Frank (@realfengliu) 's Twitter Profile
Frank

@realfengliu

Hardware, Data Storage, Artificial Intelligence, EV, Investment, History, Politics, Golfing, Running and Educators

ID: 1457216307007426567

calendar_today07-11-2021 05:20:46

4,4K Tweet

200 Takipçi

1,1K Takip Edilen

Tech with Mak (@technmak) 's Twitter Profile Photo

LLM = Smart, but works only with what it was trained on. Its knowledge is fixed after training, and when it lacks relevant information, it may generate plausible-sounding but incorrect answers (“hallucinations”). RAG = Smart + can “look things up” in real time. It retrieves

LLM = Smart, but works only with what it was trained on. Its knowledge is fixed after training, and when it lacks relevant information, it may generate plausible-sounding but incorrect answers (“hallucinations”).

RAG = Smart + can “look things up” in real time. It retrieves
Li Xiangyu (@lixiang1947) 's Twitter Profile Photo

有个朋友是奇绩校友,他从美国回国创业的时候,我们就认识了。当年他刚回来,非常意气风发,我们总是彻夜聊天到天明。

Amelia Maya (@ameliamaya98) 's Twitter Profile Photo

🚨 BREAKING: President Trump slaps 100% tariffs on imported chips. 🇹🇼 Taiwan moves into crisis mode — TSMC execs in urgent meetings, officials racing for answers. Why this is huge — and why the move is having an impact: 🧵 A thread

🚨 BREAKING: President Trump slaps 100% tariffs on imported chips.

🇹🇼 Taiwan moves into crisis mode — TSMC execs in urgent meetings, officials racing for answers.

Why this is huge — and why the move is having an impact:

🧵 A thread
Gabriele Berton (@gabriberton) 's Twitter Profile Photo

If LLMs are indeed approaching a plateau, big spendings in LLM is not justified anymore Open source LLMs will be just as good as closed source (and they already are for certain tasks) The valuation of LLM companies (and salaries of LLM scientists) will be significantly reduced

Kirk Borne (@kirkdborne) 's Twitter Profile Photo

Deep #ReinforcementLearning Hands-On — Practical easy-to-follow guide to RL from Q-learning and DQNs to PPO and RLHF: amzn.to/3MV9o60 [3rd Edition] v/ Packt Data Science & Machine Learning —— #AI #MachineLearning #DeepLearning #DataScience #DataScientist —— 𝓚𝓮𝔂 𝓕𝓮𝓪𝓽𝓾𝓻𝓮𝓼: 🟢Learn with

Deep #ReinforcementLearning Hands-On — Practical easy-to-follow guide to RL from Q-learning and DQNs to PPO and RLHF: amzn.to/3MV9o60 [3rd Edition] v/ <a href="/PacktDataML/">Packt Data Science & Machine Learning</a> 
——
#AI #MachineLearning #DeepLearning #DataScience #DataScientist
——
𝓚𝓮𝔂 𝓕𝓮𝓪𝓽𝓾𝓻𝓮𝓼:
🟢Learn with
The Prohuman (@theprohumanai) 's Twitter Profile Photo

🚨BREAKING: AI can now build you a business in 24 hours. Here are 10 insane Grok 4 prompts to turn any idea into income — starting today. 👇

🚨BREAKING: AI can now build you a business in 24 hours.

Here are 10 insane Grok 4 prompts to turn any idea into income — starting today. 👇
🥱 Sleepy (ML/DL) (@krishnanarakun) 's Twitter Profile Photo

Deep Learning with Pytorch : From Zero to GNN Deep Learning tutorials, taught by Prof Aritra Sen, teaching #DeepLearning using #Pytorch . Link in comments 😎👇

Deep Learning with Pytorch : From Zero to GNN

Deep Learning tutorials, taught by Prof Aritra Sen, teaching #DeepLearning using #Pytorch .

Link in comments 😎👇
看不懂的sol (@dtdt666) 's Twitter Profile Photo

不得不承认,川普真是一个金融交易“天才”,他搞出了稳定币这个东西。 首先照顾一下不了解事件的朋友,解释一下到底发生了什么事情。 兄弟们可以看看我最新统计的数据(图),美国的债务规模已经突破 37 万亿美元,这个数字是个什么概念? 我就这样说吧, 哪怕美国现在原地解散美军,

不得不承认,川普真是一个金融交易“天才”,他搞出了稳定币这个东西。
首先照顾一下不了解事件的朋友,解释一下到底发生了什么事情。
兄弟们可以看看我最新统计的数据(图),美国的债务规模已经突破 37 万亿美元,这个数字是个什么概念?

我就这样说吧,
哪怕美国现在原地解散美军,
看不懂的sol (@dtdt666) 's Twitter Profile Photo

醍醐灌顶!一位A8交易员的20条观点整理,字字珠玑! 观点一:假如我错了,我得赶紧脱身,有道是留得青山在,不怕没柴烧。我必须保存实力,卷土重来。 观点二:有的交易者之所以会不断亏钱,最主要原因是耐性不够,以致忽略交易原则,无法等到大势明朗或自己可以把控的行情,就贸然进场交易。

醍醐灌顶!一位A8交易员的20条观点整理,字字珠玑!
观点一:假如我错了,我得赶紧脱身,有道是留得青山在,不怕没柴烧。我必须保存实力,卷土重来。

观点二:有的交易者之所以会不断亏钱,最主要原因是耐性不够,以致忽略交易原则,无法等到大势明朗或自己可以把控的行情,就贸然进场交易。
0xKaKa (@0xkaka03) 's Twitter Profile Photo

(1/2)再次分享一门宝藏课程,加州伯克利(世界上最好的大学之一)公开的硬核Defi入门课程,帮助你从入门到精通。而且这门课所有的课件,课程录像,阅读材料(各种经典paper)都是齐全的。课程在小白友好的同时附带了几十篇Defi领域的经典paper作为参考材料,话题涵盖了传统金融基础,智能合约,Dex,预

(1/2)再次分享一门宝藏课程,加州伯克利(世界上最好的大学之一)公开的硬核Defi入门课程,帮助你从入门到精通。而且这门课所有的课件,课程录像,阅读材料(各种经典paper)都是齐全的。课程在小白友好的同时附带了几十篇Defi领域的经典paper作为参考材料,话题涵盖了传统金融基础,智能合约,Dex,预
Ken Berenger (@kenberenger) 's Twitter Profile Photo

Cathie Wood just made the boldest prediction of her career on DOAC: “These 5 industries will grow by 5000% and create the largest wealth transfer in human history.” Trillions will go to a tiny group of people. Here’s who gets rich—and how you can capitalize: 🧵

Cathie Wood just made the boldest prediction of her career on DOAC:

“These 5 industries will grow by 5000% and create the largest wealth transfer in human history.”

Trillions will go to a tiny group of people.

Here’s who gets rich—and how you can capitalize: 🧵
Bobby Freiler (@bobbyfreiler) 's Twitter Profile Photo

The Bible names 8 foods that contain "God's medicine." I’ve eaten these, grown these, and raised my boys on these. Here's the full list of elite, Biblical foods: 1. Watermelon

The Bible names 8 foods that contain "God's medicine."

I’ve eaten these, grown these, and raised my boys on these.

Here's the full list of elite, Biblical foods:

1. Watermelon
墓碑科技 (@mubei7777777) 's Twitter Profile Photo

1963年 一个叫约瑟夫·墨菲的人, 发现了宇宙最大的秘密。 他1898年生于爱尔兰,从小质疑信仰。 年轻时,贫穷,迷茫。 他只问一个问题: 是什么让一些人兴旺,而另一些人痛苦? 这个问题,定义了他一生的使命。 墨菲去了美国。 他研究宗教、哲学、心理学。 他向精神领袖和科学家学习。

墓碑科技 (@mubei7777777) 's Twitter Profile Photo

吴恩达与AMD苏大姐的对话: 学编程的人会变少吗? 可能正好相反。 AI正在让写代码变简单, 有人说那就别学编程了, 因为AI会替你写。 但历史告诉我们:这可能是个极其糟糕的建议。 还记得从打孔卡到键盘的过渡吗? 编程门槛降低了,结果是,学的人更多了,而不是更少。再之后从汇编语言到COBOL,

墓碑科技 (@mubei7777777) 's Twitter Profile Photo

专为机器人大脑而生 NVIDIA英伟达刚刚发布Jetson Thor, 机器人界的“芯中之王”。 它比上一代Orin快20倍,还省电,每瓦性能提升3.5倍。 Amazon Robotics、Meta、Caterpillar和Figure率先上手,已经开始部署。 Jetson Thor其实是全套机器人“大脑”。 摄像头、雷达、激光雷达获取的数据,

墓碑科技 (@mubei7777777) 's Twitter Profile Photo

英伟达NVIDIA黄仁勋: 我们在没有中国市场的情况下, 创下了创纪录的一个季度。 我们预计下一个季度也将再创纪录, 同样不包含中国市场 AI计算,正在离开中共国。 不靠中国,英伟达依然季度破纪录, 而且还连续两季新高。 它背后的动力, 已经不是传统数据中心负载, 而是一种全新的计算范式。