The real intelligence behind AI

Artificial Intelligence (AI) is revolutionising the way we live, work and do business. 

As with many software-based new technologies, it’s not always obvious to the casual observer that AI is involved. But if you’ve used Google maps to find the best route across town in rush hour, or you’ve asked Alexa for a quick pizza recipe, or even if you’ve followed a recommendation on Netflix, you’ll have experienced AI in action. 

We’ve been involved in AI since the early days, not only helping to push the technology forward, but seeking ways we can harness the power of AI right across our business. For instance, can self-adapting scheduling help allocate engineers’ time more effectively by planning their routes while they’re on the road? How can AI use historical networking data to optimise our fibre rollout? Will AI be able to spot network faults before they manifest themselves? 

What is AI?

Artificial Intelligence (AI) is the ability of a machine to display intelligent behaviour.  

By ‘intelligent’ we mean an activity that we would normally expect a human to perform: reasoning about a problem, explaining a decision, understanding spoken language, recognising images or faces, and so on. A machine can perform some of these actions by perceiving, analysing, and adapting to data about its environment. 

But do these abilities make the machine truly intelligent? Or is it just faking it?  

So far, all AI is essentially Narrow (or Weak) AI – it can perform a single task very well. But that’s all. General AI is the next level, which remains hypothetical for the moment. And that’s AI with true human-level intelligence, the ability to understand or learn any intellectual task that a human being can do. For the time being, General AI is the stuff of science fiction. 

All current AI systems are essentially software (albeit sometimes embedded into hardware, as in robots), and that leaves them open to bugs and attack like any other computer program. And as many of today’s AI systems rely on foundations of data, their performance ultimately depends on the quality of that data. 

What are we doing with AI?

Readiness

AI isn’t just something you buy off the shelf. As a business, we need to be prepared to implement AI applications so we work out which data, compute infrastructures and tools we need to put in place now. Our projects cover concepts for an end-to-end AI architecture; how we can employ AI to manage our networks; and how to detect anomalies in networks and systems that could lead to faults.

Awareness

We’re aware of the pitfalls for the unwary organisation that treats AI as a niche interest. To successfully deploy AI at scale, there needs to be a broad understanding of the opportunities, skills needed, and risks of using AI systems. That’s why we run training events like the ‘Data Science and AI Week’; we design training materials; we hold AI masterclasses; and we consider how to use AI ethically and in line with legislation and regulation. 

Future

AI does not stand still. 

There have been some amazing developments over recent years and more are on the cards, fuelled by our collaborations with academic institutions like MIT.  

Amongst other things, we’re looking at: 

  • new types of machine learning 
  • how to create secure AI 
  • how to spot and stop exploits of AI systems (like tricking Deep Networks with manipulated images)