Ironman, Terminator, 2001: A Space Odyssey. Artificial Intelligence is still brought to life with tales of superheroes and films.
But at a recent event hosted by BT, Augmented Intelligence was the preferred term, brought to life with examples of man and machine, not man versus machine.
Of course, AI is nothing new, as Dr Detlef Nauck, chief research scientist for data science, BT, pointed out. It’s been used for more than 20 years, and is commonly behind things we already take for granted, like Google Maps telling us how long it’s going to take us to get home, helping block nuisance calls and Amazon suggesting other books we might like.
But globally, it’s also being used in high stakes areas like helping judges assess a suspect’s likelihood of reoffending, recruitment and even predictive policing, resulting in discussions on the implications on security, ethics, governance and regulation of AI.
AI driving business transformation
Dr Robert Hercock, chief researcher in disruptive technologies, BT, said, “Despite the hype, AI isn’t going to go away. It’s driving business transformation.” At the event in London, 75 per cent of the audience said they were already using AI in their multinational organisations.
From predicting maintenance schedules to warehouse safety, oil pipeline protection to temperature anomalies in cold storage units, AI is being used by large enterprises all over the world, with 63 per cent of attendees said they were seeing greatest impact on the process efficiency of their business.
Within cyber security, AI can be used by criminals as a tool to hack into drones and autonomous vehicles, and turn them into potential weapons. But it can also be used by teams protecting organisations to filter through ten million events a second to look for early warning signals of attacks. AI is useful to help protect organisations from attacks they know about, but is less useful when it comes to unknowns. What it can do is bring together sources of diverse data which can be presented for analysts to review and interrogate. As one of BT’s security analysts said. “I’ve got too much data at my fingertips, I need the tools to tell me where to focus my attention first.”
In BT’s own network operations, AI is transforming how we predict service outages. Anomaly detection and predictive analytics have helped us move from hundreds of alarms being monitored by analysts to a prioritised view of those we think are going to cause a potential outage. We’ve also moved from maintenance teams having to manually inspect network equipment for potential failures to capturing photos during routine work which AI tools can use to classify the state of repair and prioritise maintenance schedules.
Other examples included AI being used to cut the fuel bill for ferry companies, lower energy consumption and showing the creation of integrated work orders based on predictive maintenance schedules.
Considerations of AI
But AI isn’t easy. From considering what the base things you need in place are, testing you need to do before, during and after, to training, security, governance and ethical questions, it’s a complex landscape.
Sixty five per cent of the audience said they weren’t doing anything about AI governance, yet Rob Claxton, senior manager in research, big data, insight and analytics, BT, discussed some of the challenges, including understanding the provenance of models, how they work and delegating decisions and actions to AI but not responsibility. Model bias was also discussed, as AI is only as good as the data it has to learn from, yet only five per cent of the audience said their data was in a good enough state for AI. Because AI is evolving and autonomous, governance needs to go beyond the normal software engineering, deployment and lifecycle management. Algorithms that learn and change and run themselves are difficult to audit and as humans we can longer know to any degree of certainty what their rules and parameters are.
Another area which particularly caught the interest of the audience was around the ethics and regulation of AI, with 80 per cent of the audience saying they needed more clarity on regulation. AI is learned from data, has no human readable code that can be fixed and relearning can lead to a completely different system. Whilst there are plenty of examples of AI for good, the potential impact on society through fairness, accountability, privacy and trust issues is already being felt, not to mention on the reputation of brands who get it wrong. Moira Oliver, head of policy and chief council, human/digital rights, BT, suggested starting with understanding where the business is using AI, scrutinising its wider impact, checking compliance with existing systems and monitoring developing external discussions on regulation.
One of the much discussed questions around the impact of AI is whether it is going to steal all our jobs. Nicola Millard, head of customer insight and futures, BT doesn’t think so. She described jobs as jigsaw puzzles, with some aspects easy to automate, and others not. In a world where increasingly people want to chat, she called out Lidl’s Margot the winebot as an example of a chatbot that makes things easy for customers, using a deep set of data about a fairly narrow subject. With people increasingly the differentiator in customer service, the most likely scenario isn’t about man vs. machine – it is man plus machine, allowing us to become more productive, valuable, and more human.
With people increasingly the differentiator in customer service, it isn’t about man vs. machine – it is man plus machine.”
Dr Nicola Millard
Head of customer insight and futures
AI working with people
As we’ve seen, AI can be used to present a set of potential solutions to allow humans to be able to make an informed decision that often neither man nor machine would have been able to find independently. AI understands data, but not context. Humans can’t process the volume of data that AI can.
With no ability to audit or understand how algorithms fit together, Simon Thompson, principal, research, customer experience and AI readiness, BT, believes in the future, AI will have to explain to us what it is doing and why. Certainly in a world that is demanding more transparency over how decisions are being made, being able to ‘show your homework’ will be a crucial part of AI.
AI is necessary to help drive digital transformation, but there are limitations on what technology can do. The example shown at the event was using a voice command on your mobile to “Call me an ambulance”. The response? “Ok, I’ll call you ‘an ambulance’ from now on.”
The expert speakers closing the Augmented Intelligence event agreed that there is more understanding needed on where the best hand offs are between machine and man to play to the strengths of both humans and the machines. There is an important role for AI, but it’s very much an unfinished area of technology.