There's no denying that Artificial Intelligence (AI) is now the next big thing. The term was coined in 1956, but only incredible computing power and data storage capabilities of today enabled AI to truly unfold its wings. It's enough to look at McKinsey's or Forrester's reports to get a sense that business all over the world starts to take advantage of this new opportunity. And staying behind on this is not an option. But how to implement it? And why it's so popular now, after 60 years of existence?
AI moves imaginations and ignites hopes. We've all heard AI enthusiasts claiming it could propel us towards building a heaven on Earth, and we've all seen Hollywood productions that tell stories about humans doomed under AI's dominion. Even Stephen Hawking who used intelligent word prediction to speed up his writing said that the "development of full artificial intelligence could spell the end of the human race". Elon Musk claims that the "AI is a fundamental risk to the existence of human civilisation".
Who is right? Can and should we benefit from AI? To find out how we can use it and what dangers it poses realistically, we first have to answer the question: what AI really is?
AI - let's give it a name
Michael Chui, a partner with the McKinsey Global Institute, defines AI as follows:
...basically it refers to using machines to do things that we consider to be “intelligent”—being able to either simulate or do things that we describe people as doing with their cognitive faculties.
Artificial intelligence (AI) makes it possible for machines to learn from experience, adjust to new inputs and perform human-like tasks. Most AI examples that you hear about today – from chess-playing computers to self-driving cars – rely heavily on deep learning and natural language processing. Using these technologies, computers can be trained to accomplish specific tasks by processing large amounts of data and recognizing patterns in the data.
In general, Artificial Intelligence's task is to mimic human cognitive functions. AI programs can:
- solve specific problems
- analyse and interpret graphical inputs
- analyse and interpret language (written and spoken)
This is a quite broad spectrum, but because of that AI gives us a wide array of very handy solutions:
- intelligent chatbots and virtual agents
- image and video recognition (like face recognition)
- automated translations
- autonomous cars
- intelligent schedulers
- targeted advertising
- predictive analytics and anomaly detection
- self-optimising devices (like wind-farms)
It's no surprise such possibilities make AI look like it can easily overtake humans, let alone replace them at work. But when we zoom in to see the underlying principles on which AI is built, we see that there is less to worry about and more to look forward to.
Machine Learning and two types of AI
AI is the result of incredible data storage possibilities and computing power. The idea of AI itself is very broad and has virtually limitless applications, but nowadays AI is based on Machine Learning.
Machine Learning takes huge amounts of data, feeds them to a semi-autonomous program, which is nurtured and taught by humans to recognise specific patterns. Gradually the program learns to figure out by itself the best way to achieve a given task - be it suggesting a new product, deciding if a car should slow down, or understanding what a user of a chatbot really needs.
The reason I prefer calling today's AI solutions as semi-autonomous is simple. Although they can decide and interpret, they operate in a narrow window of tasks and have to be taught by humans. Move a car from point A to B, interpret a written text, trade stocks and shares. This is in no way a small feat, and it's not easy even for humans, but still - these AI programs cannot and will not go outside their bubble of specialisation.
What we have today is defined by experts as a "narrow" AI. What the Hollywood and influencers fear is "generalised AI". This would be an all-round software that can theoretically handle any task and find solutions to unfamiliar problems. There is no example of such AI today, and personally, I see no reason this kind of AI should ever come to be. We already immensely benefit from "narrow" AI, that gets more and more sophisticated. And we already worry enough about current AI solutions.
Real business benefits of AI
McKinsey recently did an analysis of over 400 use cases across 19 industries to evaluate the practical applications and the economic potential of AI in business. What they found out is:
- AI can extend the life of revenue-generating machines beyond what was previously possible. If you want your machines to live long and perform well, predictive maintenance is the way to go. AI can use huge amounts of data collected by relatively cheap sensors, process this data, and predict failures, allowing you to plan maintenance in advance. At Objectivity, we have created a cool Proof of Concept.
- AI can cut costs in many sectors by optimising logistics. Using sensors and data AI can find optimal routes and even coach drivers in real-time to encourage them to drive more economically. For one of our Clients, we have created a solution that helps them save a lot of fuel.
- AI is a perfect tool for improving customer experience and increasing conversions. A program that analyses audio can detect customers' emotional tone and respond accordingly (e.g. relay a call to a human operator). AI can also analyse customers' demographic and purchase history and propose relevant purchase options (up to twofold increase in conversions).
- AI can greatly improve analytics performance. Depending on an industry, the value from analytics can increase from 30 to 128 per cent. Humans cannot process vast amounts of data, but with the help of AI, we can better see how our businesses perform under certain conditions. For one of our clients, we have built an anomaly detection engine based on machine learning.
- AI has the potential to create up to $5.8 trillion in value annually across 19 industries.
With that being said, we have to recognise some of the drawbacks or challenges that AI brings. All emerging technologies do, but since AI has a huge potential, the challenges it brings are also wide-spread.
Artificial Intelligence and its challenges
One of the biggest challenges - or fears - is that AI could someday potentially replace humans in their jobs and that we would be left without an income. The economic solutions to such a scenario are beyond this article, but even the technology perspective gives hope.
At Objectivity, we practice something we call ethical automation. This is automation - be it AI, business intelligence, advanced analytics, cognitive interfaces, or RPA - built around humans' needs. Automation made in line with human centred design support our daily work. It takes care of repetitive, mundane, or tedious tasks, and lets people do what they do best - think creatively and solve problems.
Sure, it is probable that in the future cars will be autonomous and lorry drivers could face some shift in their work. But today experts say that this shift will not be as drastic as some people predict. AI will be able to drive lorries between points of delivery, but we will still need human eye and reasoning to navigate through crowded spaces – more unpredictable than highways.
Another challenge is data. Once AI is up and running, it can propel your business forward, but to do that AI needs data. McKinsey says:
Deep-learning methods require thousands of data records for models to become relatively good at classification tasks and, in some cases, millions for them to perform at the level of humans.
And the data must be of good quality. This means it has to be comprehensive and unbiased. It's easy to design a data gathering mechanism that supports our thesis, but an AI solution that brings value needs to see the big picture. Gathering such sets of data is no small feat, you need expert Data Scientist to build your data sets. And even when you have those, you have to train (and keep retraining) AI to recognise patterns and make decisions. You can read here:
Along with issues around the volume and variety of data, velocity is also a requirement: AI techniques require models to be retrained to match potential changing conditions, so the training data must be refreshed frequently. In one-third of the cases, the model needs to be refreshed at least monthly, and almost one in four cases requires a daily refresh; this is especially the case in marketing and sales and in supply chain management and manufacturing.
All this requires a clearly defined business vision and expected results from an AI solution.
It's easy to follow trends and build the next conversational interface. But as Forrester reports: "Consumers want convenience and choice — not conversational interfaces. (...) Digital business leaders should only build conversational interfaces where they offer more convenience than existing means of interacting with customers." You need a viable use case to implement a conversational interface.
Experts agree: if you want to stay ahead of your competition, you need to adapt AI as fast as possible. It is not easy to do, because you need a team of expert Data Scientists to build relevant data sets, and a lot of specialists that will train (and keep retraining) your AI to support your key business goals. Predicted $5.8 trillion in AI value across 19 industries is no joke, and we'd better adapt now than later. What we have to do is look at our business strategy, decide which aspects AI should support, and then let AI experts do their work.