Researcher's 'Weightless' Neural Networks Skip Multiplication to Cut AI's Energy Use

UT Austin's Lizy John says lookup-table-based 'weightless' neural networks can be far smaller or faster than conventional models with comparable accuracy on some tasks.