Month: January 2017


How AI and Machine Learning are Reshaping Industries

Artificial Intelligence is at a breakthrough point. Investment is at a record high, and adoption rates are on the rise. Success stories are going viral across social media platforms, boosting confidence in the new tech, and every day more companies are jumping on the bandwagon.

Below is a summary of how AI is revolutionising the following five industries:

Electric Utility
Health Care

We’ve all been in that situation (either as an employee or customer) where there is endless frustration and arguing over an out-of-stock product on Christmas Eve. AI technology has reached a point where it can predict demand, analyse stock levels and automatically order products to completely eradicate this problem and deliver a better shopping experience.

AI has insight into what consumers want. Deep learning systems analyse data on past transactions, weather forecasts, media trends, online viewing history and even facial expressions of shoppers to anticipate what will sell and to automate orders from suppliers. Accuracy in this stage means more efficient supply management, and AI can also give direction on the optimum assortment and pricing of existing inventory stock. Automating in planning and stocking means more attention can be paid to customer service. German e-commerce merchant Otto adopted AI in forecasting and stocking found it helped them reduce product returns by over 2 million per year and cut surplus stock by 20% (Bughin et al, 2017). The system’s 90%+ accuracy enables Otto to order 200,000 items a month with no human input.

In the warehouses, automation can also be used to increase productivity by working faster and minimising the chance of human error. Swisslog reduced stocking time by 30% with AI-powered autonomous vehicles (Bughin et al, 2017). Ocado (an online supermarket in the UK) has a warehouse filled with conveyor belt, robots and AI applications to move products from the store to shopping bags to delivery vans to customer – a futuristic concept that mirrors Amazon’s existing distribution centres.

AI can also help optimise personalised marketing in real time through insight-based selling (customised promotions and assortment, tailored displays etc). Online, this can boost sales by up to 30% as it is so easy to collect customer information instantaneously (Bughin et al, 2017). Surprisingly, insights can be used to support in-store sales as well – face recognition technology identifies customers and analyses their shopping history to identify ways to target them or products they may consider. AI also helps analyse and identify prospective locations for new stores based on demographics and sale patterns, as well as proposing suggestions for what stock to focus-on in new stores.
Electric Utility
From power generators to consumers, there are endless opportunities for AI technology to increase efficiency.

AI can offer highly accurate short-term forecasts for energy use, cutting wasted energy by 10% and maximising the potential of renewable sources (Bughin et al, 2017). Wind stations can also predict their energy production levels by analysing past performance, wind direction, and speed, boosting production up to 20%. Sensors and machine learning also allow for real-time adjustments to the turbines with consideration to wind strength and direction to maximise energy efficiency.

Similar technology also helps improve preventive maintenance. With the help of AI, some coal plants were able to predict which machines would fail six to nine months in advance, minimising down time by performing repairs before they escalate. Smart meters feed information back to utilities instantaneously so technicians can spend time problem solving rather than logging data. They can also be used to measure the energy suppliers and households put back into the grid from renewable sources to help calculate owing costs instantaneously.
Manufacturing is a long and complex process from the first supplier to the store. With the help of AI, development cycles can be shortened, faults avoided, costs reduced, and injuries minimised.

In the future, manufacturing chains will be interconnected – manufacturers, suppliers, and businesses will seamlessly communicate and coordinate to ensure a smoother production and delivery process. AI helps achieve transparency in machine availability, performance, production, and down time, which means forecasts can accurately be communicated in real time to other members of the manufacturing chain.

Advanced analytics can identify problems earlier and rectify the situation before it becomes an expensive error. Manufacturers can similarly identify when after-sale service is needed rather that doing time consuming tests and checks. A complex network of sensors and actuators allows companies to monitor product use, identify potential issues, and resolve errors early to minimise disruption and downtime. This is particularly useful for airlines as it means technicians can identify where the issue will be in advance and schedule a maintenance session, minimising potential delays. In the future, bug sized robots will be used to inspect aircrafts internally using computer vision, avoiding the need remove panels.

Advance analytics can also rethink processes and practices to reduce costs, minimise waste, and accelerate time to market. Machine learning and numerical optimisation helps factory managers increase accuracy of operating predictions, identifying where errors or clashes could occur in the production schedule. General Electric used AI to develop a system to plan their flight paths based on wind and weather conditions and other air traffic to maximise fuel efficiency, resulting in 12% greater efficiency and less strain on the aircraft (Bughin et al, 2017). Across the fleet, this reduced annual fuel costs and extended the distances planes can fly before requiring refilling.
Health Care
Generally, the health care industry is slow to adopt digital trends. A study by McKinsey Global Institute (2015) found 25% of American hospitals had not adopted an electronic health record system, and those who do have histories digitised often do not share them with other health care services – meaning tests are being repeated unnecessarily and patients need to repeatedly detail their medical histories. Simple digital applications are able to streamline this process to save time and money for both the provider and patient.

It is becoming easier for individuals to track their health and the potential of devices to act as a health support source is ever increasing. Machine learning in mobile devices (such as the popular Apple Watch) are already able to track the user’s health patterns and will alert them to possible health issues (e.g. an abnormally high heart rate with trigger an alert to check for heart disease).

AI can also be used to increase the accuracy and speed of diagnosis by comparing the patient’s data with a large data library of past cases – common symptoms generally point to a common health problem, and most effective treatments are easily identifiable. As beneficial as this would be, it is hard to make this a widespread reality as people are hesitant to have their medical histories made public and shared with computer systems and other providers, preventing AI in the health industries reaching its maximum potential.

AI can also look at a patient’s medical history alongside weather conditions and identify people at risk of allergies and public health threats, directing them to preventative measures. If asthmatics are warned before a dust cloud hits, they can prepare for it and minimise the impact, drastically reducing the time patients spend in hospital to avoid overcrowding and unnecessary strain on the healthcare system. Providers can also optimise hospital staffing and inventory by using information from trends and weather forecast to predict disease probability, preparing them to efficiently combat viruses and minimise the spread of disease.

Virtual agents have also been built to use basic patient knowledge to match the patient with the most appropriate doctor based on their symptoms and specialisations to provide the best treatment possible. They can also use natural language processing techniques to analyse and summarise journal articles for doctors, assisting with accuracy and speed in diagnosis and treatment. AI applications can sift through endless pages of medical data in seconds to suggest a well-informed treatment plan, and see identify more subtle details in MRI and X-ray scans than the human eye.
Education is becoming more personalised and intense than ever before – each day, an individual is bombarded with 174 newspapers worth of information, and it is important that they are able to absorb and remember the relevant content (Alleyne, 2011).

There are increased efforts to shift from a standardised learning approach to tailored learning, with consideration for the student’s existing knowledge, learning preferences, and progress. Adaptive learning through AI helps deliver content in the right way, at the right time, at the right speed, boosting confidence and performance. An example of this is the virtual tutor, which personalises content delivery by testing students to analyse which delivery methods ensure the best information recall and retention. This is a big growth area – private investment in educational technology has increased by 32% each year from 2011 to 2015 (EdTechXGlobal 2016).

Similarly, in the classroom, computer vision can track the eyes and movements of students to analyse whether they are engaged, confused or bored and identify their learning preferences. This data is then used to direct lesson planning to maximise learning and engagement. All this information is synthesised to form effective learning groups based on student profiles, learning styles and knowledge levels to further increase the learning experience.

Similarly, universities are using AI to recognise trends in students before they drop out. This allows universities to intervene and get the student back on track by offering support and better suited learning before they drop out. Forecasting methods also helps plan for the future by preparing students with the demanded skill sets. Universities look at the requirements of employers and where there is demand for employees in the coming years and use this insight to shape their course marketing plans. They direct students to the paths with greatest demand, bridging the skills gap in employment and preparing society for coming years.

Machine learning can be used to speed up the administrative and basic tasks of teachers, such as marking and documentation, to allow them more time for the students. Natural language processing can decipher student handwriting and mark objective questions in assessments. The University of Akron built a predictive platform in 2012 to assess 16,000 essays. The computer grades matched the same mark given by a human teacher approximately 85% of the time (University of Akron, 2012). Remaining consistent while grading written assessments is one of the challenges of being a teacher, and automating this would save considerable time for the teacher.

McKinsey Global Institute, December 2015 ‘Digital America: A tale of the haves and have-mores’

Field, D, December 2015 ‘See the heart in 7 dimensions: This team of researchers attacks world’s biggest killer withsoftware’ GE Reports

EdTechXGlobal and Ibis Capital, 2016, ‘2016 Global EdTech industry report: A map for the future of education’

University of Akron April 12, 2012 ‘Man and machine: Better writers, better grades,’ University of Akron press release

Alleyne, R 2011 ‘Welcome to the information age’ The Telegraph<>

Bughin, J, Hazan, E, Ramaswamy, S, Chui, M, Allas, T, Dahlstr.m, P, Henke, N, Trench, M 2017 Artificial Intelligence: The Next Digital Frontier? McKinsey Global Institute