h a l f b a k e r yCall Ambulance, Rebuild Kitchen.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Evolv-a-Drink
Grind %QUANTITY %INGREDIENT and pour %QUANTITY %INGREDIENT on them | |
Firstly, I would like to apologize for any incoherency in the following idea. I'm currently suffering from a slight sugar rush- Although this augments my typing speed considerably, my brain isn't going quite fast enough to keep up.
*********
This is an idea for the algorithmic creation of drinks,
alcoholic or otherwise.
First, one would randomly create several "chromosomes", each a recipe for some sort of drink. It would probably be a list of ingredients (Coke, V8, Salt, Vanilla Extract, Blue-Flavored Kool-Aid Powder, etc) and actions (Shake, Refrigerate, Microwave, Boil, Stir, Wait, etc), each appended with numbers explaining the quantity to add or the duration to carry the action out.
Each of these drinks would then be made, preferably by some kind of automated bartender robot, and tested, by which I mean drank. Before the testing of a drink, the current state of the drinker would be measured- temperature, hydration level, drunkeness, and perhaps some other factor(s). The drinker would then consume each drink, and then rate the drink on a 200-point scale ranging from "Vomit-Inducing" to "Perfection".
The data from each drink would then be weighted, based on the drinker data collected, and the resulting fitness values would be used to select the recipes to "breed" to produce the next generation.
The chromosomes would be "bred" to produce new chromosomes, "mutations" would be applied in the form of adjustments to the recipes- say, replace Coke with Pepsi, or add an extra gram of sugar, or delete the "shake" step, or throw in one.
The cycle would repeat, particularly delicious variants would be saved, and the drinks would slowly be optimized into perfection.
(Once fMRI technology or other neural monitoring technology advances enough, perhaps one could simply monitor activity levels in the pleasure center instead of relying on subjective and inaccurate data. One might also make this an iPhone application, so that a much larger gene pool could be tested without taking forever.)
xkcd: Recipes
http://www.xkcd.com/720/ [jutta, Apr 17 2010]
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Destination URL.
E.g., https://www.coffee.com/
Description (displayed with the short name and URL.)
|
|
The problem with this sort of in vitro evolution is that you
tend to reach local maxima. In other words, you would risk
evolving towards a drink which was quite unpleasant, but
where changing any one of the ingredients made it more so. |
|
|
If you added a chance for mutation in each few generations,
this would be even better. Also, testing simultaneously in
several different markets could lead to allopatric speciation -
the formation of several drinks at once, each pandering to
regional tastes. |
|
|
I vote that the first triumph of this process be called "The
Dodo". |
|
|
"The Amoeboid Blob" would be more appropriate. |
|
|
Actually, [DrWorm], mutations are already included in the design. Look at paragraph 5 more closely. |
|
|
Oh, and I'm currently trying to implement this. Seeing as I'm so newb at coding that I've never even worked with random numbers, let alone GA's, before, this could take a while. |
|
|
Many years ago, my father was involved in market research on snack foods. He explained to me that, if you give a range of similarly-flavoured products to a focus group, they will usually state a preference for whichever one has the strongest flavour. That's not because they really like it more, but because they're under pressure to pick one, and the stronger flavour is the one they remember. Hence, generally over-flavoured snack foods. |
|
|
So, I suppose my point is complementary to [MaxwellBuchanan]'s; not only might you get local maxima, but those maxima might well not mean what you think they mean. |
|
|
A very apt link [jutta]. Is there no situation that Randall Munroe has not drawn a comic for? |
|
|
I hadn't even known about that. Now everybody's going to think I stole my idea! |
|
|
Cheerios and Vermouth ? see link |
|
| |