Rancid flavors and unpleasant odors in food items like nuts, chocolates, and oils can now be more precisely addressed with the help of artificial intelligence (AI) tools. A group of chemists specializing in food preservation has unveiled a study highlighting the benefits of employing AI to extend the shelf life of food products, particularly those prone to rancidity. This innovative approach has broader implications for various sectors, including cosmetics and pharmaceuticals.
The Culprit: Rancidity and Oxidation
Rancidity is a result of the oxidation process that occurs when food is exposed to the air over time. Lipids, which encompass fats and oils, are particularly susceptible to this phenomenon. Factors like heat and UV light can expedite the oxidation process, leading to the formation of undesirable byproducts such as ketones, aldehydes, and fatty acids. These byproducts are responsible for the characteristic, off-putting scent and taste associated with rancid food. Consistently consuming rancid foods can pose health risks.
Fortunately, the food industry employs a potent defense against rancidity in the form of antioxidants. These antioxidants encompass a range of natural and synthetic molecules, such as vitamin C, that safeguard food products from oxidation. While there are several mechanisms through which antioxidants operate, they are primarily responsible for neutralizing the processes leading to rancidity, thereby preserving the taste and nutritional quality of food. Often, consumers are unaware that antioxidants are added to food during the manufacturing process, typically in small amounts.
The Challenge of Antioxidant Selection
Selecting the right combination of antioxidants and determining their precise quantities is a complex endeavor. The efficacy of antioxidants can be influenced by various factors, including their specific types and proportions in a food product. Suboptimal choices and ratios may result in antagonism, reducing the protective effect of antioxidants. This necessitates extensive experimentation, consuming time, human resources, and increasing overall food production costs.
AI as a Solution
Artificial intelligence, a technology that has become increasingly prevalent in various domains, now offers a solution to this intricate problem. AI tools, like the popular ChatGPT, are capable of processing vast datasets to identify patterns and generate valuable insights.
In the context of food preservation, researchers sought to harness AI’s potential to discover novel antioxidant combinations. An AI system that operates with textual representations of chemical structures was chosen for this purpose. Initially, the AI was educated on fundamental chemical concepts, enabling it to recognize crucial molecular features and interactions. Subsequently, the AI was fine-tuned by integrating more advanced chemistry knowledge using a comprehensive database of approximately 1,100 antioxidant mixtures documented in existing research.
Upon completion of this training, the AI exhibited the ability to predict the outcomes of combining two or three antioxidants in less than a second, aligning with the literature’s descriptions in 90% of cases. However, the transition from AI predictions to real-world experiments proved to be more challenging.
Continual Improvement
AI models are dynamic learners, and their capabilities can be honed over time with the introduction of new data. The research team found that adding approximately 200 laboratory test examples enabled the AI to predict the outcomes of experiments performed with genuine lard, resulting in only minor discrepancies between predicted and actual results.
This AI model has the potential to serve as a valuable tool for scientists seeking the best antioxidant combinations for specific food products. The project is ongoing, with researchers exploring ways to enhance the AI model’s training process and further refine its predictive capabilities. This innovative approach could revolutionize food preservation methods and contribute to improved product quality and shelf life.