AIup Regogni开发了Pareto方法,以降低电耗较低的AI芯片的体积、加快速度和降低其成本。 AI startup Recogni developed Pareto method to reduce size, increase speed, and lower costs of AI chips with lower power consumption.
AIup Recogni开发了一种新的计算方法Pareto, 可以减少AI芯片的体积、加快速度和降低成本。 AI startup Recogni has developed a novel computing method, Pareto, which could reduce the size, increase speed, and lower costs for AI chips. 专利对数法优于现有方法,将乘法转换为附加操作,从而降低电力消耗,同时保持准确性。 The patented logarithmic approach outperforms existing methods, converting multiplication into addition operations, thus lowering power consumption while maintaining accuracy. Recogni 已经在 Meta Platforms 和 Stability AI 的模型上进行了测试, 现在 Recogni 正与一家未透露姓名的公司合作, 扩大 Pareto 的可用性, 可能在数据中心提供硬件租用. Tested on models by Meta Platforms and Stability AI, Recogni is partnering with an unnamed company to expand Pareto's availability, possibly offering hardware in data centers for rent.