The work is very interesting. The title is misleading.
A better title would be: "all of human ingredients compressed into 1,800 primitives"
There is little to substantively nothing about the actual cooking: preparation methods, proportions, etc.
But the idea that tomato goes well with beef the whole world over is very interesting and useful for creating flavors that will go together, perhaps surprisingly. It will be a nice resource in the future.
I have a wonderful book that explores this idea of an atlas of flavours that work together.
The flavor bible.
I can assure you that it does not contain 1800 ingredients in all of there combinations, but it does a remarkable job of covering a widely used selection of herbs spices vegetables and meats. I doubt a compressed version of the text would even be very large.
The trouble I find with LLM generated recipes is they miss the nuance of the technique. Often the success of a depends on a single step or ratio. For instance “fried chicken” has a million incarnations the world over, but you can’t just average out the recipes and end up with tasty fried chicken.
Ruhlsman's "Ratio" is also quite good at distilling the mechanics of food into an algorithm of sorts.
https://www.simonandschuster.com/books/Ratio/Michael-Ruhlman...
yeah I took a look at this and others and tried to pull together some helpful 'flavor maps':
https://transcendent-choux-d1b930.netlify.app/
Have you also come across this? ‘Flavor network and the principles of food pairing’
https://arxiv.org/pdf/1111.6074
Sounds like The Flavor Bible - it's a great reference book to find pairings of ingredients.
> Sounds like The Flavor Bible - it's a great reference book to find pairings of ingredients.
I think fpshero says exactly that (https://news.ycombinator.com/item?id=48294459):
> I have a wonderful book that explores this idea of an atlas of flavours that work together.
> The flavor bible.
> The trouble I find with LLM generated recipes is they miss the nuance of the technique. Often the success of a depends on a single step or ratio. For instance “fried chicken” has a million incarnations the world over, but you can’t just average out the recipes and end up with tasty fried chicken.
Specify what technique you want. Explicitly say you want to correctly follow all the techniques of the chosen cuisine.
All the LLMs have ingested nearly every cookbook ever made, across multiple languages.
You can upload a photo of your spice rack (with visible labels) to ChatGPT and tell it to save your pantry ingredients as a memory.
LLMs are absurdly overpowered for cooking, when used right. If you ask it for a week long meal prep plan the results will be meh, but ask it for kheer inspired rice crispy treats (which everyone reading this should to, kheer rice crispies are the best!) and you'll get some solid results.
You may notice at first the LLM will still water things down for "American" tastes. With Claude/ChatGPT you only need to remind it once or twice not to do that and it'll course correct all future conversations.
> All the LLMs have ingested nearly every cookbook ever made, across multiple languages.
That's not a positive thing, good recipe developers are Rare. For every recipe that's been meticulously tested and documented there are 1000 that haven't been. Many cookbooks are riddled with errors.
Sounds like the flavour theaurus
Sounds like the flavour colour wheel.
Sounds like the flavour roulette wheel.
This is exactly like when Skippy is trying to work out the flavor profiles for the kinds of mush that monkeys will find most appealing. Yes, we know that "chocolate peanut butter bananas" is the true king of flavor latent space, but even a slight error in floating point precision could stick you with "cantor's peach butt nut abalone".
Unless I'm missing something, there's also nothing in the paper to indicate this is "all of human ingredients"? It looks like it's 11 data sources covering a bunch of common cuisines, with the English + Chinese sources accounting for 90% (!) of the dataset. Among others, Africa and the Arab world are not present in the data (good for about 25% of the global population).
Also, all non-English terms were AI-translated to English which is methodologically understandable but surely leaves room for error.
translation is an interesting problem in and of itself still. its kind of a miracle we can do it at all, yet in some circumstances it seems obvious for there to be objective answers (cooking ingredients being one of them), but even then you never really know even with human translators if you've got it correct. even within the same language nearly every individual has their own version of it.
for example, how would you translate "chips" to another language without first knowing which version of English you are translating from? could be an american speaker with a british relative and they use the british definition of chips while otherwise mostly speaking american english.
there's a level of pragmatism in translation that needs to be assumed, and ultimately we have to accept that translated knowledge will always have low resolution. There is a layer of work that needs to be done with the source of the materials involvement to get written content to a level of formalism needed to be representative of the language it is written in. Generally, the work of editors. Which means successful translation for wide distribution, while still not guaranteed, is predicated on the editorial skills of the translator which begs for dialogue with the source.
Meanwhile, AI provides this super convenient band aid to get translation results you can't disprove.
I genuinely think people are severely underestimating the power held by these models for being translators and how literal truth is going to be determined by them deep behind the scenes under the disguise of accessibility. Not in a dangerous way necessarily, just in a way where what languages are and what words mean is going to shift towards whatever the models think they are.
In a way, over extended time, the models will not be wrong about the translations because their results will redefine what successful formal editing of language looks like, and disagreeing with them will amount to the same difference as having local slang.
Leaving out Indian, Southeast Asian and Arab cuisine means this is nigh useless.
There are 2,000+ varieties of mangoes alone. You could literally end up with a larger file using only mangoes.
Was going to give the same example with chili peppers. Tons of varieties and not exactly interchangeable
> But the idea that tomato goes well with beef the whole world over is very interesting
I saved a beef stew I was making for twelve people once by adding tomato sauce.
Beef hardens if stewed incorrectly and tomato acid tenderises it again.
EDIT: removed incorrect information about store bought tomatoes.
I have a hard time believing that that weak an acid can have a noticeable effect vs just the extra cooking time. Sorry for being a skeptic :-) Most of the time I deal with beef either it's the particular chunk of meat I bought that day or marinating it in salt for over a day or just stewing it for a long amount of time.
If you are interested in that you might want to check out this paper:
https://www.nature.com/articles/srep00196
I would like one day to have a database which measure how strongly every food ingredient in use binds to every human smell receptors.
It's also based on 11 sources across just 7 of the 7,150 human languages (English, Chinese, Russian, Vietnamese, Spanish, Turkish, Indonesian, German, and Indian-English)
One that has long tickled me is cabbage +/- pickling. I eat both sauerkraut and kimchi from the jar and enjoy them as additions to _roughly_ the same foods, and when friends/family ask I insist they are basically the same thing anyways, but they are uninterested in such shenanigans. I'd love to learn more about these cross cultural shared foods.
Tomatoes are high in glutamate, which accentuates beef flavor.
+1.
On a side note (and maybe off topic), I am thinking about food pairing which is based on the idea that two ingredients sharing volatile aroma compounds or certain molecular families may have a potential sensory compatibility (broccolis and strawberries for example). I'd love to test those ingredients and find some unknown food pairings. But .. time is what it is (for now).
Wouldn't it be great if we had a simulator like the MIT violin simulator [1] but for cooking ingredients! Then you don't have to throw out pounds of perfectly good ingredients just because broccoli doesn't go with Nutella.
[1] https://news.mit.edu/2026/mit-engineers-virtual-violin-produ...
I think it's a lot simpler than that. A common pattern for sauces is fat + acid + salt
- (Mexican) avocado and lime/lemon + salt
- (Chinese-southwestern) chili oil and vinegar + salt/fermented bean paste
- (Italian) olive oil and tomato + salt
- (Turkish) olive oil and lemon + salt
- (Thai) coconut milk and lime + salt
...
https://www.saltfatacidheat.com/
This book was an eye opener for me. Obvious in retrospect; I wondered how I did not notice it myself.
I was gonna post the same, as a lifelong cook (and eater!) Samin Nosrat's book/show was essential for giving me the confidence to improvise in the kitchen while retaining authenticity to regional cuisines.
> all of human ingredients
Depending on who you ask, this may also sound misleading
It's cheaper to train a robot how ingredients go together than to cook for humans.
can anyone check whether pineapple goes well with baked dough with cheese/tomato sauce?
only in Hawaii ...