VinIntell SEPTEMBER 2017, ISSUE 33

2017 THE YEAR OF ARTIFICIAL INTELLIGENCE CONTENTS BACKGROUND gence techniques are pervasive and are too numerous to list here. High-profile examples Background...... 3 Flying defibrillators, Whoppercoins buying of AI include autonomous vehicles (such as Introduction...... 4 hamburgers, drone-based parachutes, your drones and self-driving cars), medical diagno- Defining artificial intelligence ...... 4 Friday night take-out sushi prepared by a sis, creating art (such as poetry), proving robot and upper-assembling robots that make Drivers of the AI Surge ...... 6 mathematical theorems, playing games (such Nikes 20 times the pace of human workers ... as Chess or Go), search engines (such as Type and examples of AI Technology ...... 7 virtual sommeliers ... educational robots that Google search), online assistants (such as AI in real life ...... 8 teach children how to code ... welcome to the Siri), image recognition in photographs, spam world of this is the world of Artificial Intelli- Policy and strategic pointers ...... 15 filtering, prediction of judicial decisions and gence (AI or machine learning). This world is Conclusion ...... 18 targeting online advertisements. Major pub- characterised by the term “The Fourth Indus- lishers now use artificial intelligence (AI) tech- Endnotes ...... 19 trial Revolution” (after steam power, electricity nology to post stories more effectively and and information technology); an era that will generate higher volumes of traffic. be defined and driven by extreme automation and ubiquitous connectivity. Robots and This edition of VinIntell will broadly explore machine learning will play an increasingly the concept of AI, provide some definitions prominent role in all industries including finan- and then have a high-level view of how it is cial, agriculture and medicine and we will see impacting in various economic sectors most a closer merger of biological intelligence and notably in agriculture. Finally, a number of digital intelligence.1 Modern artificial intelli- strategic pointers are provided.

2 3 INTRODUCTION The potential of artificial intelligence and tfried Leibniz’ extended the concept of the methods and scientific standards. As recently advanced robotics to perform tasks once calculating machine intending to perform as 20 years ago (in 1997), Deep Blue became A blog entry on the World Economic Forum reserved for humans is no longer reserved for operations on concepts rather than numbers. the first computer chess-playing system to (WEF) website earlier in 2017 states that spectacular demonstrations by the likes of However, it really started as a field at Dart- beat a reigning world chess champion, Garry Artificial Intelligence (AI) is now a “hot topic” IBM’s Watson, Rethink Robotics’ Baxter, mouth College. Attendees Allen Newell (CMU), Kasparov.6 in business and government circles. Social DeepMind, or Google’s driverless car. Just Herbert Simon (CMU), John McCarthy (MIT), Today, AI is an umbrella term that encom- robots, including companion and entertain- head to an airport: automated check-in Marvin Minsky (MIT) and Arthur Samuel (IBM) passes everything from robotic process ment robots for homes, are attracting great kiosks now dominate many airlines’ ticketing became the founders and leaders of AI automation to actual robotics and has investment deals. A May 2017 forecast areas. Pilots actively steer aircraft for just research. They produced programs that the become a buzz word in a time when there is stated that revenue generated from the direct three to seven minutes of many flights, with press described as “astonishing”: “computers an increase in the speed, size and variety of and indirect application of AI software would autopilot guiding the rest of the journey. were winning at checkers, solving word prob- data businesses are now collecting. It grow from US$1.4 billion in 2016 to US$59.8 Passport-control processes at some airports lems in algebra, proving logical theorems and includes the terms machine learning and billion by 2025. This forecast represents an can place more emphasis on scanning docu- speaking English.” In the beginning of the deep learning (see figure 1). upgrade of a previous projection for AI mar- ment bar codes than on observing incoming 2000’s, AI began to be used for logistics, data ket growth, which was published in 3Q16, passengers. Claims of AI advances are eve- mining, medical diagnosis and other areas UBS South Africa has a succinct definition of owing to a greater than anticipated pace of rywhere these days, coming even from the showcasing its computational power, greater AI. It says AI can be viewed as being pro- 2 change and development in the AI sector. marketers of fast food and toothbrushes. emphasis on solving specific problems, new grammes and tools that make software more Recent advancements in AI, and specifically Even boasts from solid research teams can ties between AI and other fields and a com- intelligent “in a way an outside observer thinks in machine learning, have contributed to the be difficult to assess. Microsoft first an­­ mitment by researchers to mathematical the output is generated by a human.” AI uses growth of Autonomous Things such as nounced it had matched humans at speech drones, robots and self-driving cars. Besides recognition in October 2016. these tools and functions however, more Let us view some popular definitions of AI. significant is how software is being rede- signed and how systems are built, what can DEFINING ARTIFICIAL be programmed and how users interact. Machines have started to grasp and antici- INTELLIGENCE pate what we want to achieve and soon they AI is really intelligence generated by machines will do it for us. Computing will never be the (as opposed to natural intelligence namely same and things we never imagined to be intelligence generated by humans) or getting possible are now happening. An example: machines to do the things which at the There are now flying defibrillator drones that moment humans do better (including machine comes right to the site of an emergency in a learning and deep learning).4 The term Artificial matter of minutes meaning medical help can Intelligence was coined by John McCarthy, an take a path that’s as short as the crow flies. American computer scientist, in 1956 at The Similarly a drone-based parachute drops Dartmouth Conference where the discipline blood donations and other medical supplies was born while some literature says that AI to Rwanda while a number of Swiss hospitals began in the early 1960’s.5 Still others claim began using drones to transport lab samples some of the earliest AI tools include Ramon Figure 1: Evolution of AI back and forth.3 Llull (c. 1300 CE)’s Calculus ratiocinator, Got- Source: Nvidia, 2017

4 5 self-learning systems through the use of data DRIVERS OF THE AI SURGE Google Drive, Apple iCloud, Netflix, TYPES AND EXAMPLES OF mining, pattern recognition and natural lan- Yahoo Mail and Dropbox).11 A number of factors are driving the stunning AI TECHNOLOGY guage processing (NLP) and emulates human 2. Scalability. Processors (chips) are turned growth in AI and experts argue that should brain processes for simple tasks. In business out more frequently and cheaper than Broadly the types of AI we see today (machine these conditions continue, the types of AI we AI has value in its high scalability (think the ever before. Their ubiquity means chips learning and deep learning) include optical see today will continue to flourish and, possi- Nike robot which makes shoe uppers), its abil- are in everything and in turn computing character recognition, voice recognition, bly, more general AI might actually become a ity to decrease errors amongst others and its capacity is accessible at a fraction of the autonomous vehicles and content delivery reality. But one thing is certain: if everything is ability to perform tasks such as identifying cost of a few years ago all the time from networks e.g. video streaming, software a connected computer device, and all infor- patterns in the data more efficiently than everywhere. downloads, web and mobile content accel- humans, enabling businesses to gain more mation can be known, processed and ana- eration and cloud intelligence. Arend Hintze, 3. Data proliferation is the third factor. The insight out of their data.7 8 Put in another way, lysed intelligently, then the things we would be an assistant professor of integrative biology volumes and types of data available digi- it is about outsourcing cognitive processes to able to program and change are unthinkable. and computer science and engineering at tally have proliferated exponentially over a computer so it could think, analyse, antici- What are these driving factors that fuel the Michigan State University stated there are the last decade, as everything has moved pate, pre-empt and act independently without growth of AI? basically four types ranging from those we see online, been made mobile with smart- explicit programming.9 1. Connectivity is reaching ubiquity as never today to those that do not yet exist. phones, and tracked via sensors. New before. Sensors are built into everything AI is fast moving in the direction where it more sources of data emerged through things 1. Reactive machines (e.g. Deep Blue, the and the Internet now connects comput- than imitates human thinking. Since the like social media, digital images and IBM chess program that beat Garry Kasp- ers to mobile devices while buildings, 1980’s machine learning and deep learning video. This is the language that machines arov in the 1990’s) can identify pieces on infrastructures, machinery, homes, have created even larger disruptions than the understand and it is this data that is ena- the chess board and make predictions. clothes are connected through the cloud. early AI breakthroughs. Already we have Siri bling machines to learn. Reactive machines have no memory and Everything turns into a device that facili- which can understand language, look for rel- cannot use past experiences to inform tates connectivity so data and orders are 4. Data has become more usable: The evant information and provide that information future ones. A reactive machine analyses being sent all the time. When we say we emergence of machine learning due to a in an appropriate answer. Further on the spec- possible moves (its own and its opponent) work in the cloud it really means that data need to make data more useable includ- trum of AI is self-driving cars but again there and chooses the most strategic move. is stored in different places and that there ing how algorithms and mathematical are limitations as is has only one ‘person- Other examples include Microsoft’s devel- is a network of servers that find what we models are used to discover patterns byte’10 – it cannot clean the house, shop for opment of a Skype system that can auto- need and deliver it (think web browsers, implicit in data. The machines then use food and be a complete human. AI does one those often complex patterns to figure matically translate from one language to or two of these tasks well and they save time out on their own whether a new data another and Facebook’s system that can and effort and efficacy. The promise of AI is point fits, or is similar, or to predict future describe images to blind people. just astounding as the pace in which AI is outcomes. Robots learn to make bever- 2. AI systems with limited memory which developing. Some experts say we should put ages and cook using YouTube videos for can use past experiences to inform future a speedometer on its development. In August example. More machine learning models decisions: Some of the decision-making 2017, Microsoft said it reached a new record have emerged recently that seem to be functions in autonomous vehicles have for the accuracy of software that transcribes able to take better advantage of all the been designed this way. Observations speech claiming its system missed just one in new data. For example, deep learning used to inform actions happening in the 20 words on a standard collection of phone enables computers to ‘see’ or distinguish not-so-distant future, such as a car that call recordings, matching humans given the objects and text in images and videos has changed lanes. These observations same challenge. better than before. are not stored permanently. Source: Dreamstime

6 7 At the other end of the spectrum, there are two types of during transportation offer a wealth of infor- The ways in which AI provide benefits are AI that do not yet exist. Theory of mind refers to the BOX 1 mation about soil, seeds, livestock, crops, numerous. Here are a few examples: understanding that others have their own beliefs, desires Jeff Heepke knows where to plant costs, farm equipment or the use of water • Agricultural drones help already farmers and intentions that impact the decisions they make. mealies on his farm in Illinois and fertilizer. Internet of Things (IoT) tech- because of artificial intelligence (AI). scan fields, monitor crops and seeding or Finally, there is self-awareness which is AI that has a nologies and advanced analytics help farm- He uses a smartphone app called analyse plant health. Farm activities can sense of self. Such machines understand their current ers analyze real time data like weather, tem- Climate Basic, which divides become much more effective when such state and can use the information to infer what others are Heepke’s farmland (and, in fact, the perature, moisture, prices or GPS signals and drone data, IoT and computer vision tech- feeling. entire continental US) into plots provide insights on how to optimize and nologies join forces to optimise strategies. that are 10 meters square. The app increase yield, improve farm planning, make So from robotic process automation (robots programmed draws on local temperature and These AI systems will save time, increase smarter decisions about the level of resourc- to perform high-volume, repeatable tasks normally per- erosion records, expected safety and reduce potential human error es needed, when and where to distribute formed by humans), machine learning (computers per- precipitation, soil quality, and other while improving effectiveness. forming without programming) we move to machine agricultural data to determine how them in order to prevent waste. Logistics vision (the science of making computers capturing and to maximise yields for each plot. If company Knapp AG developed a picking • The challenge to find good labour at a a rainy cold front is expected to technology using augmented reality. Pickers time of urbanisation will be alleviated by AI analysing visual information using a camera, analogue- pass by, Heepke knows which and many operations will be done remote- to-digital conversion and digital signal processing areas to avoid watering or irrigating wear a headset that presents vital information unbound by biology) to natural language processing that afternoon. on a see-through display, helping them locate ly, processes will be automated, risks will (NLP) and robotics. NLP is the processing of human Source: Strategy+Business, May 2017 items more quickly and precisely. And with be identified and issues solved before language (as opposed to computer language) by a com- both hands free, they can build stronger and occurring. Farmers will be able to take puter program e.g. spam detection, text translation, and more efficient pallets, with fragile items safe- more informed and rapid decisions. In the sentiment analysis and speech recognition. Robotics is about the design and manufacturing of guarded (Winetech, 2017). future, the right mix of skills will probably increasingly be technology and agricul- robots that are used to perform tasks that are difficult for humans to perform or perform consist- With AI, efficiency and productivity will tural skills rather than pure agricultural. ently. Typically, we are familiar with the assembly lines for vehicle production for example. More increase in the next years as precision agri- recently, researchers are using machine learning to build robots that can interact in social settings. culture grows bigger and farms become • The use of cognitive technologies in agri- smarter and more connected. While the culture could help determine the best crop growing number of connected devices rep- choice or the best hybrid seed choices for look at what is happening in the wine and AI IN REAL LIFE resents a big opportunity for food and agri- a crop mix adapted to various objectives, other industries. conditions and better suited for farm’s Newsweek in an article published in March business players, it also adds more complex- ity for farmers and organizations. Moreover, needs. IBM Watson for example can use 2017, said while all along we were concerned In agriculture and the wine industry in the explosion of unstructured data, like social diverse capabilities to understand how about the loss of low-income low-skilled jobs, particular media posts, imagery or video content drives seeds react to different soil types, weath- AI also threatens certain high-end jobs e.g. Modern agriculture is integrated in a more the need to know more, to receive real time er forecasts and local conditions. By hedge fund managers and stock brokers.12 complex arrangement than ever before driv- recommendations on close at hand devices, analysing and correlating information Goldman Sachs shows just how devastating en by multiple sociological, economic and like smartphones or tablets. It has become a about weather, type of seeds, types of soil automation can be to traders. In 2000, its US environmental forces. Digital and technologi- challenge to the human brain to consistently or infestations in a certain area, probabil- cash equities trading desk in New York cal advancements are taking over the indus- gather and process so much information and ity of diseases, and data about what employed 600 traders. Today, that operation try, enhancing production while adding value turn it into operational and strategic deci- worked best, year to year outcomes, has two equity traders, with machines doing to the entire farm-to-shop supply and value sions. The solution? The use of cognitive marketplace trends, prices or consumer the rest. Is AI and will it have an equally direct chain. Data generated by sensors or agricul- technologies that help understand, learn, needs; farmers can make decisions to impact on other industries? Let us have a brief tural drones collected at farms, on the field or reason, interact and thus, increase efficiency. maximize return on crops.

8 9 • Chatbots are conversational virtual assis- wine selectors have been around for some In terms of robotics, AI is finding its place in In healthcare tants who automate interactions with end time but now intelligence is being integrated the South African wine industry. The Institute Artificial intelligence is breaking into the users. Artificial intelligence powered chat- into these solutions making them highly per- for Viticulture and Oenology Sciences healthcare industry in a major way. The big- bots, using machine learning techniques, sonalised.15 Some of the apps and names (IWWW) launched a robotics project. Its Das- gest bets are on improving patient outcomes understand natural language and interact have a familiar ring: Wine Ring Wine Ring sie is a remote controlled machine that aims and reducing costs. According to Bloomberg with users in a personalised way. While it (www.winering.com) bases suggestions on an to achieve sustainability in the South African Technology, Microsoft has developed AI to individual’s preferences using algorithms to wine industry and make it more future is still early, and chatbots are used mostly help doctors find the right treatments for develop a personal profile based on the per- focused. The Dassie has sensors that are by retail, travel, media or insurance play- cancer and there was a recent study by sur- son’s rating of wines16 and Google’s “My Wine monitored and can amongst others map soil ers, agriculture could also leverage this geons at the Children’s National Medical Guide,” a conversation action added to Goog- difference. It is Wi-Fi controlled from a com- emerging technology by assisting farmers Center in Washington which successfully le Assistant for wine pairing suggestions. puter or a mobile phone.19 And just like good with answers to their questions, giving demonstrated surgery with an autonomous wine needs time, good grapes require con- advice and recommendations on specific Also on the consumer note, we see AI in robot. The team supervised the robot while it tinuous attention and reliable assessment farm problems and downstream make it wine locating and pricing: Wine Searcher performed soft-tissue surgery, stitching 13 tools. An EU consortium has developed easier for marketers and sales people. (www.winne-searcher.com) is an AI a tool for together a pig’s bowel during open surgery, VineRobot, an ‘Unmanned ground vehicle’ Looking specifically at AI in the wine industry locating and pricing wine, beer and spirits and doing so better than a human surgeon, (UGV) equipped with non-invasive sensor globally and in South Africa it is clear that AI across all online stores. Wine Searcher is also the team claimed. integrating label recognition technology and technology as a convincing alternative to has crept into the wine industry in a number developing a chatbot to improve user interac- manual sampling and analysis.20 The major Companies are applying machine learning to of ways. On the production side of the indus- tion with the site. Yet another AI tool, Vivino, advantage to growers is the availability of a make better and faster diagnoses than try, AI increases production efficiency by (www.vivino.com) uses label recognition large quantity of automatically obtained data humans. One of the best known healthcare optimising machine use. Ailytic, an Australian technology to help guide wine purchases. that any user will be able to interpret easily, technologies is IBM Watson. It understands tech firm, has developed an AI program that Consumers take a photo of the wine label since it is represented on simple maps, as natural language and is capable of respond- uses an AI technique called ‘prescriptive they are considering and are instantly pro- well as the wireless transmission of the infor- ing to questions asked of it. The system analytics’ to account for all the variables that vided the wine’s rating, average price and mation from the vineyards (see figure 2).21 mines patient data and other available data go into mass-producing wines such tem- review from the community of 22 million perature, wine changeover and inventory. users. The app then tracks which wines you The program then creates the best possible scan and rate, but does not at this point offer operation schedule, allowing companies to suggestions based on your profile.17 18 save considerable time and money.14 Along the same lines, Knapp AG with its picking Figure 2: The VineRobot technology is introducing AI into the industry. Source: Science 2.0, 2017 From the market side we see that consumers who are unsure of pairing wine with food now call on the advice and guidance of a new breed of virtual ‘sommeliers’. These tools aim to take the mystery out of picking wine by applying AI. It could be the very modern appli- cation of AI that makes wine and wine selec- tion relevant to today’s consumers. Virtual

10 11 sources to form a hypothesis, which it then disruption of education is yet to arrive how- In finance digital era.26 In finance there will certainly be presents with a confidence scoring scheme. ever. AI can automate grading, assess stu- Given the huge amount of data to be ana- job losses. In 2000, Goldman Sachs’ US Other AI applications include chatbots, a dents and adapt to their needs, helping them lysed and measured, the use of AI (digital cash equities trading desk in New York computer program used online to answer work at their own pace. AI tutors can provide technology and algorithms) has become employed 600 traders. Today, that operation questions and assist customers, to help additional support to student and could mandatory in the finance world. Indeed, AI has two equity traders, with machines doing schedule follow-up appointments or aiding change where and how students learn, per- has infiltrated everything from trading to cus- the rest (see Box 2). haps even replace some teachers. Then the patients through the billing process, and vir- tomer service to regulation, and this domi- way students are taught will change: Educa- tual health assistants that provide basic nance will only become more widespread as BOX 2 tion experts argue that the factory model i.e. medical feedback. Lately there are drones technologies improve and possibilities AI is impacting heavily on peoples’ roles in teachers broadly teaching the same subject that deliver defibrillators and medical supplies increase.25 AI applied to personal finance the financial sector. In an article called to remote site. The possibilities are endless. to all students is no longer sustainable. We “Goldman Sacked” Newsweek says applications, such as Mint or Turbo Tax, is Goldman Sachs and many of the biggest Imagine the time freed up for doctors to need to look into delivering skills for New revolutionising the way interaction with finan- hedge funds are all switching on AI-driven Collar Work.23 A New Collar Worker is typi- reach more accurate diagnosis and allocate cial institutions take place. Such applications systems that can foresee market trends and cally involved in technical (but not special- time for treatment if they no longer have to can collect personal data and provide finan- make trades better than humans. “The ised) work and the skills are self-taught and machines can ingest books, tweets, news rely solely on reading medical journals or find- cial advice. Other programs, e.g. IBM Watson ing treatments. By using cognitive comput- learned but not necessarily through formal reports, financial data, earnings numbers, are applied to the process of buying a home international monetary policy, even Saturday ing, they can start with an understanding of higher education and learning institutions. At and banks use artificial intelligence systems Night Live sketches – anything that might the patient by extracting information from present the most in-demand hard skills in the to organise operations, maintain book-keep- help the software understand global trends”, medical records.22 US are cloud computing expertise, data min- ing, invest in stocks, and manage properties. says the article. “AI can keep watching this ing and statistical analysis, smartphone app information all the time, never tiring, always AI has also reduced fraud and crime by In education development, data storage engineering and learning and perfecting its predictions” and it monitoring behavioural patterns of users for In the past years, software and online service management and user interface design.24 does not wear expensive suits so where any changes or anomalies. does that leave the clothing retailers one have managed to bring changes and reforms Training institutions will have to start deliver- would ask tongue in cheek. Goldman Sachs to classrooms and teaching methods. True ing those skills into the job market. Furthermore, AI is most helpful in the complex shows just how devastating automation can compliance sphere while product, policy and be to traders. procedural complexity is increasing and, along with increasing legislative oversight, the pressure on service staff to become deep In law content specialists is mounting. This has cost implications because these companies are Will AI replace lawyers? Similarly, to finance, required to train and maintain a pool of deep AI in the legal profession has a useful place specialists who cannot be deployed across and while it would not be the end of lawyers, multiple product lines. Further, expecting staff there are a number of potential applications to be walking, talking, legislative and policy for AI in the legal industry notably where they encyclopaedias that never tire and that know relate to the automation of repetitive and all about compliance 100% of the time just is routine tasks to help lawyers provide superior not realistic and distracts them from their legal counsel at a higher level e.g. conducting actual role of providing the customer service legal research, administrative legal support, that is becoming increasingly important in the and performing due diligence. The volumes

12 13 an optimally personalised vehicle purchase competitive advantage the time has now and finance experience. Given the vast selec- come to start discussions around an Arti- tion of cars and finance providers available, ficial Intelligence strategy that involves all machine learning has the potential to help car aspects of business from production, buyers quickly find the vehicles and financing service delivery, marketing, business deci- options that are right for them, vastly simplify- sion-making, data processing, to sales ing their customer journey. and compliance and providing insight and The aforementioned small peek into the infil- foresight. Thoughts should also focus on tration of AI in various industries and sphere long-term investment planning in AI, new of life is just that, a peek. The pace and vast- business models based on improved pro- ness of spectrum that AI develops in all walk ductivity and integrating AI into existing of life is just astounding. Indeed, a new wave digital and analytics plans. Consider devel- of AI has arrived in business. AI is generating oping a company-wide AI capability (rede- new approaches to business models, opera- sign processes, automation, upgrade AI tions, and the deployment of people and skills base) and ensure compliance (poli- resources that are likely to fundamentally cies re data privacy and transparency). change the way business operates. And if it Also consider how you as business leader will approach the introduction and applica- can transform an earthbound industry like tion of AI in the company e.g. if you are of data are just too large to ingest and then bly lines. A Tractica market intelligence study agriculture and in long-standing industries like interested mainly in existing processes, we are not talking about the time it takes. forecasts that the demand for automotive AI automotive and law, how long will it be before reducing costs and improving productivity Automating this process is a better use of hardware, software, and services will explode any company in any industry is affected? The it is perhaps best to become an expert of time and a more efficient process. An AI sys- from US$404 million in 2016 to US$14 billion race for competitive advantage by AI has 28 have access to expert advice on AI tem may provide an opportunity to perform by 2025. As of 2016, there are over 30 begun, in every industry. It is therefore apt to because technology is changing daily. such due diligence in a faster, cheaper, and companies using AI into the creation of driver- provide a few strategic pointers. more thorough fashion instead of relying on a less / self-drive cars. A few companies Innovation, especially with analytics and high-priced and bleary-eyed team of law- involved with AI include Tesla, Google, and POLICY AND STRATEGIC artificial intelligence is a constant challenge yers.27 AI start-ups are already building ques- Apple. We are also seeing significant AI- to be evaluated, re-evaluated and im­­ POINTERS 29 30 tion-and-answer computer assistants that related investment for self-driving cars from proved with every step. AI can multiply intelligence density and lever- can sift programmed-to-answer questions by across the design space. In February • Industries can learn from one another in age of firms far beyond human capabilities examining the taxonomy and ontology asso- 2017, Ford invested US$1 billion in the self- terms of the benefits of AI in certain busi- and this opens up a vast array of opportuni- ciated with a database and certainly it can be driving car start up Argo AI, which was found- ness activities. The wine industry could expected that AI will increasingly play a role in ed by a partnership between two top engi- ties to optimise processes and cost and learn from the automotive industry in terms the legal profession. neers from Google and Uber. On the retail enhance competitiveness. Therefore, for a of customer service and productivity and side there is an equally important role for AI. new strategic agenda is required. It is per- automation; conversely the automotive In automotive The competitive vehicle industry continues to haps opportune to impart with a few policy industry could learn from the financial Advancements in AI have greatly contributed search for new ways to engage customers guidelines and strategic pointers to think industry in terms of AI for customer intel- to the growth of the automotive industry through existing and new channels. AI pro- ‘outside the bottle’ so to speak: ligence and compliance. AI provides a through the creation and evolution of self- vides a completely new set of tools to better • Whereas digital strategy and data analytics completely new set of tools to better driving vehicles and in sophisticated assem- understand customer behaviour and deliver were the tools of businesses to gain a understand issues such as climate pat-

14 15 terns, customer behaviour personalised intelligence will be key skills. That said, the purchase experiences through advanced lack of capable talent skilled in deep learn- analytics. Put to optimal use, AI provides ing technology and analytics may well turn for enhanced data understanding and out to be the biggest obstacle for large therefore better insights and reliable pre- companies. Those managers capable of dictive models for various challenges assessing what the workforce of the future including customer understanding (accu- will look like can prepare themselves for racy, predictive, fast, volumes, customis- the arrival of AI. They should view it as an able). Machine learning methods are par- opportunity to flourish.34 ticularly applicable when it comes to pow- ering new insights within industries in which data sets are large, diverse, and Box 3 change quickly. “AI is the new electricity,” meaning that it will be found everywhere and create new jobs • When looking at talent and skills against that weren’t imaginable before its the background of automation (keeping in appearance.”- Andrew Ng, Baidu Research mind the concept of person bytes) the head and deep learning pioneer same fashion, managers no longer bog­ Some AI tools build shopper profiles by reality is that at least for the foreseeable The lack of capable talent — people skilled in ged down by reporting and administrative accessing previous purchase data, asking deep learning technology and analytics — future, only a few jobs will be totally auto- tasks will be able to focus on executing on shoppers whether they enjoyed those may well turn out to be the biggest obstacle their judgment-oriented skills of creative purchases through an intuitive mobile mated. We are more likely to see certain for large companies. The greatest thinking and experimentation, data analy- application and then applying proprietary activities being automated which in turn opportunities may thus be for independent will impact on the jobs that people per- businesspeople, including farmers like Jeff sis and interpretation, and strategy devel- wine characterizations and learning algo- form. The types of job being redefined and Heepke, (see Box 1) who no longer need opment. These are the skills that will be rithms. With that, you can begin giving scale to compete with large companies, automated to a degree range from low- required to succeed in the future. With accurate, personalised wine suggestions because AI has levelled the playing field. skilled jobs to financial managers, lawyers that their talent to tap into the knowledge immediately while the individual is stand- and CEOs and the impact of AI will and judgment of partners, customers, and ing in that wine aisle trying to identify the increase on business in terms of labour • From a management perspective there will communities to bring together diverse right wine. savings, skills requirements, improved reli- also be a redefinition of the tasks. AI will perspectives, insights, and experiences. • Machine learning algorithms are being ability, higher quality and the ability of soon be able to do the administrative AI will bring new criteria for success: col- integrated into analytics and CRM plat- managers to lead increasingly automated tasks that consume much of managers’ laboration capabilities, information shar- forms to uncover information on how to organisations.31 Professions at greatest time faster, better, and at a lower cost. ing, experimentation, learning and deci- better serve customers. These are typi- risk are telemarketers, loan officers, cash- Reporting and monitoring are other exam- sion-making effectiveness, and the ability cally highly repetitive tasks normally per- iers, paralegal and legal assistants, taxi ples of a manager’s tasks that will be sup- to reach beyond the business / organisa- formed by humans. Chatbots have been 35 drivers and fast food cooks.32 The road ported by AI. A large media compa- tion for insights. incorporated into websites to provide ahead is less about automating individual ny expanded its quarterly earnings report- • Competing for the future will make it immediate service to customers. Indeed, jobs wholesale, than it is about automating ing from approximately 300 stories to imperative for businesses to make optimal chatbots and messaging apps are revolu- the activities within occupations and rede- 4,400 with the help of AI-powered soft- use of data (e.g. collecting purchase his- tionising the future of marketing. Con- fining roles and processes.33 Robots and ware robots. In doing so, technology freed tory with a loyalty program or keeping sumer behaviour has shifted from social people will develop a close and integrated up journalists to conduct more investiga- records of past sales or adjusting produc- networks to messaging platforms such as work relationship while social and creative tive and interpretive reporting. In much the tion processes to environmental factors). SMS, Facebook Messenger, Apple iMes-

16 17 sage, Slack, and WeChat. The growth of ger, Kik, and SMS to talk directly to users. the four largest messaging apps exceeds Imagine how this will impact on marketing ENDNOTES 36 that of the four largest social networks. and sales and distribution pipelines. They 1 www.theverge.com/2017/3/27/15077864/elon-musk-neuralink-brain-computer-interface-ai-cyborgs Companies are now creating bots for can no longer be ignored and should be 2 www.tractica.com/newsroom/press-releases/artificial-intelligence-software-revenue-to-reach-59-8-billion-worldwide- Slack, Amazon Echo, Facebook Messen- part of any marketing strategy workshop. by-2025/ 3 https://curiosity.com/topics/this-defibrillator-drone-could-be-the-robot-that-saves-your-life-curiosity?utm_ campaign=daily-digest&utm_source=sendgrid&utm_medium=email 4 http://searchcio.techtarget.com/definition/AI 5 P Grenier, I Alvarez, J-M. Roger, V. Steinmetz, P Barre and J-M. Sablayrolles, 2002. Artificial Intelligence In Wine- Making. J. Int. Sci. Vigne Vin, 2000, 34:2, 61-66 6 http://searchcio.techtarget.com/definition/AI 7 http://searchcio.techtarget.com/definition/AI 8 www.ubs.com/microsites/artificial-intelligence/en/new-dawn.html 9 http://searchcio.techtarget.com/definition/A 10 www.ft.com/content/d8270fda-152e-11e5-a587-00144feabdc0 11 www.recode.net/2015/4/30/11562024/too-embarrassed-to-ask-what-is-the-cloud-and-how-does-it-work 12 www.newsweek.com/2017/03/10/how-artificial-intelligence-transform-wall-street-560637.html 13 IBM Watson: www.ibm.com/blogs/watson/2016/12/five-ways-agriculture-benefit-artificial-intelligence/ 14 https://phys.org/news/2017-07-artificial-intelligence-boosts-wine-bottom.html#jCp 15 http://thespoon.tech/wine-and-ai-a-perfect-pairing-of-technology-and-tradition/ 16 www.cnbc.com/2016/07/29/artificial-intelligence-and-wine.html 17 http://thespoon.tech/wine-and-ai-a-perfect-pairing-of-technology-and-tradition/ 18 https://mobilebusinessinsights.com/2017/02/thinking-outside-the-bottle-with-analytics-and-artificial-intelligence/ 19 www.farmersweekly.co.za/agri-technology/farming-for-tomorrow/south-africa-builds-its-own-vineyard-robot/ 20 https://phys.org/news/2017-08-wheeled-robot-grape-growth.html#jCp 21 www.science20.com/news_articles/wine_production_now_with_more_robots-152703 22 https://mobilebusinessinsights.com/2017/02/amplifying-human-cognition-with-cognitive-computing/ 23 www.nytimes.com/2017/06/28/technology/tech-jobs-skills-college-degree.html?_r=0 24 www.gnapartners.com/blog/new-collar-job/ 25 www.imperial.ac.uk/business-school/knowledge/finance/artificial-intelligence-is-changing-the-face-of-finance/ 26 www.businesslive.co.za/bd/opinion/2017-08-12-the-link-between-ai-risk-management-and-customer-service/AI in The Wine Industry 27 https://www.lexisnexis.com/lexis-practice-advisor/the-journal/b/lpa/archive/2017/06/07/preparing-for-artificial- intelligence-in-the-legal-profession.aspx 28 https://venturebeat.com/2017/06/24/how-ai-will-play-a-major-role-in-the-auto-industry/ 29 https://mobilebusinessinsights.com/2017/02/thinking-outside-the-bottle-with-analytics-and-artificial-intelligence/ 30 In PwC’s 2017 Digital IQ survey of senior executives worldwide, 54% of the respondents said they were making substantial investments in AI today. But only 20% said their organisations had the skills necessary to succeed with this CONCLUSION technology http://www.pwc.com/us/en/advisory-services/digital-iq.html 31 For the full MGI report, see “Disruptive technologies: Advances that will transform life, business, and the global In conclusion, the lack of capable talent and knowledge of AI (deep learning technology economy,” May 2013, on mckinsey.com. This research has examined the economic potential of disruptive technologies and analytics skills) may well turn out to be the biggest obstacle for businesses. On a that can automate physical work (for example, advanced robotics, 3-D printing, and autonomous vehicles) as well as those that can automate knowledge work requiring intellectual effort and the ability to interact with others (for example, positive note, AI creates immense opportunities for businesses, regardless of size or various types of artificial intelligence, machine learning, and deep learning). industry, to become more competitive because AI has opened the game to all sized 32 www.theguardian.com/us-news/2017/jun/26/jobs-future-automation-robots-skills-creative-health players. Like Farmer Jeff Heepke (Box 1), small businesses no longer need scale to 33 https://mobilebusinessinsights.com/2017/02/thinking-outside-the-bottle-with-analytics-and-artificial-intelligence/ 34 https://hbr.org/2016/11/how-artificial-intelligence-will-redefine-management compete with large companies. But to get there needs deep learning and investment in 35 https://hbr.org/2016/11/how-artificial-intelligence-will-redefine-management the right AI technologies. 36 www.topbots.com/why-chatbots-are-future-marketing/

18 19 vrg rpahics.co.za_0218633165

Compiled, in collaboration with SAWIS, by Dr Marie-Luce Kühn, IBIS Business and Information Services (Pty) Ltd PO Box 7048, Stellenbosch 7599 Tel +27 21 883 2855 e-mail: [email protected] website: www.ibis.co.za

A SAWIS Publication. ©SAWIS, 2017