Is there some device that can analyze an arbitrary sample and create a “taste profile” of it that can later be reproduced? Or would something like a mass spectrometer be able to distinguish the compounds in a thingie enough that scientists would be able to guesstimate how it would taste to a human without actually having to put it on their tongue?
If you must know, this thread was inspired by my other one… I wonder how this would taste…
The mass spectrometer is going to be the closest thing you could have.
Consider what taste is… It’s a purely subjective response to what our chemical receptors (tastebuds and olfactory nerves) determine what’s in a substance. We instinctually consider edible items to “taste good” and inedible or dangerous items to “taste bad”. We further have widely varying opinions on what is good or bad, and some palates are able to distinguish flavors others can’t. We try to categorize flavors by assigning them various labels (spicy, sweet, savory) or saying it “tastes like” something else. You can’t objectively measure something like that with a machine.
What you can do is use a machine to accurately and objectively identify the chemical composition of a substance which is what our sense of taste is crudely trying to do. And that function is exactly what a mass spectrometer does.
So I guess the best way to summarize is that no, machines can’t analyze taste, but they can do better than taste.
The answer to the rest of your question is simple. If a scientist can determine that something has cinnamon in it and knows what cinnamon tastes like then he can probably make a reasonable estimate about what it would taste like without putting it on his tongue. (It would taste like cinnamon.) The same way he could conclude from a noise dosimeter that rated a noise at 90 dB would be “loud” without listening to it himself.
There are a lot of newer machines that are specialized to “taste” a particular product. They have a variety of electronic sensors and typically aren’t doing what a mass spectrometer does to determine the chemical makeup of a product.
The machines are very specialized though, and are typically used for quality control for something like coffee, wine, or beer production. The machines can recognize a good tasting product vs. a bad tasting product because they have been programmed to recognized the difference, but these machines most definitely could not taste some arbitrary product and tell you if it tastes good or not (in other words, a beer tasting machine wouldn’t have a clue what good coffee is supposed to taste like).
While these machines do exist, they aren’t in widespread use. I’ve done a lot of work in large scale breweries and the quality control was basically done by taking small samples of each batch and letting employees taste them in the lunchroom during breaks. While employees were strongly encouraged to taste samples, they were not required to, and excessive oversampling (aka getting drunk) was not only strongly discouraged but could get you fired.
I used to work for a big food company, and we occasionally used mass spec/gas chromatography (MS-GC) for food analysis. It was pretty much hit or miss. We would get a print out of which compounds appeared in food samples with a corresponding set of possible flavors that the compound is associated with. There would be a bunch of compounds, which tended to group into a number of sets of (maybe) identifiable flavors.
There were some attempts to teach the machine by using samples spiked with known flavors, which would be used to help identify what was in the product. A coworker had success identifying an unknown flavor in a competitor this way (white pepper). The task in general was only somewhat helpful. A compound at a level of 100 may be a very weak flavor, while another compound at a level of 0.1 may be a very strong flavor. So sorting through all of the compounds and their levels rarely led to startling insights. They did help a bit in some cases though.
As far as the basic flavors of salt, sweet, and sour (and maybe bitter), they were usually pretty easy to test for.
A chemical test could give you an accurate figure for the amount of salt (sodium) in the sample. But what tastes too salty to me might be just right for you. Same with sweetness, bitterness, etc.
And besides this variation between individuals, it can also vary within a single individual based on expectations. The amount of sodium that I find as ‘too salty’ in beef gravy might be just fine in fried rice with soy sauce.
As I understand it, sight, hearing, and (probably) touch are pretty linear sensations. Hearing is simply a matter of detecting fluctuating pressure. There’s more pressure, there’s less pressure, but it’s a single variable. Lightwaves effectively, could be classified by two variables, wavelength and strength. Our eyes have opted to specialize receptors for certain ranges of wavelength, so that the variable count is expanded, but it wasn’t technically necessary for that to be done. Touch is basically, again, just pressure. Hearing and touch both have a “location” variable, I suppose, but still it’s pretty straightforward what the signals are that are being sent into the brain.
Now a molecule is a set of atoms that are bonded together, into a particular shape. Swap out the atoms with different ones, but preserve the shape, and you have a different molecule. Use the same atoms but arrange them into a different shape and you have a different molecule.
Taste and smell are based on having receptors that react with a wide variety of different molecules. It’s sort of like those wood boards with holes cut out for square blocks, round blocks, and triangular blocks. If you try to put a block that doesn’t fit into one of the holes, well it doesn’t fit, it doesn’t go in. Find the right shape and it will go through. In the case of our senses, the wood board (e.g., our tongue and olfactory system) has been shaped to have slots for specific molecules and if a molecule which is floating by fits into the keyhole, then a signal will go to the brain. If it doesn’t fit, it will just bounce off and keep going past different slots which, again, may or may not fit the shape of the molecule (or some branch off of it, like a thiol).
In a sense, this is a simpler system than the other ones, since it’s just a “present” or “not present” (boolean, for programmers and mathematicians) variable of whether a molecule has slipped in to fit the slot. Other senses are a range of values. But, there may be hundreds or thousands of different slots that support different types of molecule, and there are probably thousands or millions of duplicates of each type of receptor and so a “stronger” signal is sent when more of these sensors are triggered at once for one particular type of molecule, so we can tell how large a component of the smell/taste the molecule is.
But so basically each molecule that is supported is like its own sense, somewhat independent of every other type of molecule that can be detected. Or, you might think of it like our light sensors, where we have three (or four, if you count our low-light grayscale sensors), each receptive to a particular subset of all possibilities, except we have hundreds or thousands of different types of them instead of just three or four.
A machine would need to work in a somewhat similar way, being able to interact with each type of molecule that we can detect, and able to interpret it in a similar way to how we are hardwired to detect it. This means either having machinery which can image molecules and recognize what all bits could interact with a human receptor, or having microscopic receptors of its own. And it means determining how humans are hardwired to interpret different molecules. Some of those are probably straightforward. Thiols are universally bad. Nearly all humans are instinctively put off by receiving a thiol. But some molecules are probably left to experience or are a bit randomly defined. Potentially, they vary by country and what their cuisines consist of. One molecule might be a marker of “delicious” in one cuisine and a marker of “spoiled” in another. We have to be taught by our upbringing how to interpet those molecules. And so you would need to isolate the molecule and do population studies by country in order to make a determination for how a person would interpret the sensation, so that you could make a flavor or scent for people in that market.
And that’s assuming that we interpret each molecule indpendently of one another. Like light, we probably don’t interpret each of the three colors independently of the other. They’re merged together, by our brain, into a single “color”. Likely, our nose and tongue merge together the sensations to create a unified “flavor” that is a mixture of the subcomponents that make it up.
Making a machine that can recreate this, then, means not just isolating the individual molecules and performing market-specific research for each of them, but also mixing those molecules and testing those out, and using data mining to make determinations of what does and doesn’t “go together”.
Overall, it’s going to take some effort. It’s a lot harder than the other senses.