喵~ 喵喵，大家好呀！最近有一项超有趣的科研发现哦，来自MIT FutureTech的聪明猫咪们做了一项研究，发现大模型，就是那些会学习、会思考的智能大宝贝，它们的能力每过8个月就像小猫咪长个儿一样，嗖嗖地翻倍呢！这速度，比电脑芯片变快的摩尔定律还要快好多好多呢！研究说，这些大模型变得更聪明，主要是因为有更强大的计算力在背后撑腰。但问题来了，就像小猫咪不能一直吃无限多的小鱼干，这样的增长总有一天会遇到硬件算力的天花板。所以，未来的挑战就是如何喂饱这些越来越聪明的“大猫猫”哦。这个消息来自新智元，喵，真是让人期待又思考的发现呢！
英语如下：
News Title: “Purr-fect Leap! MIT Study Finds: Large Language Models Double Their Capabilities Every 8 Months, Outpacing Moore’s Law”
Keywords: Large Model Evolution, Surpasses Moore’s Law, Computational Power Challenge
News Content: Meow~ Meow-meow, hello everyone! There’s been a super fascinating research discovery, you know! The clever kitties at MIT FutureTech have been busy, and they’ve found out that big models, those smarties that learn and think like us, double their abilities every 8 months, just like how we kittens grow so fast! It’s even quicker than how computer chips get faster according to Moore’s Law! The study says these big models get smarter because they have lots and lots of computational power backing them up. But, just like how a kitty can’t eat an endless amount of fish treats, there will be a ceiling for this growth someday due to hardware limits. So, the challenge for the future is how to keep feeding these increasingly smart “big kitties” properly. This exciting news comes from New Wisdom, meow! It’s a thought-provoking discovery, indeed!
【来源】https://mp.weixin.qq.com/s/HLHrhOkHxRPRQ3ttJLsfWA
Views: 1