Ana Perez-Neira

AI keeps getting bigger—but does bigger always mean better? As Large Language Models (LLMs) take center stage, their enormous resource demands raise pressing concerns about sustainability and efficiency. In this webinar, the presenter questioned the assumption that AI must be massive to be powerful. They introduced the Expressive Neural Network (ENN), a novel architecture that reimagines activation functions through the lens of classical signal processing—specifically, the Discrete Cosine Transform (DCT). This innovative design boosts a network’s flexibility and expressiveness, enabling faster convergence and much smaller models that cut both energy use and computational costs. Their discussion bridged traditional signal processing and modern AI, showing how established mathematical tools can inspire the next generation nonlinear signal processing. They explored how ENNs could transform edge computing by bringing efficient AI to resource-constrained environments, and why expressiveness—not mere size—holds the key to the future of neural networks. If LLMs represent the brute force of AI, ENNs may become a precision instrument. Looking ahead, they anticipate the rise of new model types—such as Small Language Models and specialized architectures—opening the door to more sustainable, purpose-driven AI deployment.