通用情报(General Intelligence)

OneZero’s General Intelligence is a roundup of the most important artificial intelligence and facial recognition news of the week.

OneZero的General Intelligence是本周最重要的人工智能和面部识别新闻的摘要。

Google Street View emerged from a seemingly insane vision: Put cameras on cars, and drive them around the entire world to capture every street on the planet.

摹oogle街景从一个看似疯狂的视力出现了：把相机上车，把他们赶了整个世界捕捉每一个街道的星球上。

Over time, that data became more and more valuable. The footage from those cars automatically updates Google Maps with new business signs and changes street names. Data from the car’s trip can be used to correct satellite imagery.

随着时间的流逝，这些数据变得越来越有价值。 这些汽车拍摄的镜头会自动使用新的商业标志更新Google Maps并更改街道名称。 来自汽车旅行的数据可用于校正卫星图像。

Now, Facebook has an even grander ambition. On Wednesday, the company announced a plan to map the entire world, beyond street view. The company is launching a set of glasses that contains cameras, microphones, and other sensors to build a constantly updating map of the world in an effort called Project Aria. That map will include the inside of buildings and homes and all the objects inside of them. It’s Google Street View, but for your entire life.

现在，Facebook的哈秒的更加宏伟的雄心。 周三，该公司宣布了一项计划，要绘制街道以外的整个世界。 该公司正在推出一套包含照相机，麦克风和其他传感器的眼镜，以建立一个不断更新的世界地图，这被称为Project Aria 。 该地图将包括建筑物和房屋的内部以及其中的所有对象。 这是Google街景视图，但终其一生。

Facebook can already see almost every aspect of how the world uses the internet, through its ad tracking of users and nonusers on and off its site. Project Aria would add a physical dimension to Facebook’s omnipotent gaze.

通过对网站上和下网站上的用户和非用户进行广告跟踪，Facebook已经几乎可以看到世界使用互联网的方方面面。 Aria项目将为Facebook的万能注视增加一个物理维度。

About a hundred Facebook employees and contractors will use the glasses to record as much as they can of their immediate surroundings, according to Facebook’s website.

据Facebook网站称，约有一百名Facebook员工和承包商将使用这副眼镜来记录尽可能多的周围环境。

This effort will give Facebook an enormous dataset to build future augmented reality and A.I. products. The data could be used for everything from building an AR application to telling you where you left your keys, to a more personalized virtual assistant, according to Facebook employees who presented the project.

这项工作将为Facebook提供庞大的数据集，以构建未来的增强现实和AI产品。 据介绍该项目的Facebook员工称，这些数据可用于从构建AR应用程序到告诉您密钥的存放位置，再到更具个性化的虚拟助手等所有功能。

If the project goes according to plan, the final result would be a live, 3D map of the world, constantly updated and refreshed by people walking around with AR headsets. Facebook says that maps of public places would be publicly viewable, while the maps of users’ homes and belongings would be private.

如果该项目按计划进行，最终结果将是实时的3D世界地图，并由使用AR头戴设备的人不断更新和刷新。 Facebook表示，公共场所的地图将是公开可见的，而用户房屋和所有物的地图将是私有的。

The company insists any faces and license plates captured by Aria glasses wearers will be anonymized. But that won’t protect the data from Facebook itself. Ostensibly, Facebook will possess a live map of your home, pictures of your loved ones, pictures of any sensitive documents or communications you might be looking at with the glasses on, passwords — literally your entire life. The employees and contractors who have agreed to wear the research glasses are already trusting the company with this data.

该公司坚称，Aria眼镜佩戴者捕获的任何面Kong和车牌都将被匿名化。 但这并不能保护Facebook本身的数据。 从表面上看，Facebook将拥有您的住所的实时地图，亲人的照片，戴眼镜可能正在看的任何敏感文件或通讯的图像，密码-实际上就是您的一生。 同意戴研究眼镜的员工和承包商已经将这些数据信任了公司。

The project brings up an extraordinary amount of questions because it doesn’t just affect the people who might buy Facebook’s AR glasses one day. The product and research decisions made by Facebook right now can determine whether Facebook is actually creating a distributed surveillance panopticon. Who has access to see the data from this live 3D map, and what, precisely, constitutes private versus public data? And who makes that determination? Faces might be blurred, but people can be easily identified without their faces. What happens if law enforcement wants to subpoena a day’s worth of Facebook’s LiveMap? Might Facebook ever build a feature to try to, say, automatically detect domestic violence, and if so, what would it do if it detected it?

该项目提出了很多问题，因为它不仅影响有一天可能会购买Facebook的AR眼镜的人们。 Facebook现在做出的产品和研究决策可以确定Facebook是否实际上在创建分布式监视泛黄酮。 谁有权访问此实时3D地图中的数据，以及究竟是私人数据还是公共数据？ 谁做出这个决定？ 面Kong可能会模糊，但没有面Kong也可以轻松识别人。 如果执法部门想要传唤一天一整天的Facebook LiveMap价值，会发生什么？ Facebook可能会建立一种功能来尝试自动检测家庭暴力，如果是这样，那么如果检测到它会怎么办？

It’s just the beginning for Facebook’s augmented reality plans, but these are questions the company has to consider before putting anything in the hands of consumers. Facebook has been pushing away users’ concerns about their privacy for more than a decade. That’s led to major data breaches and legal challenges for its invasive ad tracking.

这只是Facebook增强现实计划的开始，但这是公司在将任何东西交给消费者之前必须考虑的问题。 十多年来， Facebook一直在消除用户对其隐私的担忧。 入侵式广告跟踪导致重大数据泄露和法律挑战。

One thing is for certain: If the company handles this platform the same way it’s handled Facebook.com, we’re in for a lot of trouble.

可以肯定的是：如果该公司使用与处理Facebook.com相同的方式来处理该平台，那么我们会遇到很多麻烦。

And now, here’s some of the most interesting A.I. research of the week:

现在，这是本周最有趣的AI研究：

假货的心如何跳动？ (How do the hearts of deepfakes beat?)

This question, asked by researchers from Intel and Binghamton University, sounds like the start of a poem, but it sets up research that tries to weed out deepfakes by detecting biological markers in video. That could mean visual signs of breathing, or even a heartbeat. If the biological signals are atypical, that might mean the video was manipulated, and is not, in fact, footage of Joe Biden endorsing Antifa.

英特尔和宾汉姆顿大学的研究人员提出了这个问题，听起来像是一首诗的开头，但是它建立了研究，试图通过检测视频中的生物标记物来清除伪造品。 那可能意味着呼吸的视觉信号，甚至是心跳。 如果生物信号是非典型信号，则可能意味着该视频已被篡改，实际上并非乔·拜登(Joe Biden)认可安提法的镜头。

将大脑活动转化为单词 (Translating brain activity to words)

By analyzing an fMRI while the patient is reading a word, researchers from Zurich have been able to achieve a sliver of success at matching that person’s brain activity with a word. That’s a big step in the direction of brain interfaces like Elon Musk’s Neuralink being able to understand human thought. In an experiment with 180 possible words, the algorithm is able to identify the correct word being read just 5% of the time. But the team stresses that this method outperforms previous attempts.

通过在患者朗读单词时分析功能磁共振成像，苏黎世的研究人员已经能够成功地使该人的大脑活动与单词相匹配。 这是向大脑接口迈出的一大步，例如埃隆·马斯克(Elon Musk)的Neuralink能够理解人的思想。 在具有180个可能单词的实验中，该算法能够识别仅5％的时间被读取的正确单词。 但是研究小组强调说，这种方法优于以前的尝试。

了解抱怨的语言 (Understanding the language of grievance)

We’ve all texted while angry. But can an algorithm detect whether that anger will translate into violence? By gathering and validating the connotations of 20,000 words from 2,300 subjects, researchers at the University College of London released a dataset called the Grievance Dictionary. It correlates words with violent behavior, taken from manifestos, posts on the white nationalist website Stormfront, and abusive letters to politicians. Researchers hope that by understanding this connection, an algorithm might soon be able to identify online posts that convey violent intent, and maybe even prevent harm.

生气时我们都发了短信。 但是，算法可以检测出这种愤怒是否会转化为暴力吗？ 伦敦大学学院的研究人员通过收集和验证来自2,300个主题的20,000个单词的含义，发布了一个名为“抱怨字典”的数据集。 它使词语与暴力行为相关联，这些词语来自宣言，白人民族主义网站Stormfront上的帖子以及给政客的辱骂信。 研究人员希望，通过了解这种联系，一种算法可能很快就能识别出传达暴力意图甚至可以防止伤害的在线帖子。