What AI and Machine Learning mean for business

First things first. There’s no such thing as genuine Artificial Intelligence.

There’s clever software. But no software will ever emulate what management gurus so often (too often, in fact!) praise: seat-of-the-pants gut feeling, ethics and all that. No computer programme can infallibly forecast election results, for these are in our hands, not its. What sophisticated code can do is enhance human intelligence, which is by itself a unique, many-splendoured thing: artistic, political, emotional, self-conscious.

What is today termed Artificial Intelligence does augment the magnificent human sort. Like the Internet of Things, AI techniques carry the promise of faster and cheaper processes, as well as processes that are more insightful with inputs and more reliable with outputs. AI techniques include Natural Language Processing (NLP), the deployment of layer after layer of neural networks, and Machine Learning (ML): algorithms that, through feedback loops, can be ‘trained’ to classify and draw inferences from data, and make probability-based decisions as a result.

First, catch your graphic processing unit (GPU); they’re able to process lots of data in parallel. Next, equip a GPU-based machine with ML. Then feed it millions of photographs, videos, text, spoken words, attributes or games. Supervise the machine’s training if your data has labels, or let it learn unsupervised if you don’t. Eventually, it ought to recognise a lot of things right. On top of its observations, reward it for accuracy, and give it memories to recall, too. Then from, say, raw pixels, it may be able to make models and select relevant actions.

What difference, then, will ‘AI’ make to your business? It can predict more accurately when your equipment needs maintaining. It can also help you see what cyberthreats might do, and what they have in common with each other; at PayPal, senior director of risk sciences Dr Hui Wang says it decides within a couple of hundred milliseconds whether a respondent is ‘a potentially bad guy’.

The hottest field at present is ML, and its more sophisticated derivative, Deep Learning. With Google’s Android’s operating system, for instance, ML is now offered to developers of mobile apps.

Zurich Insurance uses ML to assess claims more precisely. Retailing ML specialist BlueYonder claims to cut the dismal message ‘Out Of Stock’ by 80 per cent, predicting demand by taking into account patterns of shopping, the weather, holidays and major events. And at General Electric, the procurement function used ML and Dun & Bradstreet data on suppliers to save $80m within a year – rationalising suppliers, targeting more of its sourcing on lowest-cost countries, searching out interchangeable parts, and coordinating the different geographical footprints GE had for procurement, manufacturing and marketing.

In general, what is termed AI is about gaining special insights about customers. But it can also be deployed as a means of competitor intelligence. Thus Getty Images works with Mintigo to check where, on the Web, pictures from other photography houses are being displayed – in order that it can pitch to those companies publishing them, and so supplant its rivals.