Have you ever bitten into a nut or chocolate, expecting a smooth, rich taste, only to be met with an unexpected, unpleasant, chalky or sour taste? It affects almost every product there is.now artificial intelligence It will help scientists tackle this problem more accurately and efficiently.
We are a group of chemists who research ways to extend the lifespan of food, including spoiled food.we recently published research Learn about the benefits of AI tools that help keep oil and fat samples fresh for longer periods of time. Because fats and oils are common ingredients in many types of foods, including potato chips, chocolate, and nuts, the research results could be broadly applicable and have implications for other fields such as cosmetics and medicine.
Rot and antioxidants
food spoils When exposed to air for a period of time, a process called oxidation occurs. In fact, there are many common ingredients, especially lipidsIt is a fat that reacts with oxygen. The presence of heat and ultraviolet light can accelerate this process.
Oxidation results in the formation of small molecules such as: ketones, aldehyde and fatty acid It gives spoiled food a unique rank, a strong metallic aroma.repeatedly ingesting spoiled food may threaten your health.
Fortunately, both nature and the food industry provide excellent antioxidants that prevent spoilage.
Antioxidant It contains a wide range of natural molecules, such as vitamin C, as well as synthetic molecules that can protect food from oxidation.
while it’s still there Some ways antioxidants work, overall, they can neutralize many of the processes that cause spoilage and preserve the flavor and nutritional value of food for longer. Food manufacturers typically add antioxidants in small amounts during cooking, so customers often don’t even realize they’re consuming additional antioxidants.
However, you cannot expect a preservative effect just by sprinkling vitamin C on food.Researchers must choose carefully A set of specific antioxidants and accurately calculate the amount of each.
Combining antioxidants does not necessarily enhance their effects. In fact, using the wrong antioxidants or mixing them in the wrong proportions may reduce their protective effects. antagonistic. Finding which combinations work for which types of food requires a lot of experimentation, is time consuming, requires specialized personnel, and increases the overall cost of the food.
Exploring all possible combinations would require an enormous amount of time and resources, so researchers are stuck with a few mixtures that offer only some level of protection against spoilage. This is where AI comes into play.
Applications of AI
You’ve probably seen AI tools like ChatGPT in the news or tried them yourself. This type of system can ingest large amounts of data, identify patterns, and produce output that may be useful to users.
As chemists, we wanted to teach AI tools how to find new combinations of antioxidants. To this end, we have selected the types of AI we can work with. text expression, is a written code that describes the chemical structure of each antioxidant. First, they fed the AI a list of about 1 million chemical reactions and taught the program some simple chemistry concepts, such as how to identify key features of molecules.
Once the machines could recognize common chemical patterns, such as how certain molecules react with each other, they taught them more advanced chemistry to fine-tune the machines. For this step, our team used a database of approximately 1,100 mixtures previously described in the research literature.
At this point, the AI could predict the effect of combining two or three antioxidants within a second. The predictions matched the effects described in the literature 90% of the time.
However, these predictions did not completely match the experiments our team conducted in the lab. In fact, we found that the AI was able to accurately predict only a small number of oxidation experiments performed using real lard. This illustrates the complexity of transferring results from the computer to the laboratory.
Refinement and enhancement
Fortunately, AI models are not static tools with predefined “yes” and “no” paths. Because these are dynamic learners, our research team feeds the model new data until its predictive ability is strengthened and can accurately predict the effects of each antioxidant combination. You can continue. Just as humans grow through learning, models become more accurate as they acquire more data.
By adding about 200 examples from the lab, the AI will be able to learn enough chemistry to predict the outcome of the experiment the team conducted, and the difference between the predicted value and the actual value will be small. I understand.
Models like ours could help scientists develop better ways to preserve food by coming up with the best combination of antioxidants for the specific foods they study. . It’s like having a super smart assistant.
The project is currently exploring more effective ways to train AI models and further improve their predictive capabilities.