Are AI Models Working Harder Than They Need to?

Lizy K. John and her team at the University of Texas at Austin developed weightless neural networks that use lookup tables instead of multiplication, achieving similar accuracy with much smaller size and faster speed. Their work has focused on small applications but aims to scale to larger models like transformers, potentially reducing AI's energy consumption significantly.