New algorithm promises to slash AI power consumption by 95 percent
SMRTR summary
Researchers at BitEnergy AI developed Linear-Complexity Multiplication (L-Mul), potentially reducing AI power consumption by up to 95% without major performance loss. L-Mul uses integer additions instead of floating-point multiplications, requiring less energy. Tests show maintained accuracy with significantly reduced energy use. The technique requires specialized hardware, currently in development with programming APIs. This innovation could address AI's growing energy demands, as some models like ChatGPT consume electricity equivalent to 18,000 US homes daily.
SMRTR provides this summary for quick context. The original article belongs to TechSpot.
Read the original article