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.

• →Covid-Net achieved the highest accuracy of 95.5% while maintaining 99.9% lower carbon footprint than GPT-4.5-Preview
• →Smaller LLM GPT-4.1-Nano reduced carbon footprint by 94.2% compared to larger models but still had disproportionate environmental impact
• →Small discriminative models showed bias towards positive diagnosis and lacked confidence in output probabilities
• →Using LLMs purely for probabilistic output resulted in poor performance in both accuracy and environmental efficiency
• →The research demonstrates environmental risks of using generative AI tools for tasks better suited to specialized models