关键词：
title: You are what you write——沈向洋
date: 2018-02-21 13:18:28
tags: [随想,write]
categories: 个人文章
---
Are Twitter, PowerPoint, Facebook, Instagram and texting eroding our ability to think?
There is a Chinese proverb that says “见文如见人,” which literally means “reading the document is the same as seeing the author.” If we are what we write, then who have we, as a society, become?
I was sitting in a technical review recently, listening to one of our reviewers grill the engineer who was presenting: Why did you choose that design? Why is the service showing bad results? How many users will switch to the solution?
The presenter’s answers lacked depth. It seemed like he hadn’t done enough rigorous thinking, the kind where you sit quietly, sift through research, contemplate options, determine what you know, don’t know and where more work is required. The kind of thinking I did as a young researcher when peers took me and my work apart when I took short cuts. Back then, I practiced a disciplined approach, spending hours just thinking, and even more hours on the hardest part—writing it down.
Today, long-form writing is being replaced. Tweets pass for dialogue. PowerPoint condenses thoughts to bullets. Words have been traded for emojis and GIFs. And we’ve become addicted to the noise. What happens in an Internet minute? 16 million text messages. 1.8 million snaps. 452,000 Tweets. 156 million emails. Who has time to think, let alone write?
And maybe we, in the technology industry, have shaped this reality. We created the phones, apps and 24/7-connected world. We’ve enabled society to put down the pen. The only writing I do today is email or quick WeChat posts.
So now I worry that we’re losing a valuable tool that helps us to think deeply, express who we are at our greatest and expand the intellect of those around us. And for us in the technical community, this is especially troubling. The stakes are higher than ever before with AI. We’re under enormous pressure to ship quickly, to achieve more, faster, but we can’t do this at the expense of the highest engineering quality. We have to think carefully about consequences and alternatives. Who gets blamed when a self-driving car hits someone? The engineer who wrote the code is the driver. Who is accountable for the AI algorithm with bias? The engineer who created the AI.
I see fewer engineers writing and sharing deep thinking, but this is what will lead to far more true innovation across the industry. How will we achieve the big transformative breakthroughs versus the incremental milestones?
By writing. Because the way to think is actually to write.
Putting pen to paper forces you to develop and refine your thinking by iterating, revising and exploring alternatives. Anyone who can think deeply can write beautiful code, inspiring papers or develop the plan to bring the next big thing to life. I encourage you to read Reid Hoffman’s Series B pitch for LinkedIn in which he shares the thinking that helped him succeed. At the time, he shares that a partner in a venture firm was exposed to around 5,000 pitches, looked more closely at 600 to 800, and did between 0 and 2 deals.
Writing offers the possibility to create lasting artifacts. I think of papers I published that endure, albeit perhaps as reference materials. Plenoptic Sampling. Lazy Snapping. Poisson Matting. These are my work’s contribution to the field of computer vision and graphics. They will survive me and, if I’m lucky, even help shape a mind or two.
One of my favorite professors at Carnegie Mellon, Takeo Kanade, said that you have to write research papers like detective novels. You need story, suspense, surprise and ‘aha’ to explain your ideas to peers, to inspire others to contribute and advance your work and the whole field.
Writing is an equalizer to get the best from the whole team. At Amazon, presentations are done with the six-page paper. Meetings kick off with everyone reading followed by comments and questions to the author. Everyone operates from the same context, and introverts, extroverts and non-native speakers have an equal chance to get their thinking across. It’s not about the presenter’s personality, but the words.
Ultimately, writing helps make you successful. You might be the smartest person with the best idea, but if you can’t communicate your thinking in a compelling way, you won’t get far. Two engineers in our AI+R team who inspire me with their regular writing habits are Bill Ramsey and Ronny Kohavi. Bill has written over 250 blog posts at Microsoft, benefitting our entire technical community. With Ronny, you don’t even need to meet him—his highly cited A/B test experimentation papers say it all, and he’s publishing for the benefit of the industry on LinkedIn.
As you’re reading this, you may be logging your objections: I need to drive results, so I need to go straight to code. I’m known for my code, so I don’t need to write papers. I’m not a native speaker, and I speak better with my code. I don’t know what to write about. I don’t have time… But please set them aside—for your own success, for your company’s, for the industry’s advancement—and start writing.
I see so many occasions for building long-form writing back into the engineering culture—planning documents, project proposals, technology LRP’s, review articles—to inspire us to work together, collectively creating and cultivating big ideas and big thinking.
I took a first step recently, writing a research paper with my colleagues Xiaodong He and Di Li, From Eliza to XiaoIce: Challenges and Opportunities with Social Chatbots, for the first time in years, so please no judgment, only constructive feedback!
I challenge everyone reading this piece to write 500 words per week. If you’ve got an idea or you see a problem, write your proposal and share it!
Let’s rewrite our standards for thought leadership and engineering quality by writing more!
微软全球执行副总裁沈向洋：三年内用人工智能赚100亿美元
...一直在尝试回答一个最基本的问题：智能是什么。”沈向洋博士说。 这听上去不像是统治办公软件市场半壁江山的微软需要考虑的议题——并非如此。事实上，早在1991年比尔·盖茨就成立了微软研究院，以&ldquo... 查看详情
马毅沈向洋曹颖最新ai综述火了！耗时3月打造，网友：必读论文
...f01;耗时三个多月，联合神经科学家曹颖、计算机大牛沈向洋，协作完成。据本人描述，这篇论文是将他“过去五年的工作以及智能七十多年的发展有机结合起来”，并且还表示：一生从未在一篇文章上花这么... 查看详情
2022idea大会引领科技创新趋势沈向洋团队重磅发布低空经济白皮书
11月22日，2022IDEA大会在深圳顺利开幕。大会由深圳市科技创新委员会、深圳市人才工作局和深圳市福田区人民政府指导，粤港澳大湾区数字经济研究院（InternationalDigitalEconomyAcademy，简称“IDEA研究院”）主办。... 查看详情
18年后，张亚勤李开复王坚沈向洋洪小文等又干了件大事
650)this.width=650;"src="http://s2.51cto.com/wyfs02/M00/8A/49/wKiom1gsX26CdlitAAHg8UigDnk148.jpg-wh_500x0-wm_3-wmp_4-s_2513557738.jpg"title="五巨头1.jpg"alt="wKiom1gsX26CdlitAAHg8UigDnk148.jpg-wh_50"/>（上 查看详情
覆盖近2亿篇论文还免费！沈向洋旗下团队「读论文神器」登b站热搜
转自：新智元对于搞科研的同学们来说，看论文是要经历的第一关，尤其是要读好多篇论文的时候，更是着实令人头大。这不，最近无意中发现了个在线论文阅读网站：readpaper.com，号称「论文阅读笔记... 查看详情
旷视首席科学家孙剑博士去世，前辈一路走好！
...研究院和西安交通大学联合培养的研究生，他曾师从沈向洋博士。沈向洋博士是顶级计算机视觉和图形学专家，也是微软中国研究院（现微软亚洲研究院）联合创办人。2001年，孙剑和沈向洋完成了利用置信传... 查看详情
方法论：论文十问
...到的论文阅读方法论，感觉不错在此分享论文十问由沈向洋博士提出，鼓励大家带着这十个问题去阅读论文，用有用的信息构建认知模型。Q1论文试图解决什么问题？Q2这是否是一个新的问题？Q3这篇文章要验... 查看详情
创建工作生活新范式开拓经济增长新空间
...市人才研修院开幕。大会首日，IDEA研究院创院理事长沈向洋作为主讲人，发布《低 查看详情
微软亚研cv大牛王井东离职，或将加入百度
...视觉计算组作为微软亚研的灵魂团队，陆续培养出了沈向洋、汤晓鸥、马毅、孙剑、何恺明等老中青三代计算机视觉引领者，只可惜这些人都已远去。王井东原为微软亚洲研究院视觉计算组的首席研究员，研究领域包... 查看详情
突发！微软亚研cv大牛王井东离职，或将加入百度
...视觉计算组作为微软亚研的灵魂团队，陆续培养出了沈向洋、汤晓鸥、马毅、孙剑、何恺明等老中青三代计算机视觉引领者，只可惜这些人都已远去。王井东原为微软亚洲研究院视觉计算组的首席研究员 查看详情
谁是下一个丘成桐？产业界也开始关心这事儿了
...生里，能产生中国第一批菲尔兹奖得主。”丘成桐和沈向洋对谈20分钟，让清华“不用高考”的求真数院又火了一把。对，就是那个让丘成桐离开哈佛全职加入，专门培养“顶尖数学人才”的学院。在丘成桐看 查看详情
cncc2017梳理
...AN）的等价形式，通过解决等价问题加速优化过程沈向洋理解自然语言概述，对话和理解自然语言：机器学习（表述）->机器智能（对话）->机器意识（意境）图像表述：微软有一个Im... 查看详情
小爱同学与小冰将实现联合进步
...月举行的第六代微软小冰发布会上，微软全球执行副总裁沈向洋博士宣布了与小米的此项合作。此后，双方的合作及联合开发不断加深。针对本次上线，微软全球资深副总裁、微软亚洲互联网工程院院长王永东表示：“我们很高... 查看详情
2021idea大会开启ai思想盛宴，用“创业精神”做科研
...长、美国国家工程院外籍院士、英国皇家工程院外籍院士沈向洋博士作为主讲人，携IDE 查看详情
突发！微软亚研cv大牛王井东离职，或将加入百度
...视觉计算组作为微软亚研的灵魂团队，陆续培养出了沈向洋、汤晓鸥、马毅、孙剑、何恺明等老中青三代计算机视觉引领者，只可惜这些人都已远去。王井东原为微软亚洲研究院视觉计算组的首席研究员，研究领域包... 查看详情
九坤投资投身基础科研，携手idea成立联合实验室
...成立学术委员会。学委会成员包括AI大牛、IDEA创院理事长沈向洋。还有IDEA讲席科学家、IDEAAI安全普惠系统研究中心负责人王嘉平博士和九坤投资创始合伙人&CEO王琛博士。王嘉平&#x 查看详情