AI Fails at Cooking: Why It Can’t Replace Chefs
▼ Summary
– Moving a struggling sage plant to a sunnier spot led to abundant leaf production, prompting the need for sage-heavy recipes.
– The author used DishGen, an AI cooking platform, to generate a recipe for “sage infused brats skillet with caramelized onions.”
– DishGen repackages recipes from large language models into kitchen-friendly formats but lacks detailed instructions, requiring user experience to fill gaps.
– The recipe’s vague directions (e.g., onion slicing, sage measurement) and lack of descriptive cues made cooking frustrating and unclear.
– The author contrasted DishGen’s shortcomings with well-tested recipes from sources like The New York Times Cooking, which inspire confidence and clarity.
Artificial intelligence may be transforming many industries, but when it comes to cooking, human chefs still reign supreme. My recent experiment with AI-generated recipes revealed some glaring limitations that no algorithm can yet overcome.
After relocating my thriving sage plant to a sunnier spot, I found myself with an abundance of leaves, the perfect excuse to test an AI recipe generator. Typing “recipe using brats and lots of sage” into DishGen, an AI-powered cooking platform, yielded a promising-sounding dish: “sage-infused brats skillet with caramelized onions.” The description was enticing, but the execution proved far from flawless.
While DishGen and similar tools like ChefGPT and Epicure aim to simplify meal planning, their instructions often lack the precision and nuance of human-written recipes. The suggested dish called for just two tablespoons of sage, hardly “lots”, and left out critical details. For instance, the recipe instructed me to use a “large yellow onion, thinly sliced” without clarifying whether to peel it first, how to cut it, or what “thinly” actually meant. Seasoned home cooks might improvise, but beginners would likely stumble.
Even basic phrasing became a stumbling block. The ingredient list specified “2 tablespoons fresh sage leaves, chopped,” raising questions: Should I measure whole leaves before chopping, or chopped leaves after? A well-written recipe would avoid such ambiguity by clearly stating “2 tablespoons chopped fresh sage leaves.”
The cooking process introduced more frustrations. The instructions directed me to caramelize onions in butter for about 12 minutes but offered no guidance on heat level, stirring frequency, or visual cues to determine doneness. When the dish was finished, the portion size seemed meager for four servings, a detail my wife immediately noticed. “Is that all of it?” she asked, eyeing the modest serving.
This experience highlights a fundamental truth: great cooking relies on intuition, experience, and attention to detail, qualities AI still struggles to replicate. For comparison, I searched The New York Times Cooking app for sage-heavy recipes and found Samin Nosrat’s fried sage salsa verde. Just reading the title triggered an almost instinctive craving, a reaction no algorithm can yet inspire.
While AI recipe generators provide convenience, they fall short in delivering the depth, clarity, and sensory richness that make cooking truly rewarding. For now, human chefs, and well-crafted recipes, remain irreplaceable.
(Source: Wired)