AI’s impact on our world is undeniable, as foundational models like ChatGPT demonstrate. Yet, the public’s understanding of AI remains limited. People often think of ChatGPT or humanoid robots and fear job loss or dystopian scenarios. While these concerns are exaggerated, they reflect AI’s potential in the physical world, especially in industries like manufacturing, food, and transportation, which constitute 90% of global GDP. AI-enabled robotics will significantly affect the manual labor market, a $45 trillion industry.

The American food industry urgently needs AI-enabled robotics. With over a million unfilled jobs and a high turnover rate, many in the US are unwilling to take on repetitive, hazardous food preparation and production roles. This labor shortage pressures food manufacturers to keep their supply chains domestic amid high demand. Many consider offshoring parts of their supply chain, threatening the US food industry.

Chef aims to address this labor shortage by deploying AI-enabled robots in commercial kitchens across the US and eventually worldwide. Traditional automation in the food industry has been limited by its inflexibility. Robots have been designed to perform specific tasks, suitable for low-mix production lines but inadequate for high-mix environments requiring the assembly of diverse meals. Consequently, food companies have relied on millions of manual workers. In commercial kitchens, workers assemble meals, while in industrial kitchens, the process involves tedious, cold conveyor line work. Americans are increasingly reluctant to perform such labor-intensive tasks.

AI can transform traditional hardware, making it adaptable enough to provide labor through robotic automation. However, creating a functional food manipulation AI faces a cold start problem. Unlike cloud AIs or self-driving cars, the food industry lacks off-the-shelf training data and suitable simulation engines due to the deformable nature of food. Therefore, Chef must deploy robots in various production environments to gather diverse training data. This widespread deployment will enable robots to handle different foods and ultimately meet production demands effectively.