The online magazine
dedicated to the
discussion & revival
of British foodways.


Cognitive Cooking with Chef Watson: Recipes for innovation from IBM & the Institute of Culinary Education.

Robophobes despair. It has come to this; a cookbook created not just by computer, but by ‘Watson,’ a specific computer. To compile Cognitive Cooking with Chef Watson , IBM tracked historical ingredient combinations, or “food pairings” (ginger and scallions; potatoes with rosemary), and loaded them along with “tens of thousands of recipes” into its culinary computer. ( Watson 9, 10)

An apparent inspiration has been the current rage for molecular gastronomy. Watson obtained from its human handlers the chemical compounds of each ingredient in its database. IBM also delved into recent research involving “hedonic psychophysics” or, in plainer English, the molecular foundation of human foods and human taste:

“In recent decades, scientists have given people different flavor compounds to smell and taste, and then asked them to rate each compound for pleasantness. This has produced quantified rankings of our wide-spread preferences at the molecular level.” ( Watson 10)

Watson’s programmers understand that none of this means much without something more. Especially in recent years, restless chefs have sough to create novelty through the collision of culinary cultures and juxtaposition of jarring ingredients. Their solution is the attempt simply to outdo the innovators by sifting an avalanche of data, hardly a surprise given the mundane reality: That is what computers do.



Even so the Watson crowd call their database the “surprise factor” based on the combinations their computer would disgorge. “To create bold, new recipes, Watson had to be programmed to produce combinations not yet tasted. This relentless hunt for novelty,” they insist without succumbing to humility, “would fuel its creativity.” ( Watson 10)

As IBM itself concedes in a flourish of puffery, however, “[o]nly talented human chefs… could turn those surprising ingredients into wonderful food.” IBM chose the Institute of Culinary Education in New York as its kitchen collaborator for the Watson project, and “the ICE staff invested hundreds of hours in experimenting and recipe testing.”

Watson may have handed, or rather printed out, the combinations to them but the ICE instructors supplied all the technique, which should appear to most people acquainted with a kitchen to be the key to good cooking. According to Cognitive Cooking , “this collaboration unleashed a torrent of culinary creativity and… never-before-seen recipes…. ” ( Watson 11)

Most of the innovation is more apparent than real. Indonesian Chili Con Carne is recognizably Indonesian but bears no resemblance to chili; the only innovation involved with the Portuguese lobster roll is the encasement of the lobster in gel, which does nothing at all good to the texture of the otherwise ancient marriage of shellfish and pork; an India paella is merely a misnamed pillau; the roast Tomato and Mozzarella Tart is of course a pizza altered only with the unsatisfactory substitution of puff paste for pizza dough. These and other recipes get the wrong name simply because it was input into the computer along with elements of an otherwise traditional preparation.

Another fundamental constraint of the computer culture is the iron law that output depends on input. Apparently the IBM programmers did not deign even to dabble in British foodways because Watson appears ignorant of them. Even an undeniable British contribution to global cuisine, the savory pie, gets a different and unlikely, Scandinavian, pedigree from Watson.

Given the avowed purpose of the project, to create the unexpected and the novel, IBM has missed an opportunity through their omission. British cuisine, with its robust seasonings of anchovy, acid and spice, its juxtaposition of textures and temperatures, could have provided Watson with many more chances to achieve the shock of the new.

Some of the never-before-seen recipes are so gratuitously complex they will not-been-seen-again, at least not in any domestic kitchen. A recipe for roast duck, for example, requires five distinct preparations, while “Carrot Pearls” fashioned from carrot juice, agar, sumac and other spice are prerequisite to the assembly of Turkish bruschetta.

In other cases, festooning a dartboard with the chemical and molecular formulae, the food pairings and compounds which were provided to Watson might achieve results concurrent with those of the computer itself. Some of the recipes, like the one for “Turkish-Korean Anchovy Caesar Salad,” assemble random ingredients.

Tomorrow’s Kitchen, Today!

For this one, “regular old croutons just wouldn’t do,” so jelled artichoke puree “diced, coated with bread crumbs, and fried to look just like a typical crouton” appear in their stead. The salad dressing based on eggplant of all things “tastes like a spicy, sweet tahini.” ( Watson 50) Whatever this thing is, it is not a Caesar salad and the enterprise itself seems something like painting with the elephant dung when, as Brian Sewell noted, perfectly good pigments from Winsor & Newton have been readily available since the eighteenth century.

Watson is not much of an editor. Its book is rife with repetition; the input process is described in indistinguishable terms three times, while the anchovy promised in the Caesar salad gone rogue on an overdoes of tahini fails to appear.

Finally, the humanly generated prose of Cognitive Cooking springs straight from the hideous hyperbole of Marketing 101. Notwithstanding that, however, Cognitive Cooking does not amount to much. Its recipes are too scattered, even incoherent, both in terms of each one and of each other, for the book to make a distinct impression.

It can be done: Heston Blumenthal, for example, has done it with Historic Heston , his equally aspirational but more coherent assay into the strange and the new based on the recipes of the British past. Seventeenth and eighteenth notions become recast with ingredients unfamiliar to their creators and techniques impossible to take on given the technology of the day.



Sean Brock has done something similar but substantially more accessible with the historic foodways of the American south at his restaurant Husk and with his book Heritage. Either author has fulfilled the mission of Cognitive Cooking better than its computer. Perhaps it still takes a person after all.

So despite the sensationalist introduction to this review, Luddites may allow themselves a sly smile and slink back into their analog kitchens lined with their anachronistic books.