Artificial intelligence (AI) and machine learning (ML) are two buzzwords that have emerged in popularity over the last few years. Opinions on the subject vary greatly, some believe that AI represents the future of business, while others think of AI as robots surpassing humans and taking over jobs. Around the world, the giants of Silicon Valley like Facebook and Alphabet have been grabbing headlines with their significant investments in AI companies. In 2016, China launched an ambitious plan to become the centre of global AI innovation by 2030 and in 2018, McKinsey released a study predicting that AI would create $13 trillion of value by 2030. The complexity of the field combined with the excitement around the technology has created a movement of information overload. This raises the question - what really is artificial intelligence and how is it impacting businesses today? Definitions
What is referred to as artificial intelligence can really be divided into two categories:
1) Artificial narrow intelligence (ANI) also known as “Weak AI”, is the AI that exists in our world today. This form of AI is rapidly developing, contributing to the headlines that we see in mainstream media. While the machines may seem intelligent, ANI exists to perform a singular or limited task within pre-determined ranges.
Machine learning is an umbrella term commonly interchanged with artificial intelligence. Technically speaking, machine learning is a subset of artificial narrow intelligence, and the most successful modern AI solutions are powered by machine learning algorithms. The process with machine learning is two-fold:
Step 1: Learning – An algorithm is provided datasets of inputs (A) and outputs (B), and generates its own model to infer the connection between A à B.
Step 2: Application - The algorithm uses the model A à B from Step 1, applies it to new data (A1), and makes new predictions (A1 à B1).
Examples of ANI include: Facial recognition (Apple’s Face ID), speech recognition (i.e. Apple Siri, Amazon Alexa, Google Home, chatbots for customer service), image recognition, self-driving cars, spam filtering, fraudulent transaction detection, medical diagnosis (e.g. A machine that can identify cancer in an x-ray), manufacturing robots, recommendation engines (Netflix or Amazon’s “Recommended For You”)
2) Artificial general intelligence also known as “Strong AI” refers to capabilities equivalent to what a human can do, or greater. This is otherwise the common reference to AI as robots taking over our jobs. One of the key differences between ANI and AGI is that ANI lacks the cognitive ability or the ability to reason. In other words, ANI cannot decipher emotions nor solve unfamiliar problems like that of a human brain could. As it stands, there has been no progress on AGI to date, and it is understood that we are decades and multiple technological breakthroughs away from any AGI progress.
How can businesses benefit from AI? As the business landscape continues to change with AI adoption and automation driven growth, there are benefits but also limitations to be considered: Benefits:
Better understanding of consumer behaviour – Some of the larger areas of growth and adoption of AI have been in marketing and e-commerce. With customer data, companies may use AI tools to forecast future trends and predict consumer behaviour, which in turn impacts marketing strategies. Better insights allow an organization to bring more value to their customers, which could increase customer lifetime value and customer loyalty.
Efficiency and better decision making – When an organization needs to make a decision, it takes time. Perhaps one of the biggest advantages of AI is its ability to speed up the decision-making process by automating parts of the process. Such efficiencies will only continue to increase as the technology matures and the cost of prediction decreases over time. For repetitive tasks, AI will also reduce the risk of any human error influence. An example of an AI tool that would result in faster decision-making is demand forecasting, which would also impact inventory and supply chain management.
Shift human focus to higher value and complex tasks – With AI handling basic tasks, the value of human judgment in the workplace will increase. As a result, organizations can focus on optimizing and maximizing their human talent. AI is not displacing humans but can be seen as a technology that will empower humans to focus on more complex issues that require deeper thought and creative thinking.
Limitations:
Cost of data acquisition – Data is expensive, and in order to work effectively, large datasets are required to train and provide feedback to the systems. AI cannot predict on data that is different than the original dataset and may even generate inaccurate outputs if the training or feedback datasets were not large enough.
Susceptibility to bias, inaccuracies, or even discriminatory outcomes – Machine learning systems are taught how to perform the work they are designed to do, however, one of the biggest weaknesses of machine learning is when a machine confidently yields the incorrect answer. Machines do not know when they have predicted incorrectly and it ultimately depends on the type of data set that the system was originally trained on; if the training and feedback data contained any bias, the algorithm will learn this bias. This highlights the importance of ensuring that data is inclusive and that any data outliers are identified and removed.
Not all problems can be solved with AI - While AI represents a tremendous shift in technological advancement, AI is only effective when applied to simple concepts with plentiful data that does not require judgment. While some tasks in a workforce can be fully automated, other tasks will still require human judgment and decision-making. In a case where there is little data or uncertainty, humans will always be better at deciding what to do.