Auto Byte



Science AI




Synced Global AI Weekly | 2018.10.6—10.12

Watch This Incredible Parkour Robot!Watch This Boston Dynamics Robot Perform Incredible Parkour

Robotics company Boston Dynamics today posted a YouTube video of its humanoid robot Atlas doing parkour. There’s a wow moment when Atlas smoothly jumps over a log and leaps up a series of 40cm steps without breaking pace.


Pepper The Creepy Robot Will Appear Before UK Parliament As A Witness Next Week

The Commons Education Selection Committee will host Pepper among other witnesses to be questioned by members of parliament in an effort to better understand robots and what role they will play in humanity’s future, including their replacement of McDonald’s workers.


Amazon Developing ‘Picking’ Robots for Warehouses

When Amazon announced plans last week to bump the minimum wage for 250,000 employees to $15 an hour, it avoided a related, but sensitive topic: When might machines replace those workers, many of whom work in its warehouses?

(The Information)

Using Deep Reinforcement Learning to Control Dexterous Robot Hands

Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations proposes a low-cost and high-efficiency control method that uses demonstration and simulation techniques to accelerate the learning process.



Learning Acrobatics by Watching YouTube

In this work, we present a framework for learning skills from videos (SFV). By combining state-of-the-art techniques in computer vision and reinforcement learning, our system enables simulated characters to learn a diverse repertoire of skills from video clips.

(Berkeley AI Research)

Open Sourcing Active Question Reformulation with Reinforcement Learning

"we are releasing a TensorFlow package for Active Question Answering (ActiveQA), a research project that investigates using reinforcement learning to train artificial agents for question answering. Introduced for the first time in our ICLR 2018 paper “Ask the Right Questions: Active Question Reformulation with Reinforcement Learning”, ActiveQA interacts with QA systems using natural language with the goal of providing better answers."

(Google AI)

You May Also Like

Huawei Leaps Into AI; Announces Powerful Chips and ML Framework

Chinese tech giant Huawei is now aggressively expanding its artificial intelligence footprint. The company made a series of AI-related announcements today at the Huawei Connect 2018 Conference in Shanghai, introducing two AI chips and a machine learning framework.


New Research Examines The Current State of Deep Learning on Android platforms

Researchers studied acceleration resources on four main mobile chipset platforms: Qualcomm, HiSilicon, MediaTek and Samsung, while also comparing real-world performance results of various SoCs. AI Benchmark collected results covering all main existing hardware configurations.


Global AI Events

15–17 OctMinds Mastering Machines [m3]London, UK
15–17 OctThe Conference on Predictive APIs & AppsBoston, USA
17–18 OctPredictive Analytics WorldLondon, UK
17–18 OctNordic Data Science and Machine Learning SummitStockholm,Sweden
21–24 OctAI Deep Dive at Money 2020Las Vegas, USA
22–26 OctCKIMTurin, Italy



TensorFlow是一个开源软件库,用于各种感知和语言理解任务的机器学习。目前被50个团队用于研究和生产许多Google商业产品,如语音识别、Gmail、Google 相册和搜索,其中许多产品曾使用过其前任软件DistBelief。


张量是一个可用来表示在一些矢量、标量和其他张量之间的线性关系的多线性函数,这些线性关系的基本例子有内积、外积、线性映射以及笛卡儿积。其坐标在 维空间内,有 个分量的一种量,其中每个分量都是坐标的函数,而在坐标变换时,这些分量也依照某些规则作线性变换。称为该张量的秩或阶(与矩阵的秩和阶均无关系)。 在数学里,张量是一种几何实体,或者说广义上的“数量”。张量概念包括标量、矢量和线性算子。张量可以用坐标系统来表达,记作标量的数组,但它是定义为“不依赖于参照系的选择的”。张量在物理和工程学中很重要。例如在扩散张量成像中,表达器官对于水的在各个方向的微分透性的张量可以用来产生大脑的扫描图。工程上最重要的例子可能就是应力张量和应变张量了,它们都是二阶张量,对于一般线性材料他们之间的关系由一个四阶弹性张量来决定。


机器人学(Robotics)研究的是「机器人的设计、制造、运作和应用,以及控制它们的计算机系统、传感反馈和信息处理」 [25] 。 机器人可以分成两大类:固定机器人和移动机器人。固定机器人通常被用于工业生产(比如用于装配线)。常见的移动机器人应用有货运机器人、空中机器人和自动载具。机器人需要不同部件和系统的协作才能实现最优的作业。其中在硬件上包含传感器、反应器和控制器;另外还有能够实现感知能力的软件,比如定位、地图测绘和目标识别。之前章节中提及的技术都可以在机器人上得到应用和集成,这也是人工智能领域最早的终极目标之一。


问答系统是未来自然语言处理的明日之星。问答系统外部的行为上来看,其与目前主流资讯检索技术有两点不同:首先是查询方式为完整而口语化的问句,再来则是其回传的为高精准度网页结果或明确的答案字串。以Ask Jeeves为例,使用者不需要思考该使用什么样的问法才能够得到理想的答案,只需要用口语化的方式直接提问如“请问谁是美国总统?”即可。而系统在了解使用者问句后,会非常清楚地回答“奥巴马是美国总统”。面对这种系统,使用者不需要费心去一一检视搜索引擎回传的网页,对于资讯检索的效率与资讯的普及都有很大帮助。从系统内部来看,问答系统使用了大量有别于传统资讯检索系统自然语言处理技术,如自然语言剖析(Natural Language Parsing)、问题分类(Question Classification)、专名辨识(Named Entity Recognition)等等。少数系统甚至会使用复杂的逻辑推理机制,来区隔出需要推理机制才能够区隔出来的答案。在系统所使用的资料上,除了传统资讯检索会使用到的资料外(如字典),问答系统还会使用本体论等语义资料,或者利用网页来增加资料的丰富性。