如果你对我挑选的话题感兴趣，欢迎订阅 Dailyio 会员服务，作为 I/O 会员计划的 2.0 产品，Dailyio 将继续通过付费的门槛过滤、筛选读者群体，为优质读者提供全球视野的科技商业洞察。
新闻
Waymo 推出自动驾驶出行服务。这个名叫「Waymo One」的出行服务只局限在亚利桑那凤凰城的特定区域，出于安全方面的考虑，Waymo One 的自动驾驶汽车会配备司机，司机只会在紧急状况下对车辆进行干预。
中国 62% 的机器学习人才前往美国。机器学习公司 Diffbot 的一份报告显示，中国四所顶尖大学的 12521 名机器学习人才，只有 31% 留在国内，62% 的人才则前往美国。
Google 投资了一家日本 AI 创业公司。该公司名叫 ABEJA，专注于通过机器学习来打造一个供企业使用的通用平台。具体投资金融并未透露。
高通骁龙 855 更多细节披露，AI 性能提升。根据高通的说法，他们正在和 Google 一起合作，计划将机器学习框架 TensorFlow 迁移到手机上，以此来进一步提升手机的 AI 性能。不过值得注意的，此次骁龙 855 依然没有独立的 NPU。
观点
Facebook：数据中心的深度学习需求将持续升温。FB 的研究者发布了一篇论文指出，数据中心的 AI 硬件需求正在变得越来越大，由此带来的是服务器的设计工作将发生相应的变化，同时，数据中心的工作负载也会有新特性。
故事
250 页的 FB 内部文件里，有 6 个值得关注的点。
这六个关键点来自 MIT TR，大家可以看一下：
—Facebook “whitelisted” certain companies, meaning that they still had full access to users’ friends’ data after platform changes in 2014-15, including Airbnb, Netflix, and Badoo. “It is not clear that there was any user consent for this, nor how Facebook decided which companies should be whitelisted or not,” said Damian Collins, a member of Parliament and committee chair.
—Facebook aggressively tried to shut down the competition. When Twitter launched the six-second-video-clip platform Vine, Zuckerberg approved revoking its access to Facebook’s API.
—Friends’ data has been a big source of revenue for Facebook, thanks to growing revenues from app developers. The idea of tying access to this data to the developer’s relationship with Facebook is a recurring feature throughout the documents.
—Mark Zuckerberg wanted “full reciprocity” between Facebook and app developers: you share all your data on users with us, and we’ll share all of ours with you.
—Facebook found ways to access users’ call history without alerting them, in order to make “People You May Know” suggestions and tweak news-feed rankings. Facebook planned to make it as hard as possible for users to know that this was happening.
—Facebook used Onavo (an Israeli analytics company it bought in 2013) to check customers’ usage of mobile apps without their knowledge. They used this to find out how many people had downloaded apps and how often they used them. This information was used to suss out potential companies to acquire.
