TurboQuant: Breakthrough in AI Memory Efficiency
Google developed TurboQuant, a method using learned quantization to compress conversational AI data to 3.5-bit precision without accuracy loss. This reduces hardware costs and energy use, enabling more efficient AI.