Artificial intelligence (AI) is one of the hottest topics right now, covering a range of industries. The trouble with a hot topic is that it is often accompanied by a lot of hot air. So here is our guide to help you demystify what it is and how it can be used by marketers to shape customer relationships.
What is AI?
As AI gets more and more mainstream coverage, it is becoming increasingly synonymous with algorithms and automation - but the fundamental thing that makes artificial intelligence more than just intelligence is that the code learns for itself. In other words, if you’ve written an algorithm and understand precisely what it will do, it’s not AI. That doesn’t mean it isn’t clever or useful, but there is nothing artificial about the intelligence. AI is about getting machines to learn for themselves from data.
Why is it important?
Through enabling machines to learn for themselves, far faster and more effectively than humans, we can use them to make ourselves more effective, leaving us free to focus on the things we are good at - creativity, strategic thinking and innovation - while they help us with tasks that are more mundane and repetitive.
What skills, technology and resources are needed to develop AI?
AI is a developing and highly-technical field and fully exploiting it requires superior mathematicians and computer scientists. In addition to the requisite skillsets, AI demands a lot of data from which to learn.
Critically-important AI techniques revolve around deep neural networks. These are based on how our brains store and process information, make connections and develop new insights.
On the technology side, significantly improved data processing ability has come from some clever repurposing of graphics processing units (GPUs as opposed to CPUs) to allow the computation of neural nets. But there are also current projects delivering technology specifically designed to support AI, such as Google’s Tensor Processing Units.
Theoretical advances have also spurred improved performance, such as the introduction of convolutional neural networks, which underpin the ability of AI to make sense of image data.
Is AI dangerous?
A TED talk by Joy Buolamwini considered whether AI can be racist, though the issue really stems from data rather than techniques. If an AI algorithm is trained from a set of images that are predominantly of white people, it probably won’t give great results when applied to images of people of other races.
AI is only as good as the data it has to learn from - duff or biased data may produce erroneous or incorrect conclusions. However, this is not a new problem - data scientists have been aware of, and have successfully navigated, this issue for many years. But, as AI becomes more and more pervasive in our lives, it is vitally important that we are mindful of this issue.
What will be the impact of AI?
At a philosophical level, we have a choice about how we relate to AI. The fear today largely stems from how AI will affect our privacy and our livelihoods and, in turn, our quality of life.
Privacy is a concept that has been amplified by technology and it is a fundamentally personal thing. It’s also one that changes depending upon the reward for sharing information and how we each value that reward. As we are nearing evolutions such as the realisation of the internet of things, or the demise of passwords as we use biometrics or natural language processing, our relationship with the concept of privacy will change.
Following every significant advance in technology and automation to date, there has been a fear the machines will take our jobs. Equally, there can be no doubt that some tasks currently performed by humans will be done by machines in future. However, invention and development is a human condition, not to mention a commercial imperative.
A positive view of the promise of AI is that is frees us from the work that bores us to death or takes up our time in routinely doing, rather than thinking and improving.
Where is AI right now?
Most AI today is for specific applications with a series of fixed parameters or inputs, so called “narrow AI”. We’re still largely in the realms of things like natural language processing with business rules on how to respond. Artificial general intelligence is the version of AI that is generally capable and more rounded, more akin to our own intelligence. But, as of now, even experts in the field admit that achieving it seems a way off.
On the flip side, technological advance is exponential - in essence, that means that things are likely to progress faster that many of us think and, when we look back at 2017 twenty years from now, we’ll probably say, “that wasn’t proper AI”.
What can AI be used for?
The applications of AI in the world per se are endless. Certain tasks are increasingly becoming commoditised and price-driven. This is where AI can revolutionise what we do and provide us with the opportunity to advance and use our deep specialist skills to add real value to our client relationships.
In terms of our clients, these could be driving real-time pricing changes in response to competitor pricing or an individual’s price sensitivity. Other applications might include intelligent approaches to lapsed customers, crisis PR, biddable media, display, attribution and influencer marketing, web development, creative pursuits, image processing and recognition, chatbots, self-learning statistical models and much more besides.
AI is such a hot topic right now, but it’s important that it doesn’t become:
AI is being badged as the fourth industrial revolution and its potential is limited only by our own imaginations. Like it or not, it is here to stay. We’re positive about the change that AI will bring and believe it can help us become free to further invention, have a better work-life balance, richer experiences and better relationships.
You can download a copy of this paper here.