Artificial Intelligence is eating the world, but how can we use this new power for good (and profit)? And should we be worried about the killer robots?
Artificial Intelligence (AI) is a major topic these days, and promises to help businesses achieve their goals of optimizing their operations. Many of today’s big players like Google and Facebook have AI as their core competency. Telenor has a great potential to leverage this technology in our digitization, but what does this actually mean? For many people AI is a black box, and with this article we will hopefully make it a bit more approachable.
When thinking about AI, humanoid robots is one of the first things that come to mind. Are we close to a society where we will not know if a person is human or not? Will we have machines that act like human and are sentient? These thoughts consider a machine that can operate as a “general AI”, an AI that can do general things without being specifically told what it should do – it can learn on its own. Essentially, something that acts and thinks like a human. It is easy to expect something like this, especially if people have heard about “training an AI” to complete a task. However the reality is not exactly as one might think.
The AIs we have today can do complex tasks, but are not general AIs. They can only do a single specific task that they have been taught how to do. When you hear people talk about “training” an AI, what they really mean is that the AI program tests many different configurations until it finds one configuration that solves the specified task at hand. This task has to be extremely specific. Most of today’s problems solved by AI have to be explicitly transformed into a form that can be understood by a computer, and the result transformed back, so we’re not on the verge of a robot uprising just yet.
The most important technique being used for cutting-edge AI is “deep learning” with neural networks. This technique means that we define a large, layered, mathematical network of nodes that can perform simple arithmetic operations in each of its nodes. The arithmetic operations are combined through the network, which is how interesting and complex computation arises. In simpler terms, one can think of the network as a network of roads with cars on them, and only if certain criteria are met will the roads be open for the cars to drive on. When the computations are done, you simply check how many cars arrived at the destination (or destinations), and that is your answer.
These techniques are not new. Neural networks were first being studied in the 1940s, and by the 1980s we had most of the knowledge to do what we do today. So why has it taken so long? There are several different reasons for this. One big reason is that computing these networks, i.e. trying to figure out what roads to leave open and what to close, so to speak, is computationally expensive. We’re just now, with the nascence of general purpose parallel computation on graphics processing hardware, getting fast enough. Another reason is that vast amounts of training data is needed, and large datasets are now more readily available than ever.
Perhaps most importantly for AI, the tools used by the big players have been open-sourced for everyone to use! Not only the tooling but also huge labeled data sets from various domains (needed for “training” the AI). These two things together mean that AI engineers can focus on solving business problems with AI, rather than having to spend time creating tools to solve AI problems first. Lastly, the research itself is remarkably open. Most, if not all, of the research is open and not behind any paywalls, despite the fact that it is being done in commercial labs. This openness helps the whole field move forward faster as a whole.
We know AI can recognize faces in images, steer a car and warn us about potential threats in our IT systems, but are these tasks very different? They are in fact the same task, just in different instances. Tasks solved by neural networks these days boil down to being able to recognize patterns in data. We show the computer data (e.g. an image) and tell it what is in that data (e.g. a human face), and with enough data and enough repetition, it will learn to correctly identify in new data it has never seen before. So if you want to know if you can benefit from AI then try to think of places where noticing patterns might be useful, noticing that a cell phone tower acts differently than usual, preemptively prompting users with offers based on their patterns, and so on.
In fact, AI is being used in Telenor quite a few places already! One example is the work to detect anomalies in connection statistics when making video calls with Appear.in. There are many indicators that might signal a connection loss before the caller notices it, which may in turn be used to fix the issue before it impacts, improving the call experience. Another is in the classified ads vertical, where AI-fueled image recognition is now being used to radically improve listings and search for Mudah, Kaidee and imSold! in Asia. Finally, at Telenor Group Forum in Myanmar this year, a customer care chatbot prototype was unveiled, which sparked a lot of interesting discussions. This chatbot was a joint effort between the Telenor Expert Arena experts, Telenor Research and ourselves in Telenor Digital, which stands to illustrate what we can achieve when we combine forces!
There is a lot of potential in leveraging AI to improve the operational efficiency of Telenor, as well as bringing modern products and features to the table. Work is already being done in silos here and there throughout the organization, but there is no bigger picture. The AI know-how from the different efforts can help kick-start new projects, but not if they remain isolated. The next step in order to be able to capture the latent value of our vast amounts of data and knowledge through AI is to unify in a concerted effort to make AI accessible across the board globally in Telenor. AI needs to become a core competency for our business if we don’t want to be left behind.