Trang 1
THETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHETHE EXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDEEXECUTIVE GUIDE TOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTOTO ARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCEARTIFICIAL INTELLIGENCE How to identify and implement applications for AI in your organization ANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESSANDREW BURGESS
Trang 3
Andrew Burgess The Executive Guide to Artificial Intelligence How to identify and implement applications for AI in your organization
Trang 4
Andrew Burgess AJBurgess Ltd London, United Kingdom ISBN 978-3-319-63819-5 ISBN 978-3-319-63820-1 (eBook) https://doi.org/10.1007/978-3-319-63820-1 Library of Congress Control Number: 2017955043 © The Editor(s) (if applicable) and The Author(s) 2018 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or informa- tion storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Cover illustration: Ukususha/iStock/Getty Images Plus Printed on acid-free paper This Palgrave Macmillan imprint is published by Springer Nature The registered company is Springer International Publishing AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Trang 6
Foreword I remember well the AI work I did whilst at college studying computer science, how different and fascinating it was and still is. We were set a very open challenge to write an AI programme on any subject. I decided to write mine so that it could tell you if the building in a photo was a house, a flat or a bungalow. Somewhat impractical, but a great learning experi- ence for me, particularly in understanding how AI is different from tradi- tional software. Although my college days were a number of years ago, since that time the concept of computers learning has always intrigued me and I have since won- dered how long it will take for AI to have a truly widespread impact. In recent years, we’ve seen massive improvements in processing power, big data collec- tion via sensors and the Internet of Things, cloud services, storage, ubiquitous connectivity and much more. These technological leaps mean that this is the right time for AI to become part of the ‘here and now’ and I strongly believe we will see a dramatic increase in the use of AI over the next few years. The AI in use today is known as narrow AI because it can excel at thousands of relatively narrow tasks (e.g. doing internet searches, playing Go or looking for fraudulent transactions). Things will certainly get even more exciting when ‘general AI’ can outperform humans at nearly every task we do, but we simply don’t know when this might be, or what the world will then look like. Until then, what excites me most is how we can apply AI now to solve our day-to- day problems at home and work. So why is AI important and how can we use it? Firstly, if you are impatient (like I am), doing small manual, repetitive tasks on computers simply takes too much time. I want the computer to do a better job of anticipating my needs and to just get on with it. If I could, I would prefer to talk to Alexa or vii
Trang 7
viii Foreword Google Assistant and just tell the computer what to do. For example, I would love to be able to ask Alexa to buy the most convenient train ticket and take the money out of my account. Compare this to buying a train ticket on any website, where after something like 50 key strokes you might have bought a ticket. I don’t think future generations, who are becoming increasingly impa- tient, will put up with doing these simple and time-consuming tasks. I see my children and future generations having more ‘thinking time’ and focusing on things that are outside the normal tasks. AI may in fact free up so much of my children’s time that they can finally clean up their bedrooms. In the workplace, how many of the emails, phone calls and letters in a call centre could be handled by AI? At Virgin Trains, we used AI to reduce the time spent dealing with customer emails by 85% and this enabled our people to focus on the personable customer service we’re famous for. Further improvements will no doubt be possible in the future as we get better at developing conversational interfaces, deep learning and process automation. One can imagine similar developments revolutionising every part of the business, from how we hire peo- ple to how we measure the effectiveness of marketing campaigns. So, what about the challenges of AI? One that springs to mind at Virgin is how to get the ‘tone of voice’ right. Our people are bold, funny and empa- thetic, and our customers expect this from us in every channel. Conversational interfaces driven by AI should be no different. Today it may be a nuisance if your laptop crashes, but it becomes all the more important that an AI system does what you want it to do if it controls your car, your airplane or your pacemaker. With software systems that can learn and adapt, we need to understand where the responsibility lies when they go wrong. This is both a technical and an ethical challenge. Beyond this, there are questions about data privacy, autonomous weapons, the ‘echo cham- ber’ problem of personalised news, the impact on society as increasing num- bers of jobs can be automated and so on. Despite these challenges, I am incredibly excited about the future of tech- nology, and AI is right at the heart of the ‘revolution’. I think over the next five to ten years AI will make us more productive at work, make us more healthy and happy at home, and generally change the world for the better. To exploit these opportunities to the full, businesses need people who understand these emerging technologies and can navigate around the chal- lenges. This book is essential reading if you want to understand this transfor- mational technology and how it will impact your business. John Sullivan, CIO and Innovation at Virgin Trains
Trang 8
Acknowledgements I would like to thank the following people for providing valuable input, con- tent and inspiration for this book: Andrew Anderson, Celaton Richard Benjamins, Axa Matt Buskell, Rainbird Ed Challis, Re:infer Karl Chapman, Riverview Law Tara Chittenden, The Law Society Sarah Clayton, Kisaco Research Dana Cuffe, Aldermore Rob Divall, Aldermore Gerard Frith, Matter Chris Gayner, Genfour Katie Gibbs, Aigen Daniel Hulme, Satalia Prof. Mary Lacity, University of Missouri-St Louis Prof. Ilan Oshri, Loughborough University Stephen Partridge, Palgrave Mike Peters, Samara Chris Popple, Lloyds Bank John Sullivan, Virgin Trains Cathy Tornbaum, Gartner ix
Trang 9
x Acknowledgements Vasilis Tsolis, Congnitiv+ Will Venters, LSE Kim Vigilia, Conde Naste Prof. Leslie Willcocks, LSE Everyone at Symphony Ventures
Trang 10
Contents 1 Don’t Believe the Hype 1 2 Why Now? 11 3 AI Capabilities Framework 29 4 Associated Technologies 55 5 AI in Action 73 6 Starting an AI Journey 91 7 AI Prototyping 117 8 What Could Possibly Go Wrong? 129 9 Industrialising AI 147 10 Where Next for AI? 165 Index 177 xi
Trang 11
List of Figures Fig. 2.1 Basic neural network 21 Fig. 2.2 Training a neural network 21 Fig. 2.3 A trained neural network 22 Fig. 3.1 AI objectives 30 Fig. 3.2 Knowledge map 43 Fig. 3.3 The AI framework 51 Fig. 4.1 Human in the loop 68 Fig. 4.2 Training through a human in the loop 69 Fig. 4.3 Crowd-sourced data training 69 Fig. 6.1 Aligning with the business strategy 93 Fig. 6.2 AI maturity matrix 100 Fig. 6.3 AI heat map first pass 102 Fig. 6.4 AI heat map 104 Fig. 9.1 AI Eco-System 148 xiii
Trang 12
1 Don’t Believe the Hype Introduction Read any current affairs newspaper, magazine or journal, and you are likely to find an article on artificial intelligence (AI), usually decrying the way the ‘robots are taking over’ and how this mysterious technology is the biggest risk to humanity since the nuclear bomb was invented. Meanwhile the companies actually creating AI applications make grand claims for their technology, explaining how it will change peoples’ lives whilst obfuscating any real value in a mist of marketing hyperbole. And then there is the actual technology itself—a chimera of mathematics, data and computers—that appears to be a black art to anyone outside of the developer world. No wonder that business executives are confused about what AI can do for their business. What exactly is AI? What does it do? How will it benefit my business? Where do I start? All of these are valid questions that have been, to date, unanswered, and which this book seeks to directly address. Artificial Intelligence, in its broadest sense, will have a fundamental impact on the way that we do business. Of that there is no doubt. It will change the way that we make decisions, it will enable completely new business models to be created and it will allow us to do things that we never before thought pos- sible. But it will also replace the work currently being done by many knowl- edge workers, and will disproportionally reward those who adopt AI early and effectively. It is both a huge opportunity and an ominous threat wrapped up in a bewildering bundle of algorithms and jargon. But this technological revolution is not something that is going to happen in the future; this is not some theoretical exercise that will concern a few businesses. © The Author(s) 2018 1 A. Burgess, The Executive Guide to Artificial Intelligence, https://doi.org/10.1007/978-3-319-63820-1_1
Trang 13
2 1 Don’t Believe the Hype Artificial Intelligence is being used today in businesses to augment, improve and change the way that they work. Enlightened executives are already working out how AI can add value to their businesses, seeking to understand all the different types of AI and working out how to mitigate the risks that it inevitably brings. Many of those efforts are hidden or kept secret by their instigators, either because they don’t want the use of AI in their products or services to be widely known, or because they don’t want to give away the competitive advantage that it bestows. A persistent challenge for executives that want to get to grips with AI is where to find all the relevant information without resorting to fanciful articles, listening to vendor hyperbole or trying to understand algorithms. AI is firmly in the arena of ‘conscious unknowns’—we know that we don’t know enough. People generally experience AI first as consumers. All our smartphones have access to sophisticated AI, whether that is Siri, Cortana or Google’s Assistant. Our homes are now AI enabled through Amazon’s Alexa and Google Home. All of these supposedly make our lives easier to organise, and generally they do a pretty good job of it. But their use of AI is actually pretty limited. Most of them rely on the ability to turn your speech into words, and then those words into meaning. Once the intent has been established, the rest of the task is pretty standard automation; find out the weather forecast, get train times, play a song. And, although the speech recognition and natural lan- guage understanding (NLU) capabilities are very clever in what they achieve, AI is so much more than that, especially in the world of business. Artificial Intelligence can read thousands of legal contracts in minutes and extract all the useful information out of them; it can identify cancerous tumours with greater accuracy than human radiologists; it can identify fraud- ulent credit card behaviour before it happens; it can drive cars without drivers; it can run data centres more efficiently than humans; it can predict when customers (and employees) are going to desert you and, most importantly, it can learn and evolve based on its own experiences. But, until business executives understand what AI is, in simple-enough terms, and how it can help their business, it will never reach its full potential. Those with the foresight to use and exploit AI technologies are the ones that need to know what it can do, and understand what they need to do to get things going. That is the mission of this book. I will, over the course of the ten chapters, set out a framework to help the reader get to grips with the eight core capabilities of AI, and relate real business examples to each of these. I will provide approaches, methodologies and tools so that you can start your AI journey in the most efficient and effective way. I will also draw upon inter- views and case studies from business leaders who are already implementing AI, from established AI vendors, and from academics whose work focuses on the practical application of AI.
Trang 14
Introducing the AI Framework 3 Introducing the AI Framework My AI Framework was developed over the past few years through a need to be able to make sense of the plethora of information, misinformation and marketing-speak that is written and talked about in AI. I am not a computer coder or an AI developer, so I needed to put the world of AI into a language that business people like myself could understand. I was continually frus- trated by the laziness in the use of quite specific terminology in articles that were actually meant to help explain AI, and which only made people more confused than they were before. Terms like Artificial Intelligence, Cognitive Automation and Machine Learning were being used interchangeably, despite them being quite different things. Through my work as a management consultant creating automation strate- gies for businesses, through reading many papers on the subject and speaking to other practitioners and experts, I managed to boil all the available informa- tion down into eight core capabilities for AI: Image Recognition, Speech Recognition, Search, Clustering, NLU, Optimisation, Prediction and Understanding. In theory, any AI application can be associated with one or more of these capabilities. The first four of these are all to do with capturing information—getting structured data out of unstructured, or big, data. These Capture categories are the most mature today. There are many examples of each of these in use today: we encounter Speech Recognition when we call up automated response lines; we have Image Recognition automatically categorising our photographs; we have a Search capability read and categorise the emails we send complaining about our train being late and we are categorised into like-minded groups every time we buy something from an online retailer. AI efficiently captures all this unstructured and big data that we give it and turns it into something useful (or intrusive, depending on your point of view, but that’s a topic to be discussed in more detail later in the book). The second group of NLU, Optimisation and Prediction are all trying to work out, usually using that useful information that has just been captured, what is happening. They are slightly less mature but all still have applications in our daily lives. NLU turns that speech recognition data into something useful—that is, what do all those individual words actually mean when they are put together in a sentence? The Optimisation capability (which includes problem solving and planning as core elements) covers a wide range of uses, including working out what the best route is between your home and work. And then the Prediction capability tries to work out what will happen next— if we bought that book on early Japanese cinema then we are likely to want to buy this other book on Akira Kurosawa.
Trang 15
4 1 Don’t Believe the Hype Once we get to Understanding, it’s a different picture all together. Understanding why something is happening really requires cognition; it requires many inputs, the ability to draw on many experiences, and to con- ceptualise these into models that can be applied to different scenarios and uses, which is something that the human brain is extremely good at, but AI, to date, simply can’t do. All of the previous examples of AI capabilities have been very specific (these are usually termed Narrow AI) but Understanding requires general AI, and this simply doesn’t exist yet outside of our brains. Artificial General Intelligence, as it is known, is the holy grail of AI researchers but it is still very theoretical at this stage. I will discuss the future of AI in the concluding chapter, but this book, as a practical guide to AI in business today, will inherently focus on those Narrow AI capabilities that can be implemented now. You will already be starting to realise from some of the examples I have given already that when AI is used in business it is usually implemented as a combination of these individual capabilities strung together. Once the indi- vidual capabilities are understood, they can be combined to create meaningful solutions to business problems and challenges. For example, I could ring up a bank to ask for a loan: I could end up speaking to a machine rather than a human, in which case AI will first be turning my voice into individual words (Speech Recognition), working out what it is I want (NLU), deciding whether I can get the loan (Optimisation) and then asking me whether I wanted to know more about car insurance because people like me tend to need loans to buy cars (Clustering and Prediction). That’s a fairly involved process that draws on key AI capabilities, and one that doesn’t have to involve a human being at all. The customer gets great service (the service is available day and night, the phone is answered straight away and they get an immediate response to their query), the process is efficient and effective for the business (operating costs are low, the decision making is consistent) and revenue is potentially increased (cross-selling additional products). So, the combining of the indi- vidual capabilities will be key to extracting the maximum value from AI. The AI Framework therefore gives us a foundation to help understand what AI can do (and to cut through that marketing hype), but also to help us apply it to real business challenges. With this knowledge, we will be able to answer questions such as; How will AI help me enhance customer service? How will it make my business processes more efficient? And, how will it help me make better decisions? All of these are valid questions that AI can help answer, and ones that I will explore in detail in the course of this book.
Trang 16
The Impact of AI on Jobs 5 Defining AI It’s interesting that in most of the examples I have given so far people often don’t even realise they are actually dealing with AI. Some of the uses today, such as planning a route in our satnav or getting a phrase translated in our browser, are so ubiquitous that we forget that there is actually some really clever stuff happening in the background. This has given rise to some tongue- in-cheek definitions of what AI is: some say it is anything that will happen in 20 years’ time, others that it is only AI when it looks like it does in the movies. But, for a book on AI, we do need a concise definition to work from. The most useful definition of AI I have found is, unsurprisingly, from the Oxford English Dictionary, which states that AI is “the theory and develop- ment of computer systems able to perform tasks normally requiring human intelligence”. This definition is a little bit circular since it includes the word ‘intelligence’, and that just raises the question of what is intelligence, but we won’t be going into that philosophical debate here. Another definition of AI which can be quite useful is from Andrew Ng, who was most recently the head of AI at the Chinese social media firm, Baidu, and a bit of a rock star in the world of AI. He reckons that any cognitive process that takes a human under one second to process is a potential candidate for AI. Now, as the technologies get better and better this number may increase over time, but for now it gives us a useful benchmark for the capabilities of AI. Another way to look at AI goes back to the very beginnings of the technol- ogy and a very fundamental question: should these very clever technologies seek to replace the work that human beings are doing or should they augment it? There is a famous story of the two ‘founders’ of AI, both of whom were at MIT: Marvin Minsky and Douglas Engelbart. Minsky declared “We’re going to make machines intelligent. We are going to make them conscious!” To which Engelbart reportedly replied: “You’re going to do all that for the machines? What are you going to do for the people?” This debate is still raging on today, and is responsible for some of those ‘robots will take over the world’ headlines that I discussed at the top of this chapter. The Impact of AI on Jobs It is clear that AI, as part of the wider automation movement, will have a severe impact on jobs. There are AI applications, such as chatbots, which can be seen as direct replacements for call centre workers. The ability to read thou- sands of documents in seconds and extract all the meaningful information
Trang 17
6 1 Don’t Believe the Hype from them will hollow out a large part of the work done by accountants and junior lawyers. But equally, AI can augment the work done by these groups as well. In the call centre, cognitive reasoning systems can provide instant and intuitive access to all of the knowledge that they require to do their jobs, even if it is their first day on the job—the human agent can focus on dealing with the customer on an emotional level whilst the required knowledge is provided by the AI. The accountants and junior lawyers will now have the time to prop- erly analyse the information that the AI has delivered to them rather than spend hours and hours collating data and researching cases. Whether the net impact on work will be positive or negative, that is, will automation create more jobs than it destroys, is a matter of some debate. When we look back at the ‘computer revolution’ of the late twentieth century that was meant to herald massive increases in productivity and associated job losses, we now know that the productivity benefits weren’t as great as people predicted (PCs were harder to use than first imagined) and the computers actually generated whole new industries themselves, from computer games to movie streaming. And, just like the robots of today, computers still need to be designed, manufactured, marketed, sold, maintained, regulated, fixed, fuelled, upgraded and disposed of. The big question, of course, is whether the gains from associated activities plus any new activities created from automation will outweigh the loss of jobs that have been replaced. I’m an optimist at heart, and my own view is that we will be able to adapt to this new work eventually, but not before going through a painful transition period (which is where a Universal Basic Income may become a useful solution). The key factor therefore is the pace of change, with all the indicators at the moment suggesting that the rate will only get faster in the coming years. It’s clear that automation in general, and AI in particular, are going to be huge disruptors to all aspects of our lives—most of it will be good but there will be stuff that really challenges our morals and ethics. As this book is a practical guide to implementing AI now, I’ll be exploring these questions in a little more detail toward the end of the book, but the main focus is very much on the benefits and challenges of implementing AI today. A Technology Overview The technology behind AI is fiendishly clever. At its heart, there are algo- rithms: an algorithm is just a sequence of instructions or a set of rules that are followed to complete a task, so it could simply be a recipe or a railway time- table. The algorithms that power AI are essentially very complicated statistical
Trang 18
A Technology Overview 7 models—these models use probability to help find the best output from a range of inputs, sometimes with a specific goal attached (‘if a customer has watched these films, what other films would they also probably like to watch?’). This book is certainly not about explaining the underlying AI technology; in fact, it is deliberately void of technology jargon, but it is worth explaining some of the principles that underpin the technology. One of the ways that AI technologies are categorised is between ‘supervised’ and ‘unsupervised’ learning. Supervised learning is the most common approach out of the two and refers to situations where the AI system is trained using large amounts of data. For example, if you wanted to have an AI that could identify pictures of dogs then you would show it thousands of pictures, some of which had dogs and some of which didn’t. Crucially, all the pictures would have been labelled as ‘a dog picture’ or ‘not a dog picture’. Using machine learning (an AI technique which I’ll come on to later) and all the training data the system learns the inherent characteristics of what a dog looks like. It can then be tested on another set of similar data which has also been tagged but this time the tags haven’t been revealed to the system. If it has been trained well enough, the system will be able to identify the dogs in the pic- tures, and also correctly identify pictures where there is no dog. It can then be let loose on real examples. And, if the people using your new ‘Is There a Dog in My Picture?’ app are able to feed back when it gets it right or not, then the system will continue to learn as it is being used. Supervised learning is gener- ally used where the input data is unstructured or semi-structured, such as images, sounds and the written word (Image Recognition, Speech Recognition and Search capabilities in my AI Framework). With Unsupervised Learning, the system starts with just a very large data set that will mean nothing to it. What the AI is able to do though is to spot clusters of similar points within the data. The AI is blissfully naive about what the data means; all it does is look for patterns in vast quantities of numbers. The great thing about this approach is that the user can also be naive—they don’t need to know what they are looking for or what the connections are—all that work is done by the AI. Once the clusters have been identified then pre- dictions can be made for new inputs. So, as an example, we may want to be able to work out the value of a house in a particular neighbourhood. The price of a house is dependent on many variables such as location, number of rooms, number of bathrooms, age, size of garden and so on, all of which make it difficult to predict its value. But, surely there must be some complicated connection between all of these vari- ables, if only we could work it out? And that’s exactly what the AI does. If it is fed enough base data, with each of those variables as well as the actual price,
Trang 19
8 1 Don’t Believe the Hype then it uses statistical analysis to find all the connections—some variables may be very strong influencers on price whilst others may be completely irrelevant. You can then input the same variables for a house where the price is unknown and it will be able to predict that value. The data that is input is structured data this time, but the model that is created is really a black box. This appar- ent lack of transparency is one of AI’s Achilles’s heels, but one that can be managed, and which I’ll discuss later in the book. As well as the above two types of training, there are various other terms associated with AI, and which I’ll cover briefly here, although for business executives they only need to be understood at a superficial level. ‘Neural Networks’ is the term used to describe the general approach where AI is mim- icking the way that the brain processes information—many ‘neurons’ (100 billion in the case of the brain) are connected to each other in various degrees of strength, and the strength of the connection can vary as the brain/machine learns. To give an over-simplified example, in the dog picture app above, the ‘black nose’ neuron will have a strong influence on the ‘dog’ neuron, whereas a ‘horn’ neuron will not. All of these artificial neurons are connected together in lay- ers, where each layer extracts an increasing level of complexity. This gives rise to the term Deep Neural Networks. Machine Learning, where the machine creates the model itself rather than a human creating the code (as in the exam- ples I have given above), uses DNNs. So, think of these terms as concentric circles: AI is the over-arching technology, of which Machine Learning is a core approach that is enabled by DNNs. There are obviously many more terms that are in common use in the AI world, including decision tree learning, inductive logic programming, rein- forcement learning and Bayesian networks, but I will cover these only when absolutely necessary. The focus of this book, as you will now hopefully appreciate, is on the business application of AI rather than its underlying technologies. About This Book My working experience has been as a management consultant, helping organ- isations cope with the challenges of the time, from productivity improvement, through change management and transformation to outsourcing and robotic process automation, and now AI. I first came across AI properly in my work in 2001. I was working as Chief Technology Officer in the Corporate Venturing unit of a global insurance provider—my role was to identify new technologies
Trang 20
About This Book 9 that we could invest in and bring into the firm as a foundation client (it was what we used to call the ‘incubator’ model). One of the technologies we invested in was based around the idea of ‘smart agents’ that could be used as an optimisation engine—each agent would have a specific goal and would ‘negotiate’ with the other agents to determine the ideal collective outcome. So, for example, the system could determine the most effective way for trucks to pass through a port, or the best way to generate the most revenue from the size and arrangement of advertisements in newspapers. Although we didn’t call it AI at the time, this is effectively what it was—using computer algorithms to find optimal solutions to real problems. Fast forward to 2017, and my work is focused almost exclusively on AI. I work with enterprises to help create their AI strategy—identifying opportuni- ties for AI, finding the right solution or vendor and creating the roadmap for implementation. I don’t do this as a technologist, but as someone that under- stands the capabilities of AI and how those capabilities might address business challenges and opportunities. There are plenty of people much cleverer than me that can create the algorithms and design the actual solutions, but those same people rarely understand the commerciality of business. I see myself as a ‘translator’ between the technologists and the business. And with AI, the chal- lenge of translating the technology is orders of magnitude greater than with traditional IT. Which is why I wanted to write this book—to bring that understanding to where it can be used the best: on the front line of business. So, this is not a book about the theoretical impact of AI and robots in 10 or 20 years’ time and it is certainly not a book about how to develop AI algo- rithms. This is a book for practitioners of AI, people who want to use AI to make their businesses more competitive, more innovative and more future- proof. That will happen only if the business leaders and executives understand what the capabilities of the technology are, and how it can be applied in a practical way. That is the mission of this book: be as informed as possible about AI, but without getting dragged down by the technology, so that you can make the best decisions for your business. And it is also a heartfelt appeal: whatever you read or hear about AI, don’t believe the hype.
Hiển thị 20 trang đầu. Tải tài liệu để xem đầy đủ.
Bình luận (0)
Chưa có bình luận nào. Hãy là người đầu tiên!