Machine Learning Models Tackle Global Aid Data and Industrial Recycling
Researchers are applying machine learning to unexpected areas, from analyzing two decades of international development aid to sorting millions of discarded machine parts for reuse.
Machine learning continues to expand into unexpected industries by taking on massive datasets that overwhelm human researchers. A team from ETH Zurich and LMU Munich applies AI to international development aid, training a model on 3.2 million projects worth $2.8 trillion from the last 20 years. The algorithm analyzes 200 different dimensions to sort these projects, revealing surprising trends like increased spending on inclusion alongside a recent decrease in climate funding.
Another niche application comes from German R&D organization Fraunhofer, which develops a machine learning system to sort through the enormous volume of machine parts produced by various industries. Instead of sending countless components to the scrap yard, this AI model identifies parts so they can be recycled or reused appropriately. The system processes more than just standard camera images to make accurate identifications.
To account for visual discrepancies like rust or wear, the Fraunhofer system weighs each part and scans it with 3D cameras while pulling in metadata like origin. This multi-sensory approach allows the model to suggest what a part is, saving human inspectors from starting their evaluation from scratch. The technology promises to save tens of thousands of parts from the scrap yard while dramatically accelerating the processing of millions more.