The New York Times Cooking app has launched a built-in AI assistant designed to generate recipes and answer cooking questions directly within its interface. This feature sits alongside existing tools like Alexa Plus and Samsung Food, offering users a dedicated chatbot to handle meal planning and ingredient substitutions without leaving the application. The system successfully suggested a chicken dish for a weeknight dinner, yet it nearly failed when asked to bake a cake. The AI confused standard baking instructions with a different style of cake, resulting in a recipe that would have ruined the batch.
This incident highlights the current limitations of generative models when applied to precise culinary tasks. While the assistant functions well for general guidance, it lacks the rigorous testing required for specific recipes where measurements and techniques must be exact. Users relying on these tools for baking or complex dishes remain vulnerable to errors that a human editor would catch. The integration shows promise for convenience, but the potential for costly mistakes in the kitchen remains a significant concern for home cooks.
* The model struggled to distinguish between cake styles during generation.
* Specific measurements in the output were inaccurate for the requested recipe.
* No immediate safeguards prevent users from following flawed instructions.
