Artificial Intelligence in Practice

by Bernard Marr

  The company has also partnered with China's social media giants, Tencent and Baidu, to integrate its products into their hugely popular messaging and image-sharing apps. Once again AI plays a key part in this. It is used to match users, based on their profile data, with items sold by JD.com that might appeal to them. These appear as sponsored advertisements, and users can order and pay for them without leaving their social apps.4

  Automated Deliveries By Air And Road

  In 2016, while Amazon was carrying out its first test deliveries using unmanned aerial vehicles, JD.com was putting its drone delivery network into full operation. Drone delivery has been a reality in China ever since, with JD.com's fleet having completed over 300,000 minutes of flight time so far.5

  JD.com is working on drones that can deliver cargo weighing up to 5 tons. Currently, the drone service is mainly used to deliver to areas that are in close proximity to drone stations – the furthest delivery has been around 15 km. But in the near future, particularly when batteries with better lifespans emerge, it is hoped it will hugely reduce the cost of carrying out deliveries to remote and difficult-to-reach regions of the country, which are hard to access by trucks. As well as making deliveries to customers, they will also move goods from warehouse to warehouse, which is usually done by trucks.

  Speaking of trucks, JD.com is of course automating those as well. Autonomous truck vehicles deployed by the company have accumulated 17,000 hours of road driving experience, and are already used in some circumstances to make deliveries. At the moment, although the vehicles can handle open road driving just fine, a human driver has to be present to take over when it enters a city. President of the company's X-Business division, Xiao Jun, has said: “There isn't much value if our technology can only cut three drivers down to two or even one. We hope the truck is unmanned.”6

  Facial Recognition Technology

  JD.com's interest in facial recognition technology revolves around its usefulness for customers proving their identification. It will be used to allow customers to make purchases simply by picking up items in its physical stores, as well as to verify their identification to get automated delivery agents to hand over their packages at their doorsteps.

  Customers start by signing up through their smartphones, and using their cameras to upload high-definition detailed images of their faces, which can then be identified, even if viewed from different angles, thanks to machine learning, which can predict how the face would appear from any viewpoint. Facial recognition is generally considered to be more secure than other biometric identification techniques such as fingerprinting, which in practice can be duplicated and forged relatively easily.

  Smart Fridges

  As is the case with its US competitors such as Amazon, JD.com is also branching out into developing and marketing AI-equipped “smart” consumer goods.

  In JD.com's case, it is specifically looking at refrigerators, and it has announced a “smart fridge” that uses cameras equipped with image recognition technology. The cameras scan items in the fridge and can even tell when they are starting to pass their expiry date. It keeps users informed via a smartphone app that can be used to place orders for items that are starting to run low.7

  With China's population growing more wealthy and interested in the benefits of healthy eating, devices like the smart fridge sold by JD.com, as well as rivals including Alibaba and Baidu, could also make health-based suggestions on what their owners should be eating more or less of, as well as suggest healthy recipes that can be put together from the items at hand.

  Smart Shops

  Unlike competitors such as Amazon and Alibaba, which started out on the internet, JD.com started out with a bricks-and-mortar store in Shanghai, China, before moving online in 2014. With an ongoing interest in offline retail, it has launched its first human-free shop at its Beijing headquarters. Here, humans can pay for their purchases with a simple glance at a camera, which registers their identity through facial recognition algorithms and deducts payment from their account. Another human-less store was opened in Jakarta, Indonesia, in 2018.8

  The technology is already being trialed in its network of grocery stores, which focus on selling fresh food items. It has announced plans to open 1,000 more of these 7Fresh branded stores to feed the country's growing appetite for healthy eating. AI is used at every level, from deciding where to put stores based on the demographics of residents, to maintaining stock levels and ensuring a steady supply of whatever happens to be in demand.9 Smart screens are also used in the stores, which are able to display bespoke advertising to customers based on their gender and age as determined by facial recognition technology.10

  Key Challenges, Learning Points And Takeaways

  JD.com's founder has said that he hopes to see his company's human staff reduced from 160,000 to 80,000 within the next 10 years.11 While he says many will be retrained, it seems retaining human jobs comes secondary to driving efficiencies and improving customer experience.

  Driving efficiency within its operations and supply chain is JD.com's primary motivation for rolling out AI. Automated warehouses, delivery networks and retail outlets all form a part of this plan.

  JD.com has also partnered with social media providers to allow them to use data on their customers for AI-driven precision marketing campaigns carried out entirely through their social apps.

  Starting out as a bricks-and-mortar retailer, JD.com is blurring the boundaries between online and offline shopping through a drive to introduce e-commerce technology into its physical stores.

  Notes

  1YouTube, Richard Liu, JD.com Founder, Chairman and CEO: https:// www.youtube.com/watch?v=VTSKy9E3tcU&feature=youtu.be

  2Axiom, In China, A Picture of How Warehouse Jobs Can Vanish: https:// www.axios.com/china-jd-warehouse-jobs-4-employees-shanghai-d19f5 cf1-f35b-4024-8783-2ba79a573405.html

  3JD.com, Preparing JD.com Orders for Same Day Delivery: https:// jdcorporateblog.com/gallery/preparing-jd-com-orders-day-delivery/

  4Digital Commerce 360, JD.com and Baidu Partner on Data‑Powered Precision Advertising and Marketing: https://www.digitalcommerce360 .com/2017/08/17/jd-com-baidu-partner-data-powered-precision- advertising-marketing/

  5Wired, Inside JD.com, the giant Chinese firm that could eat Amazon alive: https://www.wired.co.uk/article/china-jd-ecommerce-store- delivery-drones-amazon

  6South China Morning Post, JD.com unveils self-driving truck in move to automate logistics operations: https://www.scmp.com/tech/innovation/ article/2148420/jdcom-unveils-self-driving-truck-move-automate- logistics-operations

  7Medium, Why China's No.2 e-commerce site JD sees smart refrigerators as a key to IoTs boom: https://medium.com/act-news/why-chinas-no-2-e-commerce-site-jd-sees-smart-refrigerators-as-a-key-to-iots-boom-d9674a8c9a45

  8Retail Tech News, Weekly Focus: JD.Com Opens Unmanned Store in Indonesia: https://www.retailtechnews.com/2018/08/08/weekly-focus-jd-com-opens-unmanned-store-in-indonesia/

  9The Drum, JD.com expands 7FRESH stores across China as it takes on Alibaba's Hema stores: https://www.thedrum.com/news/2018/09/ 24/jdcom-expands-7fresh-stores-across-china-it-takes-alibabas-hema- stores

  10Afr.com, AI Inside JD: https://www.afr.com/technology/how-chinese- ecommerce-player-jdcom-is-becoming-an-ai-powerhouse-20180719- h12vph

  11YouTube, Richard Liu, JD.com Founder, Chairman and CEO: https:// www.youtube.com/watch?v=VTSKy9E3tcU&feature=youtu.be

  9

  Microsoft: Making Artificial Intelligence Part Of The Fabric Of Everyday Life

  Microsoft is a US-based, multinational technology company. As one of the world's most valuable companies it focuses on areas including computer software, consumer electronics, video gaming, cloud computing and social media.

  Microsoft's business model has revolved around bringing technology to the masses. Its operating systems helped make computers in homes and small businesses truly useful for millions of people. And its office productivity tools elevated the understanding of spreadsheets, d
atabases and presentation software among the wider population. With projects such as Internet Explorer and its .Net framework, it was also a key player in opening up the world of information we have at our fingertips, thanks to the modern internet.

  Microsoft's artificial intelligence (AI) strategy is built along this same guiding principle. CEO Satya Nadella has spoken about “democratizing AI”1 – which means not just giving as many people as possible access to its benefits, but ensuring they have a say in the future of its development.

  With this aim in mind, Microsoft provides tools for developers to create their own AI applications through its Azure Cognitive Services platform. Microsoft's vision is that business trends involving AI will follow a similar path as they did with the internet – eventually everyone will be on board – but they believe the effects on business and society could be even more profound.2

  How Does Microsoft Use Artificial Intelligence?

  Following the path it carved with Office and Word, Microsoft's goal is to give businesses tools they can use to deploy AI in their organizations.

  At its most basic level, this could simply mean taking advantage of the AI tools contained in Office 365. PowerPoint is capable of giving design tips based on how it observes the user working, and Word uses AI to suggest meanings, alternative phrases and to check spelling, grammar and punctuation. A feature called Acronyms even works to decipher the irritating abbreviations and shorthand that can build up around an organization. It does this by analyzing emails and internal documents to understand specific language traits, and provide automated translations.3

  Azure Cognitive Services offers “pre-built” machine learning solutions for speech recognition, text analysis, computer vision and language translation. The idea is that anyone who has an idea about how their business could use AI to get value from its data can jump in without having to be an expert in AI.

  They will of course need some idea about how data can be used to drive growth and the key technologies, so to help with this, Microsoft offers an online AI School. This is a collection of resources that covers the basics of what AI can do and how to start using it.4

  Microsoft is even trying to make it easy for us to build our own robots. Their school offers courses on the open source Robotic Operating System and the robot simulator Gazebo, as well as instructions on integrating it with Azure Cognitive Services to build truly smart, self-learning robots.5

  Another tool that has the potential to be very useful is Sketch2Code, which is capable of generating working HTML websites from simple sketches. It uses computer vision to understand hand-drawn sketches and transform them into wireframes and working websites. It was trained on thousands of images of hand-drawn page design elements such as buttons and text boxes, as well as hand-written data.6

  Underwater Data Centers

  Cloud-based AI requires a lot of network bandwidth, and putting it in the hands of everybody means that everybody needs access to the bandwidth. To tackle this challenge, Microsoft is trialing Project Natick, which involves submerging data centers under the ocean close to coastal cities. The submersible data centers are the size of shipping containers and fully self-contained, so they can operate autonomously for years without causing any pollution.7 As 50% of us live near a coast, the idea has the potential to bring about a quantum leap in home internet speeds.

  Who Uses Microsoft Artificial Intelligence?

  Microsoft Face API is a facial recognition system that is used by Uber to verify the identity of its drivers when they sign into their system to work. Drivers are required to periodically update their photo, and the Azure computer vision algorithm has to be able to match identities of millions of drivers in a fraction of a second. This gives customers peace of mind that the person driving them is who they say they are.8

  Microsoft also works with the Renault Formula One team to build machine learning simulators that analyze every aspect of their F1 cars’ performance on the circuit. The cars use an array of over 200 sensors to send data on everything from tyre wear to the condition of the track and temperature readings from the engine to Azure cloud servers. There, machine learning algorithms draw out insights that can be used to create more accurate simulations and improve racing performance.9

  But as we established, Microsoft wants to give smaller businesses the power to use AI too. One interesting project is their work with Spektacom, a sports technology company formed by former Indian cricket captain Anil Kumble. Their technology uses Azure to interpret data from a tiny sensor weighing around 5 grams, which can be attached to a cricket bat. The aim is both to give coaches access to more accurate data on how players perform and to enable more engaging stats and interactivity for fans.10 The plan is to roll the sensor and analytics technology out into other sports.

  Bonsai

  This year Microsoft acquired start-up Bonsai, which specializes in a semi-unsupervised form of machine learning known as reinforcement learning. Bonsai has developed proprietary methods for deep reinforcement learning, which Microsoft says will become the “brains” of its new autonomous systems. Bonsai has been noted for its work building intelligent systems that can be used for programming industrial control systems.

  According to Microsoft, the system allows someone with experience in industrial control system programming, but no AI experience, to program a machine 30 times more quickly than they could through conventional methods.11

  Key Challenges, Learning Points And Takeaways

  Microsoft boss Satya Nadella's vision is that AI will eventually become simply a part of the fabric of everyday life – much like computers and the internet have done.

  To achieve this his company is building tools and services that let other businesses carry out machine learning through their Azure cloud infrastructure.

  It also includes AI functionality in its mainstream office productivity software, which is already used by millions, providing tools to make jobs quicker and easier with help from machine learning.

  Microsoft has partnered with businesses of all shapes and sizes to roll out AI solutions and is now diving into reinforcement learning with its acquisition of Bonsai.

  Notes

  1Microsoft, Democratizing AI: Satya Nadella on AI vision and societal impact at DLD: https://news.microsoft.com/europe/2017/01/17/ democratizing-ai-satya-nadella-shares-vision-at-dld/

  2Microsoft, Microsoft AI: Empowering transformation: https://blogs .microsoft.com/ai-for-business/2018/10/11/microsoft-ai-empowering- transformation/

  3Redmond Magazine, Office 365 Gets Serious About Artificial Intelligence: https://redmondmag.com/articles/2018/02/16/office-365-gets-serious-about-ai.aspx

  4Microsoft, AI School: https://aischool.microsoft.com/en-us/home

  5Microsoft, Intelligent Robotics: https://www.ailab.microsoft.com/ experiments/f508a96d-3255-474b-a769-d5b2cf2bb9d6

  6Alphr, Microsoft's AI-powered Sketch2Code builds websites and apps from drawings: http://www.alphr.com/microsoft/1009840/microsofts-ai-sketch2code-builds-websites

  7Microsoft, Project Natick: https://natick.research.microsoft.com/

  8Microsoft, Uber boosts platform security with the Face API, part of Microsoft Cognitive Services: http://customers.microsoft.com/en-US/story/uber

  9Microsoft, Renault Sport Formula One Team uses data to make rapid changes for an even faster race car: https://customers.microsoft.com/en-US/story/renault-sport-formula-one-team-discrete-manufacturing

  10The Seattle Times, Cricket pro teams with Microsoft for a bat that can track analytics in real-time, and send them to fans: https://www.seattle times.com/business/microsoft/microsoft-partners-with-professional- cricketer-to-make-smart-bat-technology/

  11Bonsai, Bonsai Sets State-of-the-Art Reinforcement Learning Benchmark for Programming Industrial Control Systems: https://bons. ai/blog/rl-benchmark

  10

  Tencent: Using Artificial Intelligence To Power WeChat And Healthcare

  Tencent is a Chinese multinational internet-services and technology
conglomerate. Today, Tencent's success in gaming and social media has made it one of the most valuable technology companies in the world. It is probably best known for its WeChat app, a mobile messenger service with social, picture sharing and pay functions. It's the largest social media platform in the world, with over 1 billion active monthly users.

  Tencent has a huge number of interests stretching across industries from banking and real estate to space exploration and healthcare, but its focus is always the application of new technology. It is particularly prominent in the gaming and entertainment fields. This is probably why the motto of Tencent AI Lab is “Make AI Everywhere”.1

  How Does Tencent Use Artificial Intelligence?

  Tencent invests aggressively in artificial intelligence (AI) start-ups when it finds technology that can drive efficiency in any of its myriad business activities. In 2017, it made the highest number of US AI-related investment deals of any of the Chinese giants.2

  It has particularly stood out for the advancements it has made in facial recognition technology. This is so advanced that in three Chinese provinces, citizens are allowed to verify their identity through WeChat digital ID cards, rather than having to carry physical cards.3

  The technology is also used in its video games. Amid public concern in China that prolonged gameplaying is having a detrimental effect on the health and education of children, Tencent is trialling technology that can determine if a gameplayer is under age by asking them to undergo an automated on-camera check. Once the service is live, players who refuse to take part or fail the algorithm's checks could be blocked from the game.4