Machines taught how to 'smell' by new algorithm. How will they cope with shower-dodging nerds?

Machines don’t have noses – but they can now attempt to identify scents thanks to a nifty new algorithm. A big international group of scientists pooled together to launch the Dream Olfaction Prediction Challenge, a crowd-sourced competition to learn how to predict the scent of molecules. Essentially, the idea is to craft …

[Haku]

How will they cope with shower-dodging nerds?

Easy.

Said nerds will simply hack the code to make it ignore certain smells.

[Anonymous Coward]

Dreams

It is my experience that dreams do not replay smell/taste as an actual experience. I can "know" that a dream contains a smell/taste but there is no reality to it - unlike the pictorial or audio senses which appear as in waking life.

Yet in waking life the Proustian moments work in both directions - visual (sound?) memories trigger smell/taste - and smell/taste triggers visual (sound?) memories.

Is that the same for everyone?

[frank ly]

Re: Dreams

I can't recall ever experiencing the sense of smell in a dream or ever expecting it or wondering about it in a dream.

[Mage]

Re: Dreams

Smell is rare to non-existant in dreams

Colour seems pretty rare or a bit like a colour film on a B&W TV you know well and you are tired. It's not clear how much colour perception there is.

Real sound /dialogue, like real life or a recording isn't common.

On the whole dreams may be more immediate and vivid than remembering, to the point that it seems real at the time. They are much less realistic when you wake.

Memory or Dreams are not much like a video recording, far more disjointed and erratic in detail per scene. I'd imagine the 3rd party viewing if you had a device to show / record dreams or memories would be disjointed and almost incoherent. In that respect most SF or "psychic" stories on the subject seem unrealistic.

Smell also depends often on the appearance (correct colour too) and temperature of an object, also if obviously real or fake. Taste / flavour depends on smell, texture, colour and temperature of food.

Both are affected by previous exposure to something and to an extent by emotion and health.

Taste and smell are the least reliable and consistent of senses.

[Anonymous Coward]

Re: Dreams

"Colour seems pretty rare or a bit like a colour film on a B&W TV you know well and you are tired. It's not clear how much colour perception there is. Real sound /dialogue, like real life or a recording isn't common."

Recently I have not been sleeping too well - one of the joys of old age where one wakes after a couple of hours. The dreams are now definitely in colour and dialogue does take place. They are episodic - like the G&S nightmare lyric - suddenly changing to a new scene. Each scene itself is relatively coherent - and the overall thread through all the scenes is consistent too. During the day I can get flashbacks to parts of the dream.

One the other night involved a meal. At one point some petite fours appeared on a plate. I picked a green one that had a pin holding the halves together. In my dream I was wondering why I couldn't get my tongue to separate the pin out - and why I couldn't taste anything.

[jake]

Re: Dreams

For me, when I smell something from my childhood, I can remember sights and sounds surrounding that scent. The more pleasurable, the stronger/more complete the memories. But when I see or hear something, I "know" what it should smell/taste like, but I can't actually smell/taste it.

For example, when I eat/smell my home-cured salmon, I'm instantly back in my grandfather's kitchen the first time I tasted it (I learned the process I use from him). I can remember our conversation word for word, and can see the flowers Gran brought in dropping pollen on the fish. However, when I look at the picture of myself and grandpa unloading the salmon from the smoker 50 years ago (.fav photo of mine, 8x10 on the wall of my office), I can't actually smell/taste it BUT I know exactly what it smells/tastes like. Make sense?

I have dreamed of eating smoked salmon in the past. It tasted of smoked salmon. I made a batch the following day, the craving is quite powerful ...

[A Non e-mouse]

Statistics idiots

For those of us who aren’t statistics nerds, some comparison values for the Z-Score would have been helpful. Also, is the scale linear, logarithmic, etc?

[LionelB]

Re: Statistics idiots

A z-score is a measure of how far an observation is from the mean value of the actual data (the "population"), adjusted for how "spread" the data is. The spread (or dispersion) of a distribution of data is frequently measured by its standard deviation (for a quantity that follows a standard Gaussian distribution - a "bell curve" - it's roughly half the width of the bell). The z-score is then defined to be the difference between a measurement and the population mean, divided by standard deviation (it's a dimensionless number). The rationale is that in assessing the deviation of a measurement from a population mean, you need to account for the fact that the more widely dispersed the data, the further a random data point is likely to be from the mean just by chance.

Well that's the standard version of a z-score... according to which, the scores quoted in the article seem to me to be far too large to justify its claims. As an example, the mean height of adult males in the UK is around 175.5cm, with a standard deviation of around 7.5cm. If I measured a man's height and calculated the standard score as 9, this would mean the man was 243cm tall - way off the population mean.

So I had a look at the original article in Science, and it turns out that their "z-score" appears to be non-standard. I say "appears to be", as what they actually did is atrociously badly described; I quote: " We assessed the performance of models submitted to the DREAM Challenge by computing for each attribute the correlation between the predictions of the 69 hidden test molecules and the actual data. We then calculated a Z-score by subtracting the average correlations and scaling by the standard deviation of a distribution based on a randomization of the test-set molecule identities". Now I review a lot of research articles, and, sorry, but that would be an instant fail. When they say "subtracting the average correlations", subtracting from what? From 1, maybe? And what exactly is this "distribution based on a randomization of the test-set molecule identities". A distribution of what, exactly? There is more detail in the supplementary text, but it still leaves me slightly baffled. I suspect that it is probably statistically sound, but it leaves me unsure as to what constitutes a "good" z-score under their procedure. Hmph. Science (the journal and the Thing) ought to be better than this.

[Martin an gof]

Largely subjective?

I have often thought that the sense of smell is the least "common" between people, and it does depend to a huge extent on what you are used to. About the only thing that most people seem to agree on is that certain smells are "off" and are usually a warning of rotten food or similar, even if you subsequently learn to ignore it in certain circumstances (cheese comes to mind).

I, for example, get a sharp pain behind the eyes when I smell lavender which I sometimes describe as akin to sniffing vinegar, and prolonged exposure can lead to a headache, but the vast majority of people seem to find the scent relaxing and calming.

It seems to me as if perhaps the underlying mechanisms (software and hardware) that the brain uses to categorise / understand senses have a lot in common(*), just with slightly different inputs, and that perhaps this might help explain synaesthesia at one level and the simple cross-triggering of memories described by the AC above at another. I have wondered if the problem I have with lavender is connected to the brassica problem; the way that some people (me included) taste brassica as unpleasantly "bitter", while others don't.

I've wondered too about the language we use for senses. My own languages are essentially confined to English and Welsh so perhaps others can comment on other tongues, but in Welsh it is not uncommon to "mix up" the words used to describe sensing; thus one might "hear" a smell. It's possible it's a dialect thing I suppose.

(by the way, "oder" in the sub-head? Surely "odour"?)

M.

(*)consider, for example, the "desensitisation" issue - all the senses are able to "mask out" constant inputs if they are relatively low-level; your eyes become insensitive to things that don't move, while reacting quickly to variants; your ears "mask out" background noises; your sense of touch reacts at first to extremes of hot or cold but after five minutes in that "piping hot" bath you begin to wonder about letting in some more hot water; you can completely fail to notice a smell that's a constant, but which visitors comment on immediately (for example, when the farmer's been muck-spreading on the field next-door)

[Anonymous Coward]

Re: Largely subjective?

"the way that some people (me included) taste brassica as unpleasantly "bitter", while others don't."

There is an artificial sweetener that some people can't taste. IIRC it has been correlated with a particular gene.

The ability to digest milk easily usually switches off after childhood - except in those pastoral society people who have evolved with a faulty switch. There's also a particular gene inheritance that means some people are poisoned by eating broad beans.

[Martin an gof]

Re: Largely subjective?

(re: taste) IIRC it has been correlated with a particular gene

I'd heard such stories about taste too, but isn't it interesting how these things seem to affect smell and taste more than other senses? I suppose the point I was trying to make was that trying to create a measure of how the "average" person senses a smell misses the point - nobody is average :-)

Regarding the milk thing, I'd always thought it was more of a "learned" response, and certainly not a "faulty switch"! The ability to tolerate lactin (or whatever it is) is always there - it has to be otherwise babies wouldn't survive - but the tolerance reduces if your exposure to the stuff reduces. This seems to be behind those theories that claim an apparent increase in "auto immune" type problems (asthma, allergies) and even susceptibility to coughs and colds might be explained by lack of exposure to allergens and bacteria (i.e. spotless homes). Isn't there a trial at the moment that seems to be suggesting that even severely nut-intolerant people can be "trained" to tolerate nuts, simply by giving them gradually increasing exposure?

That would correspond with the fact that I'm (mildly) allergic to cats - I had a test to prove it - yet when I lived with cats at my parents' house it was rarely a problem. Yes, I'd come up in welts if the things happened to scratch me, but no worse. Now, having lived without cats for some years, I find that I can begin to sniffle and my eyes begin to water if I'm in a house with cats for any length of time.

M.

[Mage]

Re: Milk

The dairy issue is ethnic. Only a minority of the world's population (mostly Ethnic European so maybe Neanderthal genes?) can tolerate non-human milk, or milk at all when adult, though goat and sheep dairy may be less a problem.

Milk (particularly) is designed for babies. Cheese and yoghurt production not only "preserve" milk (most of all hard cheese), but actually make it more digestible for Ethnically non-Europeans.

It's nothing to do with if a society is pastoral or not.

Most raw beans have varying levels of toxin depending on species, maybe peas too (peanuts are a legume not a nut). Raw Kidney beans are a bad idea!

[Anonymous Coward]

Re: Milk

"It's nothing to do with if a society is pastoral or not."

Pastoral societies in the past would have been the ones where lactose tolerant adults would have been able to add the available milk to their nutrition. Therefore the gene variant would have been propagated by evolutionary selection processes.

[jake]

Re: Largely subjective?

"I have wondered if the problem I have with lavender is connected to the brassica problem; the way that some people (me included) taste brassica as unpleasantly "bitter", while others don't."

Some people have fewer bitter receptors in their mouth than others. I can't remember which of the human haplogroups first got that mutation.

[jake]

Re: Largely subjective?

Re: Lavender ... My wife just told me she has two cousins, identical twins, who are subject to migraine. One of them can get one from smelling an over-abundance of lavender. The other one finds the scent of lavender relieves the headache ... the stronger the scent, the better the relief.

Humans. Go figure.

[TRT]

Luca Turin's your man

Great guy. He came up with a hypothesis about nuclear spin resonance rather than bog standard chemoreception. Shunned by the scientific community but I reckon he has something. This could prove it.

[Anonymous Coward]

Rescue operations

I wonder if this could be made a useful resource for rescue workers.

If they could make them anywhere near as accurate as a dog then you could also attach smelling devices to places dogs can't get too/send in swarms of them etc.

[LionelB]

Re: Rescue operations

Just train insects - some moths, in particular, have exquisitely sensitive olfaction, almost down to the individual molecule level. Okay, training might be a bit tricky...

[Anonymous Coward]

Cheezoid?

[William Towle]

Next: Whisky?

Automated generation of tasting notes, anyone? And if you need help calibrating, I'm available.

// recalls whisky tour of Scotland in which we devised a challenge to learn the ingredients list of our favoured breakfast cereal and claim to taste those things in one of the sessions*. If any of us was sufficiently disrespectful to any of the distilleries to actually go ahead with it, they were very subtle...

* Fruit'n'Fibre in my case, easy to learn :) but would have been easy to spot too!

[razorfishsl]

My mother ate a shed load of blue Stilton whilst she was pregnant.

It is now the only cheese i will eat,

[Calimero]

Abandon the brain and move $$$ to the nose? How far down?

We have not figured out how the brain works - or was I missing that big news while watching The Stepford Wives?

For those few, apparently, who skipped the biology class: the nose smells what gets *inside* the nose (yes, a surgery mask will keep you happy during the flight). So how would then the SW get those molecules inside *its* "nose" first then inside ours - I guess will have to call it smellware.

Before we move down to the mouth and I read in Science about the SW correlation and correctness of smell and taste: the taste buds do for the mouth/taste (and brain) what the cilia do for the nose/smell.

Anyone wants to bet that we will read about TW (tasteware) in the next 6 months?