New
March 7, 2024

Can AI solve food waste?

Researchers have developed an initial-stage machine learning algorithm to predict food spoilage based on sensory observations and photos.

Can AI help tackle the global problem of food waste?

Overview:

A researcher at UPenn conducted a research study on shelf life and food spoilage and developed a machine learning algorithm to track food waste. This algorithm can predict if food was spoiled or not and how many days until spoilage based on sensory observations from photos.

Data & Research Behind the Algorithm:

The Shelf Life Expiration Date (SLED) algorithm takes in sensory characteristics, such as color, cracks, etc, from photos and aggregates them to develop correlations between the amount of time until a food spoiled and dates on food labels. Although it’s in its initial stages, this AI algorithm can accurately determine food spoilage predictions for food.

The algorithm can be applied through a test kit that takes into account the environment in which the food is stored. Factors such as temperature, humidity, and light conditions can greatly influence the rate of food spoilage and are thus incorporated into the algorithm's predictive model.

This algorithm, while innovative, still needs further refinement and an expanded dataset to improve its efficiency and accuracy. It also requires a user-friendly interface for easy integration and accessibility for consumers. As the algorithm continues to evolve, its potential impact on reducing food waste globally remains promising.

How does this help society and the planet?

This algorithm can greatly help reduce consumer food waste by clarifying shelf life and expiration dates. This in turn, prevents illness from expired foods,  decreases greenhouse gas emissions, and lengths availability of food resources.

Beyond predicting food spoilage at a consumer-level, another potential application of this AI is in the retail and food service industry. Supermarkets and restaurants could use this algorithm to better manage their food inventory, reducing overstock and minimizing waste. This not only leads to cost savings but also contributes to a more sustainable food system across the supply chain.

With the integration of the algorithm into shopping apps, consumers can make informed decisions about food purchases and consumption. Additionally, the algorithm's continuous learning and improvement through new data inputs mean its predictive accuracy and applicability will only increase with time.

How it can be biased?

This algorithm may encode bias regarding food consumption patterns, socioeconomic factors, or cultural preferences. By not having diverse data on cultural foods, this may lead to inaccurate predictions for specific demographics.

To mitigate these issues, it's crucial to include such unique and diverse food items in data collection efforts. Also, incorporating user feedback and real-time spoilage data could enhance the algorithm's understanding and prediction of spoilage patterns for these foods.

The developers should also consider partnerships with global food organizations, restaurants, and markets. This would allow for a broader, more diverse dataset that could help in enhancing the algorithm's accuracy and inclusivity.

It’s important to actively collect food and cooking data from a wide range of locations, demographics, and cultural backgrounds to ensure the algorithm learns from more representative patterns.

Report:  

https://arxiv.org/pdf/2309.02598.pdf