Everything you ever wanted to know about: Intelligent Automation Artificial Intelligence Digital Labour 4th Industrial Revolution But were to afraid to ask 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 1 Document Classification: KPMG Public
Thinking forward 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 2 Document Classification: KPMG Public
What we ll be talking about The context: The World is accelerating Being Ray Kurzweil Narrow v General AI Ethics Examples and Data Question time? The future is already here. It s just not evenly distributed Jonny Mnemonic 1981 4 Document Classification: KPMG Confidential
The Signals in the Noise = 2017 Predictions Predictions clustered and ranked by topic areas 3,000+ articles & papers from 72 publications* ** 92 articles; 18 topic areas * Excluding economic predictions ** Sources: Quid, Thomson Reuters, Google Alerts, New Times and Financial Times 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 4 Document Classification: KPMG Public
Disruption is impacting all of us 8 Document Classification: KPMG Confidential
Why do we care? Top 10% of firms = 80% profit 3.5% per annum 5% per annum 20% 80% 1.7% per annum Source: Frontier Firms, Technology Diffusion and Public Policy, OECD, 2015 0.3% per annum 10% = Frontier firms 90% = Lagging firms Source: McKinsey 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 6 Document Classification: KPMG Public
The acceleration continues.. Being Ray Kurzweil Total power of all computers is comparable to total brainpower of the human race Ray Kurzweil Futurologist 2019 2016 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 57 Document Classification: KPMG KPMG Confidential Public
And they talk back. Most business transactions or information queries are dealt with by a simulated person. Language Translation machines are routine. Q 2019 A 6 Document Classification: KPMG Confidential
AI wants recognition US$1,000 personal computer is 1,000 times more powerful than the human brain and AI claims to be conscious and petitions for recognition 2029 7 Document Classification: KPMG Confidential
Goodbye Economic Theory.. Hello Neural Links Singularity. $1k buys 1 billion times power of the human brain. AI is smartest and most capable life form on the planet 2045 8 Document Classification: KPMG Confidential
Maturity model for the use of AI The world's largest technical professional organization for the advancement of technology Levels of automation of decision and action selection (a) Low 1 2 3 4 5 6 Allows the human a restricted veto time before automatic execution Executes that suggestion if the human approves Suggests one alternative 7 Narrows the selection down to a few 8 Informs the human only if it, the computer, decides to Informs the human only if asked Executes automatically, then necessarily informs the human The computer offers a complete set of decision/ action alternatives High The computer offers no assistance, human must take all decisions and actions 9 10 The computer decides everything, acts autonomously, ignores the human Simple four-stage model of human information processing (a) Sensory processing Perception/ working memory Decision making Response selection Source: (a) Parasuraman et al.: Types and levels of human interaction with automation. 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 11 Document Classification: KPMG Public
Strategically assume narrow super intelligence exists Application of deep neural networks to Atari 2600 Games Different forms of AI Facial Recognition e.g. GaussianFace Limited Domain Games e.g. GO Narrow Intelligence Weak AI General Intelligence Strong AI Source: Nature (2014), Susskind (definition of narrow vs broad AI) All it "knows" is that it wants to maximize the score, and by watching the game carefully and observing which actions increase that score, it learns how to play then how to play better. 10 Document Classification: KPMG Confidential
DeepMind Breakout 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 13 Document Classification: KPMG Public
Ethical and moral considerations Microsoft s experience with ChatBots demonstrates the importance of data curation 50% TMOD 40 million registered users Xiaoice have said I love you 25% 11 Document Classification: KPMG Confidential
Ethical and moral considerations Microsoft s experience with ChatBots demonstrates the importance of data curation 50% TMOD Suspended after 16 hours Tay Will be re-released when it can be made safe 12 Document Classification: KPMG Confidential
AI technologies have evolved to address the challenges of today s digital world, and have the potential to reshape the workforce of the future. Cognitive systems mimic human brain functions 1 2 3 Natural language processing Artificial intelligence Deep learning Perceive (interpret sensory input beyond traditional data) Reason (hypothesise, weigh supporting evidence) Learn (improve confidence levels with experience) Machine learning Text analytics Predictive analytics Image recognition Voice recognition Intelligent augmentation A new partnership between humans and machines Physician Nurse Wealth Advisor Insurance Advisor Academic Advisor Sales Advisor Legal Advisor Tax Advisor Auditor Service Advisor 2016 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 15 16 Document Classification: KPMG KPMG Confidential Public
Robotics will be adopted wherever high labour costs are $152 billion Global market for robots and AI expected to reach $152.7 billion by 2020. Adoption could improve productivity by 30%. Bank of America Merrill Lynch. Recent research from London School of Economics suggests a ROI in robotic technologies of between 600% and 800% for specific tasks. 600% TO 800% ROI 2016 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ) $14.9 billion According to Quid, from 2010 to 2014, private investment in AI has grown from $1.7 billion to $14.9 billion, and was on track to grow nearly 50% year-on-year in 2015 alone. a Swiss entity. All rights reserved. $1.7 billion 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ),, a Swiss entity. All rights reserved. 40 17 Document Classification: KPMG Public
Digital technology: here and now 300K COIN JPMorgan Hours of legal word reviewing financial deals after COntract INtelligence AI system JD.com robot-run warehouse in Shanghai Person 1 200k Packages processed every hour 500 employed where Previously Using robotic unloading systems, barcode scanning, automated vans and even drone deliveries FOXCONN Robots 60,000 Workers replaced 500k FTE s to be replaced by 1 Million robots 50% Big 4 firm Big 4 firm has announced a reduction of at least 50% in graduate recruitment KPMG Loans Tool 100% Of a bank s loan book can now be reviewed and reassessed by AI Tesla Level 5 by December 2017 222m autopilot miles driven (10/2017) 1.3 Bn Level 5 Total number of training miles By Dec 2017 Autodesk Generative Design In: Constraints eg weight, cost etc Out: Optimised design 15m Bank of England Jobs have potential to be automated in the UK out of 34m 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 18 Document Classification: KPMG Public
Governance of AI Governance Development Operations Content Do you know where AI is being used, whether it is under central control, and has the AC been briefed on the risks? Who in the organisation is responsible for the use of AI? What is the right governance structure regarding accuracy standards (e.g. is 97% the right answer?) What is the impact on organisation, people and skills in the context of benefit realisation? How are company s values and culture embedded within the AI solution? How do you manage and value externally supplied AI engines learning from your data? Is secure by design still possible? What are the controls to define AI s ability to learn? How does the business test the solutions and decide and approve it for release? What evidence is available of controls operating during the design and built phases? How do you ensure consistent quality, unconscious bias of teachers as well as training data? What controls and QA checks are built in and around the design of the AI engine? How is quality of all data sources ensured, including external ones? What evidence is available of controls operating during operations? Have KPIs been defined to indicate quality of output? For ethical and moral considerations? And for achieving benefits? How are changes to the algorithms and learning data controlled? How to deal with robots supervising robots, or how do you mediate when robots disagree? To what extent is the ability for human override allowed or controlled? Have the continuity aspects been covered, e.g. what if the solution goes offline and staff no longer have the skills? How is quality of the content ensured during development as well as during ongoing selflearning? How are you auditing an automated rule that nobody has designed or built? How is completeness and reliability ensured over dynamically changing content? Do you have test data sets available to run through the solution to validate output? Do you have a content strategy for starting to collect quality and unbiased data? Do you have alternative mechanisms in place to replicate the AI solutions output for further validation? 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 19 Document Classification: KPMG Public
Thank You
Thank you! 2017 KPMG LLP, a UK limited liability partnership and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative ( KPMG International ), a Swiss entity. All rights reserved. 21 Document Classification: KPMG Public
Digital is rapidly changing our expectations and behaviours of Millennials have a cellphone (versus 90% 94% of adults) 76% 54% Everything as a service Successful collaborative economy examples have a smart phone (versus 64% of adults) said they would prefer to pay for purchases using their smartphone as a mobile wallet (2) The Collaborative Economy 1 Consumer Appeal Cost-effective. Convenient. Easy. Emotionally Fulfilling. Trends Fuelling Success Economic volatility. Mass urbanisation. Ubiquitous connectivity. Sustainability. Millennials want the hassle-free experience of online ordering and free delivery (3) Everything on demand Access to digital assets are also driving increased subscription payment volumes 45% of Increase in number subscribers (2) 62million users (3) Sources (1) Forrester (2) Fortune (3) Statista 23 Document Classification: KPMG Confidential
Case study: Future applications of analytics, machine learning & AI 30 Document Classification: KPMG Confidential
Case study: The problem they re now solving Ocado s 4D optimisation problem Aim is to make sure that not only are the right totes in the right place across a three dimensional grid, but also making sure they are where they need to be at the right time Robots swarm together across the hive to access lower totes Optimisation of moving full totes back into the hive is based on looking at the in-day order well (time being the 4 th dimension) Robots built to use what is essentially 4G technology (rather than WIFI), enabling them to check in ten times a second AI used to maintain the army of robots, using machine learning to spot signs of wear and tear in the analytics data collected in the firmware of each machine Considering using drones to fly overhead for visual inspections 31 Document Classification: KPMG Confidential
Case study: Future developments and innovations 32 Document Classification: KPMG Confidential
Digital Supply Chain: digital trends & enablers Cloud Services and Platforms Analytics and Data Science Mobility solutions & role specific Apps Social Media and collaborative technologies Applications, services and data storage delivered over the network or via secure internet cloud servers. E.g. logistics and planning control tower software, powered procurement software Big data and analytics ranging from data visualisation to simulation (predictive and prescriptive analytics). Predictive modeling technologies will incorporate decision support recommendations to users e.g. predictive maintenance. Workers and transport are connected /provided with internet enabled portable devices and mobile apps to process real-time info and provide MI, analytics, mapping solutions, instructions etc. Crowdsourcing technologies; Enterprise collaboration technologies; Social media platforms. Cyber Security Internet of Things and Sensors Autonomous Vehicles and UAVs Robotics Encrypted data and protection mechanisms against cyber threats. Embedded sensors and actuators in machines or other physical objects e.g. smart products. Material movements and customer delivery is handled via autonomous vehicle (level 4 or 5 autonomy), unmanned aerial vehicles and intelligent transportation units. Use of flexible robots/cobots augments operatives, automates certain processes and creates new forms of worker-robot interaction. Include bionic enhancement? Robotic Process Automation, Cognitive technologies & AI Blockchain Additive Manufacturing Augmented Reality & Virtual Reality RPA: Replacing labour-based processing with machine-based, at dramatically reduced costs. Cognitive & AI: Automating decision-making process; natural language processing; pattern recognition, hypothesis generation. A distributed ledger to order and verify transactions which is immutable, decentralised and secure. Blocks are transparent and are validated by network consensus. Can be public or private. 3D Printing enables personalisation, rapid prototyping, spare part printing and near shoring of manufacturing. Will be combined with existing and emerging materials e.g. nanotechnology Augmented reality headsets to aid maintenance activities, assembly operations, picking activity. Virtual reality to enable training within virtual plants and warehouses 33 Document Classification: KPMG Confidential
Automation and cobotics (collaborative robotics) will be prevalent in large & small manufacturing and distribution 41
Autonomous vehicles the road to autonomy TESLA Level 2 Level 0 Human driver does everything Level 1 Vehicle can sometimes assist the human driver Autopilot 2015 MERCEDES Drive Pilot 2016 CADILLAC Super Cruise 2017 VOLVO Drive Me 2017 Level 2 Vehicle can conduct some parts of the driving task, human monitors Level 3 Vehicle can conduct some parts of the driving task, human can check out EU Accident Free 2050 KIA L4 Autonomy 2030 GM, GOOGLE L3/L4 Autonomy 2025 FORD L4 Autonomy 2020 VOLVO Accident Free 2020 Level 4 Vehicle can conduct all of the driving task in certain environments Level 5 Vehicle can perform all driving tasks, under all conditions Source: Brian Cooley, CNET 42
Autonomous vehicles Applications in Supply Chain Perth Operations Centre Rio Tinto: x3 Pilbara iron ore mines 43
The factory of the future Production line is flexible and can change products with minimal downtime through greater machine understanding Cyber-physical systems connect a network of machines to provide a whole line view of performance Machines provide real time data on performance against acceptable tolerances A central Mission Control has visibility of the production line and incorporates augmented reality devices to aid operator activity Data d riven d e m a n d p r ediction r e sults in minimal stock held Breakdowns are anticipated before they occur and maintenance is not constrained to a schedule but to real time information Logistics planning is tied into Mission Control and is optimised to balance customer requirements and cost effectiveness Goods-in stock is kept at a minimum to ensure the line is kept running at minimal cost Minimal staffing on the shop floor staff know where they are needed, when they are needed, and what they are needed for 44
Bi-modal Supply Chain: companies will need to innovate Mode 1 is Predictable: Stability Efficiency Continuous Improvement Mode 2 is Exploratory: Risky Fail Fast Transformative Gartner predictions: Source: Gartner 2018: The deployment of end-to-end supply chain visibility solutions will increase up to 50%. 2020: 10% of warehouse operations will have smart, self-navigating, autonomous vehicles operating beside humans. 45