AI-Generated Recipes: The Ethics and Limitations of Copying

AI-generated recipes have become a topic of discussion in the culinary world, raising concerns about the ethics of copying and the limitations of artificial intelligence in cooking. In an experiment conducted by The Verge, cakes were baked using recipes generated by ChatGPT and Bard, two powerful language models. Surprisingly, these AI-generated recipes closely resembled those found on popular food blogs or Instagram, with minor alterations. However, only Bard (now known as Gemini) provided attribution by linking to a recipe from Sally's Baking Addiction.

Recipe copying is not a new issue in the food industry, as non-AI recipes have been widely circulated on the internet without proper attribution. The involvement of AI intensifies this problem, potentially overshadowing human recipe developers with semi-randomized AI-generated competition. While we have seen glimpses of this scenario, it has not yet fully materialized.

The food world has long grappled with recipe similarities, leading to controversies when influencers, bloggers, or cookbook authors are accused of stealing recipes. Legal protection for recipes is limited, as a mere list of ingredients and instructions is considered an unprotectable idea. Many recipes also have cultural and oral traditions passed down through generations, further complicating the issue.

AI-powered language models like ChatGPT and Gemini possess the ability to analyze numerous permutations of ingredients and quickly generate recipes. This capability has led to suggestions that chatbots could be valuable tools for finding recipes that align with specific dietary needs. However, it's important to note that AI lacks the ability to taste or physically prepare food. While it can generate recipes that fit the pattern of successful ones, it cannot guarantee their actual taste or texture.

According to Abi Balingit, a cookbook author and recipe developer, AI's impact on recipe creation is not a significant concern. However, she worries about the potential consequences for food writers and developers, as media outlets may turn to AI-generated content instead of relying on human expertise and cultural connections. The human element in recipe creation establishes trust and adds personal experiences to the culinary journey.

AI-generated recipes are not without their flaws. Instances where AI models hallucinate and produce physically impossible recipes have been documented. For example, Instacart partnered with OpenAI's ChatGPT for recipe images, resulting in peculiar combinations such as hot dogs with the interior of a tomato or a salmon Caesar salad that mysteriously incorporated a lemon-lettuce hybrid. Proportions in these recipes were often inaccurate, leading to unintended serving sizes.

Ultimately, the value of recipe creators lies in their personal connections and expertise. Cookbooks like Mayumu by Abi Balingit offer a glimpse into cultural heritage and personal stories associated with food, fostering a deeper connection with the audience. AI-generated recipes, on the other hand, lack this emotional and cultural context, often resulting in generic and uninspiring dishes.

In conclusion, AI-generated recipes have their place in assisting with recipe discovery, but they cannot replace the creativity, cultural connections, and personal experiences that human recipe developers bring to the table. The ethical concerns surrounding recipe copying persist in the digital age, with AI's involvement adding a new layer of complexity. As technology continues to advance, striking a balance between AI assistance and human ingenuity will be crucial in preserving the art and authenticity of culinary creation.