A Survey of AutoML on Consumer Electronics in the Generative AI Era: Foundations, State of the Art, and Future Directions

Abstract

In the Era of Generative AI (GenAI), Automated Machine Learning (AutoML) has emerged as a transformative approach for building predictive models with reduced need for expert intervention, with the aid of recent technologies such as Large Language Models (LLMs). In parallel, with the surge in demand for Pervasive AI, the concept of Tiny Machine Learning (TinyML) with the promise of low latency in resource-constrained environments has emerged. However, existing AutoML systems still face limitations, including high computational costs and reliance on human oversight, which hinders deployment in consumer-facing environments requiring efficiency and rapid integration. This paper examines the intersection of AutoML, LLMs and TinyML on consumer electronics by presenting the current state of the art and exploring the potential of GenAI-enhanced AutoML frameworks. A qualitative evaluation of fifteen AutoML consumer-facing platforms is conducted on usability, scalability, transparency, and integration. Four representative tools are further selected for quantitative analysis across four ML problems: tabular classification, tabular regression, image classification, and text classification. The findings highlight strengths and limitations of current platforms, offering insights for user-accessible AutoML solutions.