What does AI mean to you? If you think it stands for artificial insemination, you’ve probably stumbled across the wrong website! But, assuming you know it means artificial intelligence, just how much do you actually know about it? Does it worry you? How much impact do you anticipate AI and digitalisation will have on your life?
In this blog post, we’ll look at the role of AI in business and examine how it is changing the way we work.
What is AI?
Artificial intelligence, or AI, has been around since the 1950s. Minsky and McCarthy, widely acknowledged to be the founders of AI, described it as any task performed by a program or a machine that would require intelligence if a human were to carry out the same activity. Today this is considered a very broad definition, and it is more meaningful to think of AI as systems that typically demonstrate some of the following behaviours we normally associate with human intelligence: planning, learning, reasoning, problem-solving, perception, motion, manipulation and possibly social intelligence and creativity.
AI in daily life
You may think your life is quite ordinary; your use of technology is mundane and you are anything but an innovator or early adopter. AI is for Generation Y and the millennials, right? Wrong! If you shop on Amazon or watch Netflix you use AI. Those Amazon recommendations that say, “people who bought this book also bought…” and “recommended for you, based on this purchase” are examples of AI. The Netflix suggestions follow a similar pattern: “here are some films you might like, based on your previous viewing habits”.
If you use an iPhone, you’re probably familiar with Siri, your digital personal assistant. Even if you don’t use her, you know she is there. Siri can find information, give directions, add events to calendars, send messages and perform various other administrative tasks.
Amazon’s Alexa is another digital helper, more commonly used in the US but gradually gaining popularity worldwide. Alexa can search the web for information, shop, schedule appointments, set alarms and power smart homes.
These are all examples…and there are many others…of AI in everyday use. AI is not the future; AI is daily life right now. Even self-driving cars are no longer confined to science fiction. Self-driving trucks are being piloted right now in the US and UK.
There is both confusion and controversy surrounding AI. Some claim that the Netflix algorithm is not AI; it is simply machine learning. Others argue that machine learning is a type of AI. So let’s look at the terms in common use.
Narrow AI – Siri is an example of narrow AI, which refers to the type of machine intelligence that can only learn how to do specific tasks, such as responding to simple customer service queries, making hotel bookings, organising your diary, etc.
General AI – General AI is more science fiction than fact…yet. General AI resembles human intellect. It is adaptable and capable of learning how to carry out diverse tasks. Think of The Terminator and you’ll get the picture.
Machine learning – Machine learning is a product of “big data”. When a computer system is fed large amounts of data, it analyses the data to learn how to carry out a specific task, such as understanding speech or captioning a photograph. This is how Amazon finds book recommendations for you. One person with the same book choice is pure coincidence. You might not like any of their other purchases. But when thousands upon thousands of customer purchases are analysed, patterns of similar preferences emerge, and it becomes statistically more accurate to suggest that you might like the same books as other customers who bought your book. The computer “learns” what readers of a particular book like and makes suggestions to the next purchaser accordingly.
Neural networks – Neural networks underpin machine learning. They are networks of interconnected layers of algorithms that feed data into each other. The inspiration for neural networks came from the workings of the human brain. These networks are trained to carry out specific tasks according to the weight attached to the data as it passes between the layers. This is where some of the risks of AI lie, as the weightings are ultimately programmed by fallible humans. We’ll come on to risks shortly.
Deep learning – Deep learning is a subset of machine learning. Neural networks are expanded into a vast number of layers that are trained using massive amounts of data. They are able to carry out much more complex tasks, such as speech recognition.
AI in business
Alexa and Siri may be widely known personal applications, but the real value of AI is in business. Industries as diverse as healthcare, agriculture and investment finance use a multitude of systems based on machine intelligence. Enterprise functions from customer support to recruitment are being enhanced by products such as ActionIQ and HireValue that are AI-driven. Shivon Zilis and James Cham, writing in the Harvard Business Review, have mapped the state of machine intelligence and it is clear that AI is everywhere…making customer interactions more efficient, security systems more effective, etc. There is even software that can analyse the tone of voice or the cadence of language to detect the state of mind of a caller to a call centre, prompting the agent to respond in the most appropriate manner. Wherever there is data, there is artificial intelligence.
Will it take over your job? Probably not…
Robots will replace humans and jobs will be lost. That’s the fear many people have when it comes to AI. Undeniably, there are industries where robots can perform certain repetitive tasks previously undertaken by humans, such as production line jobs. But in reality, AI works best alongside human input. Some of the most effective applications of AI combine machine analysis of structured data with human evaluation of unstructured data, where judgment is required. The human input feeds back into the system and improves the quality of the output in the future, in a virtuous information loop.
AI, properly harnessed, can also lead to massive productivity gains. In healthcare, predictive models that run on patient data and computer vision are able to diagnose disease from medical images and gain lifesaving insights from genomic data. Inventory management in grocery retailing is improved by machine intelligence. The financial services industry is benefitting from real-time analysis that used to take an analyst several hours to report, thus enabling investment firms to respond more rapidly to market movements.
But there are risks
However, if the human information stream contains bias, the machines learn the bias, for example in the case of a bank lending decision-making algorithm containing racial bias. Sometimes the dataset contains too few examples of a particular characteristic, and the machine is unable to learn to discern variation. This can lead to it drawing incorrect conclusions. Commercial facial-recognition algorithms in the US are far better at telling white men apart than they are at recognising women or people of other races.
An algorithm in itself is a neutral device; it doesn’t have thoughts or feelings. But it is managed by humans, and there is often little transparency about the design. So while AI offers a brave new world of customer-centric transactions and increased productivity, it must be embraced with awareness of the innate risks.
Businesses introducing intelligent systems must give serious consideration to the level of trust apportioned to these models and the amount of decision-making power instilled in them. The risk of error and the ensuing consequences must be carefully assessed, and the frequency with which the systems are reviewed must be determined. But the nature of machine learning is that the models will improve, as they will continuously learn from the data they process. AI is here to stay – learn to love it!