Artificial Intelligence in Practice Read online

  You can simply read this book cover to cover or dip in and out to explore the case studies or industries you are most interested in. We hope you will enjoy it!

  Notes

  1Preparing for the Future of Artificial Intelligence, Executive Office of the President, National Science and Technology Council, National Science and Technology Council Committee on Technology, October 2016: https://obamawhitehouse.archives.gov/sites/default/files/whitehouse_files/microsites/ostp/NSTC/preparing_for_the_future_of_ai.pdf

  2Artificial Intelligence for the American People, The White House: https://www.whitehouse.gov/briefings-statements/artificial-intelligence -american-people/

  3Summary of the 2018 White House Summit on Artificial Intelligence for American Industry, The White House Office of Science and Technology Policy 10 May 2018: https://www.whitehouse.gov/wp-content/ uploads/2018/05/Summary-Report-of-White-House-AI-Summit.pdf

  4“Whoever leads in AI will rule the world”: Putin to Russian children on Knowledge Day: https://www.rt.com/news/401731-ai-rule-world-putin/

  5A Next Generation Artificial Intelligence Development Plan: http:// www.gov.cn/zhengce/content/2017-07/20/content_5211996.htm and Three-Year Action Plan to Promote the Development of New-Generation Artificial Intelligence Industry: http://www.miit.gov.cn/ n1146295/n1652858/n1652930/n3757016/c5960820/content.html

  6Communication from the Commission to the European Parliament, the European Council, the Council, the European Economic and Social Committee and the Committee of the Regions, Artificial Intelligence for Europe, Brussels 2018: https://ec.europa.eu/digital-single-market/en/news/communication-artificial-intelligence-europe

  7A.I. is in a “golden age” and solving problems that were once in the realm of sci-fi, Jeff Bezos says, CNBC: https://www.cnbc.com/2017/05/ 08/amazon-jeff-bezos-artificial-intelligence-ai-golden-age.html

  8Google's Sergey Brin warns of the threat from AI in today's “technology renaissance”: https://www.theverge.com/2018/4/28/17295064/google-ai-threat-sergey-brin-founders-letter-technology-renaissance

  9Microsoft CEO Satya Nadella on the rise of A.I.: “The future we will invent is a choice we make”: https://www.cnbc.com/2018/05/24/ microsoft-ceo-satya-nadella-on-the-rise-of-a-i-the-future-we-will- invent-is-a-choice-we-make.html

  10The Fourth Industrial Revolution: what it means, how to respond, Klaus Schwab, World Economic Forum: https://www.weforum.org/ agenda/2016/01/the-fourth-industrial-revolution-what-it-means-and- how-to-respond/

  Part 1

  Artificial Intelligence Trailblazers

  1

  Alibaba: Using Artificial Intelligence To Power The Retail And Business-To-Business Services Of The Future

  Alibaba Group is a Chinese multinational conglomerate that operates the world's largest e-commerce network through its web portals, which include Alibaba.com, Taobao, Tmall and Ali Express. With global sales that dwarf those of Amazon and eBay combined,1 the business took what it learned from building a global online retail platform and has applied it to enterprises in just about every area of business and technology. Alibaba's success in delivering e-commerce and retail services, electronic payment, as well as business-to-business cloud services, has earned it a market cap in excess of US$500 billion.

  Its customers use artificial intelligence (AI) tools to help them find what they want when they shop at its online portals, and as one of the world's largest cloud computing providers it also licenses platforms, tools and cloud services to other businesses to help them leverage AI.

  Beyond that, Alibaba is rolling out AI across the wider society, with projects involving turning entire cities into “smart cities”. They are also planning on revolutionizing China's (and perhaps the world's) agricultural industries to ease the burden of feeding a growing population.

  How Does Alibaba Use Artificial Intelligence?

  The Chinese government has strongly supported efforts by businesses to adopt AI, clearly believing that it has enormous potential for driving economic growth. Its goal is to foster a $1 trillion industry and be the world leader in AI by 2030.2

  This, combined with the fact that the country's enormous population gives companies access to huge amounts of data on customers’ lives, makes the country a fertile ground for AI development.

  Alibaba's e-commerce portals use sophisticated AI to choose which items to display to customers when they visit and search for products they want to buy. It does this by building a custom page view for every visitor, aimed at showing them items they will be interested in, at prices that seem right.

  By monitoring customer actions – whether they make a purchase, browse to a different item or leave the site – it learns in real time to make adjustments to these page views to increase the probability of the visit ending in a purchase.

  To train its e-commerce portals to show visitors pages that are likely to result in a sale, Alibaba has deployed a form of semi-supervised learning known as reinforcement learning on its Taobao portal.3

  Because collecting enough user data to train unsupervised learning algorithms from real-time customer actions would take a long time, and involve real business risks, a virtual Taobao was built, with customer behavior simulated from hundreds of thousands of hours’ worth of historical customer data.

  This mass of data meant that it was possible for the algorithms to be exposed to a far wider range of customer behaviors, in a far shorter time span.

  Alibaba also has its own AI-powered chatbot – Dian Xiaomi – that answers more than 350 million customer enquiries a day, successfully understanding more than 90% of them. These tools are necessary to help it deal with the huge spikes generated by special occasions such as the Alibaba-created “Singles’ Day” shopping event.4

  Automated Sales Copy

  With millions of different items on sale across its sites, Alibaba has invested in automated content generation to ease the burden of writing descriptions for everything it sells. The tools are also available to third-party sellers on its platforms.

  Its AI copywriter uses natural language processing algorithms running on deep learning neural networks to produce 20,000 lines of copy in a second.5

  Traditionally, sales copywriters have had to spend hours researching keywords and click-through rates to understand what is likely to make a customer click their link in a page of product search results. The AI copywriter allows Alibaba and others selling through its platforms to do it at the click of a button.

  This is done by creating multiple versions of adverts and running them through algorithms trained on customer behavior data. The system works out which combination of words is most likely to result in customer clicks, and uses them to create its copy.

  Cloud Services

  Just like Amazon and Google, Alibaba offers artificially intelligent services through the cloud to its business customers. Its cloud service business is the largest of all the Chinese tech giants.6

  Alibaba's AI offering is called Machine Learning Platform for AI, which offers solutions for businesses wanting to take advantage of cognitive computing functions such as natural language processing and computer vision, without the upfront costs of directly investing in infrastructure.

  Alibaba's natural language processing technology was the first in the world to beat a Stanford University test designed to assess whether a machine can beat a human at reading comprehension.

  In 2018, its deep neural network language processing technology passed the 100,000 question test with a score of 82.44 – narrowly beating the human score of 82.3.7

  Smart Cities

  Alibaba has developed a suite of cloud-based AI tools designed to carry out essential jobs like managing traffic flow, lighting and waste collection in cities where infrastructure is connected through smart online technology.

  Alibaba City Brain already tracks and manages traffic flow on every street ofHangzhou, a city with a population of 9.5 million. The system is reported to have reduced traffic jams by 15%8 and is soon expected to be deployed in Kuala L
umpur, Malaysia.

  City Brain monitors the flow of traffic and builds up models that it can use to predict when congestion is likely to occur. When it recognizes signs that there is a high probability of this happening, it can alter traffic light patterns to speed up or control the flow of traffic, so jams are less likely to form.

  Alibaba's AI also powers the smart ticket kiosks at Shanghai's subway stations. The kiosks give customers route information when asked, and check customer identification using facial recognition technology.9

  Smart Farming

  Alibaba has developed an AI system for monitoring farm herds, crops and orchards.

  As the world's biggest supplier and consumer of pork, Chinese pig farmers have access to technology that records activity and health levels of herds, automating decision making over when to increase feed or provide animals with more exercise.10

  Facing the challenge of feeding an ever-growing population, the system allows farmers to optimize breeding rates by raising a healthier herd and reducing newborn death rates. The system also has applications in crop growing and land management.

  Academy For Discovery, Adventure, Momentum And Outlook

  Alibaba's AI strategy is based around distributing its cutting-edge machine learning and deep learning solutions to businesses and customers through its cloud services.

  Its business AI platform is delivered through its Alibaba Cloud subsidiary, which operates 18 global data centers. These host the hardware that powers the AI algorithms and data processing technology, which is provided as a service.

  In 2017 it announced it would invest $15 billion over the next three years, expanding its global network of AI research and development facilities.

  It calls this program the Academy for Discovery, Adventure, Momentum and Outlook – DAMO – and will involve recruiting 100 researchers for its labs in Beijing and Hangzhou, China, and San Mateo and Bellevue in the United States, as well as others in Moscow, Tel Aviv and Singapore.11

  Research at the labs will focus on machine learning, natural language processing, Internet of Things, human/machine interaction and quantum computing.

  Key Challenges, Learning Points And Takeaways

  Alibaba is China's biggest investor in research and development, which has given it a strong start in the race to become the world leader in AI.

  Its model for rolling out AI to millions of customers and businesses is to deploy its services through the cloud. This cuts customer risk and infrastructure cost, while giving Alibaba access to valuable data about how its customers behave.

  By applying technology designed to drive sales at its retail portals to other problems in business and society, it identifies new use cases for AI, within and outside its established business operations.

  Notes

  1Institutional Investor, Ali Baba vs The World: https://www. institutionalinvestor.com/article/b1505pjf8xsy75/alibaba-vs-the-world

  2CNBC, China is determined to steal A.I. crown from US and nothing, not even a trade war, will stop it: https://www.cnbc.com/2018/05/04/ china-aims-to-steal-us-a-i-crown-and-not-even-trade-war-will-stop-it .html

  3Virtual-Taobao: Virtualizing Real-world Online Retail Environment for Reinforcement Learning: https://arxiv.org/abs/1805.10000

  4SCMP, Alibaba lets AI, robots and drones do the heavy lifting on Singles’ Day: https://www.scmp.com/tech/innovation/article/2119359/alibaba-lets-ai-robots-and-drones-do-heavy-lifting-singles-day

  5BBC, The world's most prolific writer is a Chinese algorithm: http://www.bbc.com/future/story/20180829-the-worlds-most-prolific-writer-is-a-chinese-algorithm

  6Data Center News, Alibaba gives AWS, Microsoft and Google a run for their cloud money: https://datacenternews.asia/story/alibaba-gives-aws-microsoft-and-google-run-their-cloud-money/

  7Bloomberg, Alibaba's AI Outguns Humans in Reading Test: https://www.bloomberg.com/news/articles/2018-01-15/alibaba-s-ai- outgunned-humans-in-key-stanford-reading-test

  8Wired, In China, Alibaba's data-hungry AI is controlling (and watching) cities: https://www.wired.co.uk/article/alibaba-city-brain-artificial-intelligence-china-kuala-lumpur

  9Technology Review, Inside the Chinese lab that plans to rewire the world with AI: https://www.technologyreview.com/s/610219/inside-the-chinese-lab-that-plans-to-rewire-the-world-with-ai/

  10Financial Times, Alibaba brings artificial intelligence to the barnyard: https://www.ft.com/content/320fb98a-69f4-11e8-b6eb-4acfcfb08c11

  11CNBC, Alibaba says it will invest more than $15 billion over three years in global research program: https://www.cnbc.com/2017/10/11/alibaba-says-will-pour-15-billion-into-global-research-program.html

  2

  Alphabet and Google: Maximizing The Potential Of Artificial Intelligence

  Alphabet is a US-based multinational internet services, technology and life-sciences conglomerate. Its businesses include internet-search giant Google, life-sciences company Verily, self-driving technology company Waymo, smart home device company Nest, artificial intelligence (AI) company Deep Mind, among others.

  In his founder's letter in 2017, Sergey Brin, the president of Alphabet, wrote: “The new spring in artificial intelligence is the most significant development in computing in my lifetime.”1 Given that this includes the arrival of the internet, it's no small statement.

  Alphabet understands the potential of AI and is set to use it across its businesses, from improving internet searches, to self-driving cars, automated homes, intelligent virtual assistants, language translation and life-saving medical science.

  How Does Alphabet Use Artificial Intelligence?

  Smarter Searching

  Google's search engine – the most widely used in the world – is peppered with AI. Whether you use its text, voice or image search capabilities, every query is now (since at least the introduction of its Rankbrain feature in 2015) processed by smart, self-teaching systems.2

  Text and voice search both employ natural language processing, so the algorithms attempt to understand how each word you enter as part of a search query relates to every other word it is used with, rather than just what each word means individually. This is semantic analysis, the key to natural language processing.

  Google Image search uses computer vision to recognize the content of image data cataloged by Google, and classifies it so users can search for it using text or voice. Deep learning algorithms allow it to become increasingly good at recognizing and labelling different elements contained in pictures. The greater the variety of images it is exposed to, the better it becomes at knowing what they are.

  Once Google's AI has processed your query and decided what it thinks you really want, it matches it against its directory of online content – web pages, images, videos and documents. These have also been processed by machine learning systems.

  The systems are trained to sort, rank and filter all of the content in its directory. Content is assessed for how frequently cited (linked) it is, the accuracy of information it contains, the possibility that the information might be spam or advertising, and whether it is likely to be illegal or copyright infringing.

  This means a simple Google search involves a great deal of complex, blisteringly fast AI calculations. Building systems capable of processing billions of calculations every day from all around the world is what has made Alphabet and Google a genuine giant in the field of AI (as well as one of the richest companies in the world).

  Google uses AI for many of its other core applications, including security measures, which keep Gmail accounts secure, and adwords, which allow businesses to pay for their ads to appear in searches of customers who may be interested.

  Artificial Intelligence Personal Assistants

  AI personal assistants using voice technology have been around for a few years now and Google Home, Amazon Alexa and Apple Siri are familiar to most of us.

  Although these first implementations of natural language processing into consumer devices seem impressive compared to what was possible just a few years ag
o, anyone who has used one will know they have limitations. They can respond well to basic, relatively short sentences and commands, but try talking to them like you would an actual human and things start to unravel.

  This is because, in human terms, they are still very much infants. Put simply, they haven't had enough data yet. This is quickly changing, and Google's Duplex tech is leading the charge.

  Duplex is able to hold far more natural, less jilted conversations. This is because it is specifically trained for particular situations, and its algorithms exclusively specialize in gathering data that is relevant to those situations. An example used by Google to showcase its ability features Duplex making a call to book an appointment at a hair salon on behalf of its user.3 In these relatively controlled and constrained use cases, it comes very close to appearing perfectly human.

  One trick used by Google's engineers to get the machine to sound more human was to incorporate imprecise elements of our speech patterns. For example, it will utter an “umm”, an “aah” or an “mh-hmm” in places where it might seem natural for a human to do so.

  Language Translation

  Thanks to machine learning, if you can teach a computer to speak one language, it can teach itself to speak any language. That's the principle behind Google's language translation service, which uses deep learning to break languages down to their fundamental building blocks.

  Google Translate uses deep neural networks to constantly refine its algorithms as its users expose it to more languages. This means it becomes increasingly efficient at accurate translations. Google has even built the feature into its Google Assistant-powered Pixel Bud headphones, meaning users can get near real-time translations directly through their headsets.4