to another person, the more your immune system will "know" who they are by the way they smell, and even while you sleep. And when your immune system is relaxed, so are you. (And yes, immunological exposure means exactly what it sounds like.)


Post Script
Stanford Medicine, Sep 2018

For two years straight, Michael Snyder, MD, professor and chair of genetics at Stanford, sported a peculiar accessory — a little gray box strapped to his bicep taking sips of air and recording his exposome cloud.

It turns out, at any given time, we are bombarded by a combination of microbes, fungi, chemicals, viruses, particulates and even tiny microscopic animals, a new paper in Cell reported. This whirling plume of particulates is called the human exposome.

The long-term goal, Snyder said, is to simplify the device into something that resembles an exposome-monitoring smart watch that can suck up and analyze the atmosphere on its own.
 
Posted by at No comments:
Labels: , , , , , ,

Thursday, October 8, 2020

Full Cat Moon Revisited




It's that time of year again where the smell of feral cats marking their territory scents your local neighborhood. Only it's not actually the cats, it's the plants themselves. English boxwood to be specific. At the right time of year (which is now in New Jersey), these popular shrubs are announcing their presence via the sharp smell of a well-used litter box.

The offending molecule, called “cat ketone,” is also a main component in blackcurrant. Pour yourself a dram of blackcurrant desert wine, take a sip, and then smell the inside of the glass. And there you'll have it. Now take a walk around the block (anywhere from August to October), and pay attention. Once you tune in, you'll notice it every time you pass one of these bushes.

Sometimes we actually like bad smells. Rather, we learn to like them. Maybe it has something to do with our ability to control the biosphere so expertly. We have refined the practice of fermentation such that it has become an integral part of our diets. Bread, beer, cheese, kimchi, sauerkraut, chocolate; all these things are fermented. Fermentation is controlled rotting, and it's unnatural. (It's described way better in this book about fermentation that won the James Beard Award in 2013 for Reference and Scholarship.)

Uncontrolled rotting makes bad smells. Controlled rotting still makes bad smells, but less. And we learn how to ... re-evaluate those bad smells in light of their finer attributes. "Parmesan cheese" (isovaleric acid) comes up a lot in regards to this. If you put isovaleric acid in a hundred small bottles and give them to a hundred people, half will call it gross and the other half will love it. I call it quantum hedonics, but you can just call it an acquired taste.

image source - natgeo

Here’s  a picture of a boxwood shrub:


Taking this from an article by Smell-Linguist Asifa Majid:
The Chukchi and Yupik of the Beiring Straits eat fermented fish, reindeer blood, and walrus fat and have even been said to have a preference for partially decomposed food.* This is reflected in vocabulary too: for example, Chukchi veglyt’ul ‘old edible’ versus pegyt’ul ‘old, should not to be eaten’; ... .
*Yamin-Pasternak, S., Kliskey, A., Alessa, L., Pasternak, I., Schweitzer, P. (2014) The rotten renaissance in the Bering Strait: Loving, loathing, and washing the smell of foods with a (re)acquired taste. Current Anthropology 55: 619–646.

In the Soviet era, some indigenous people were no longer exposed to these foods and odors; thereafter when the Soviet Union collapsed, younger people who had to go back to the fermented foods (or starve) had difficulties ingesting these potent odors. This goes to illustrate the importance of early experienced environmental odors.*
*Beauchamp, G. K. (2014) Foul odors of rotted food: Lessons from olfactory physiology. Current Anthropology 55: 634–635.

That all being said, it's not a bad thing that your blackcurrant apertif smells like civet; it's enriching your experience, and that's a good thing, right? Or would you rather not know?

“cat ketone”
4-thio-4-methylpentan-2-one

And here's a similar molecule that you can get as part of the AROXA Beer Taint Kit:

“buchu mercaptan”
p-mentha-8-thiol-3-one
Catty, like blackcurrant juice or tom cat urine, tomato plant

Notice below how this substance is described on the Good Scents Company directory for blackcurrant. They are used by the flavor and fragrance industry, so they are less likely to use the words "cat piss." "Catty" will suffice!

black currant bud absolute
powerful
impactful
green
incense
spicy
woody
herbal
berry
fruity
blackcurrant
blackberry
jammy
currant bud
catty
narcissus
boxtree
sulfur
sulfered family
unusual fruity aroma

Notes
The Art of Fermentation: An in-Depth Exploration of Essential Concepts and Processes from Around the World
by Sandor Ellix Katz

Winner of the 2013 James Beard Foundation Book Award for Reference and Scholarship, and a New York Times bestseller, The Art of Fermentation is the most comprehensive guide to do-it-yourself home fermentation ever published. Sandor Katz presents the concepts and processes behind fermentation in ways that are simple enough to guide a reader through their first experience making sauerkraut...
-goodreads link

Post Script
Civet is a kind of animal musk taken from a kind of cat/mongoose. It’s used in perfumes because humans like the way it smells. Maybe it reminds us of the “quantum hedonic” natural body odors that we wash off every day. Civetone is the name given to the molecule that most represents the Civet smell.

Post Post Script
National Geographic attracting jaguars using Calvin Klein’s Obsession for Men 1986.


Posted by at No comments:
Labels: , , , , , ,

Tuesday, September 15, 2020

The Origin of Artificial Olfaction


Apr 2020, phys.org

MIT researchers have a new and better way to compress models.

It's so simple that they unveiled it in a tweet last month: Train the model, prune its weakest connections, retrain the model at its fast, early training rate, and repeat, until the model is as tiny as you want.



In other words, in order to make a more efficient artificial brain, you grow it from scratch, like a person. 

This is a welcome development for artificial olfaction enthusiasts, because we won't see fully-functioning synthetic olfactory systems until we can first get a "lifetime" worth of autobiographical data for that system. 

That feeling you get when "the smell of grandma's attic" hits you, it will not work if you didn't have a grandma. The data used by an olfactory system, artificial or otherwise, will come not only from infinite odorous molecules and their physiochemical properties, but also from the limbic system. And not just a limbic system, it has to be one that is preloaded with physiological datapoints as they relate to different combinations of odorous molecules. That requires a lifetime of matching bodily experiences, social experiences, and ultimately autobiographical moments to odors. 

To decode olfaction is not so much a phylogenetic (species) problem as an ontogenetic (individual) problem. There is so much variety in the way we perceive smells, that to use a phylogenetic approach would exclude the majority of what makes smell such a powerful experience. It needs meaning; it is by nature subjective, not objective. In other words, it needs a subject, and in ways that other senses can do without (see object recognition, for example).

Image source: Hiroto Ikeuchi cyberpunk

Notes
Comparing Rewinding and Fine-tuning in Neural Network Pruning, arXiv:2003.02389 [cs.LG] arxiv.org/abs/2003.02389

Post Script
June 2020, phys.org

"We study spiking neural networks, which are systems that learn much as living brains do," said Los Alamos National Laboratory computer scientist Yijing Watkins. "We were fascinated by the prospect of training a neuromorphic processor in a manner analogous to how humans and other biological systems learn from their environment during childhood development."

Watkins and her research team found that the network simulations became unstable after continuous periods of unsupervised learning. When they exposed the networks to states that are analogous to the waves that living brains experience during sleep, stability was restored. "It was as though we were giving the neural networks the equivalent of a good night's rest," said Watkins.
Posted by at No comments:
Labels: , , ,
Subscribe to: Posts (Atom)