Nancy Rademaker addressing the channel during the Nextgen Leadership Forum
When picturing the future of your business, it is important to have a high-level understanding of emerging technologies.
Often, these technologies are talked about without a real clear insight into what they actually are, but more importantly, what they might achieve.
With reference to artificial intelligence (AI), this is exactly what is going on: everyone talks about it, but nobody really knows what it means and how it may drive business value.
This mainly derives from the fact that there is no consensus on the definition of AI. But is this really important? Instead of focusing what it is, isn’t it more important to focus on what it can do?
You could describe AI as a set of algorithms and techniques that enables computers to mimic human intelligence and human behaviour. That’s exactly what scientist working on the Dartmouth Project in 1956 set out to do: “creating an artificially intelligent being”.
Programming computers to behave like humans. And in order to achieve that goal, these computers needed to have some kind of representation of the world as we know it.
They needed to be able to do some kind of reasoning and problem solving, to navigate our world and plan accordingly, to process natural language and, last but not least, to perceive the outside world and react to it, preferably in real-time. Basically, that’s what AI research has been about for half a century.
For the last decade or so, the focus within AI has been on a subset called machine learning. The word says it all: the machine learns by itself.
It is both adaptive, able to improve by experience, and autonomous, in that it doesn’t need continuous guidance by an external person.
So what is machine learning all about? One thing: patterns. It can perform statistical analysis of data to find patterns (‘algorithms’) that computers could not see before. Patterns that humans don’t even understand either. How does this happen?
In two phases: first, training data is being fed into the system and a model is created automatically, and next, the model is used to make predictions about the actual ‘production data’. Needless to say that the more data being used to train the system, the more accurate the algorithm will be.
And so, if years of research and experimentation have gone into this and algorithms of various types have been created, what can we actually use them for?
The acronym POCUS may help you remember them:
Prediction
This truly is AI’s most powerful feature and represents great opportunities for companies.
Using machine learning, you can make very diverse predictions: how your customers will behave, what their future needs will be, how your state of health is evolving, how machines will behave and whether or not they will need maintenance, how particular cells within your body might lead to cancer, and so on and so forth.
And the truly remarkable thing here is that AI is very strong at weak features: combinations of certain elements that we as humans do not regard as relevant for a particular outcome may indeed prove to be so.
Object recognition
Algorithms have become so advanced that they can now recognise still or moving objects with greater accuracy than humans.
This typically proves useful for example in visual search (such as performed by Pinterest), in categorising pictures (like most of our smartphones do), but also in precision agriculture where fertilising crops is no longer a mass action, but is performed for the individual crop based on the recognition of its development state.
Content creation
Algorithms will be used to create content automatically. This is already being done in sports, where AI can write “basic coverage” articles. During the last summer Olympics, two bots by the Washington Post and Toutiaou wrote tweets or even short articles based on the result and video footage of games.
With regards to business, Gartner already predicts that by 2020, 20 per cent of all content will be generated by algorithms.
But it’s not just words that algorithms can create. You can now have AI create photo-realistic pictures from text, compose music, make paintings or even create extremely natural-looking pictures of non-existent people.
Understanding people
This is of course the area of chat bots and voice assistants where automatic speech recognition and speech synthesis play a major role and the algorithms have greatly improved over the past decade.
The quality metric for automatic speech recognition is “word error rate”, so how many words do you get wrong. The human word error rate is around 5-6 per cent, algorithms are now at four per cent and falling.
Next to having voice assistants and home devices like Amazon Echo and Google Home, algorithms in this application area are also being developed to help you with writing texts, automatic replies to emails and real-time translation.
Self-moving vehicles
And of course, this last area is the one that probably appeals to the human imagination the most. Self-driving cars, self-flying drones, self-walking robots, no one knows exactly when they will become widespread but that they will is definitely a certainty.
Needless to say that all of these applications areas provide great opportunities for companies to make a big leap forwards. Already numerous algorithms are available with the big tech companies, but also in open-source frameworks.
So what’s next?
The challenge now for anyone leading a business is how to make optimal use of AI and this is in fact not something to be achieved easily. You need to get new capabilities on board - or hire experts - and you need to get everybody trained at least on the most basic principles of AI.
Get the right tools in place and if you prefer to buy over build yourself, make sure you ask your software provider in what way their applications make optimal use of machine learning techniques.
The most important thing is to find some projects to start experimenting. Look at the challenges you already have now and pick some of those where you think the corresponding business outcomes would benefit most from AI.
Collect all the data that is needed – they will most definitely need some cleansing as well – and just start. And of course, since they’re experiments, the outcomes will most probably not be successful every time, so perseverance is required.
With over 20 years of experience in IT and training, Nancy Rademaker – partner at Nexxworks – has always and above all passionately focused upon people: how technology influences their behaviour, how it helps them share knowledge and how it enables them to create and innovate.
