Analysing Environmental Efficiency in AI for X-Ray Diagnosis

Research comparing AI models for COVID-19 X-ray diagnosis found that smaller discriminative models like Covid-Net achieve 95.5% accuracy with 99.9% lower carbon footprint than large language models. The study reveals that while LLMs like GPT-4 are versatile, they create disproportionate environmental impact for classification tasks compared to specialized smaller models.