Roundtable: Does AI dream of electric housing?

AI is a fascinating topic, but as we heard from delegates at Northern Housing’s latest roundtable, what really matters is harnessing the power of data and automation to catalyse profound change

In association with

THERE’S nothing worse than chairing a meeting where nobody wants to speak,” joked Liz Haworth, chair of Northern Housing’s latest leaders’ roundtable, as she called time on the discussion. No worries here, it had proved itself a lively gathering.

Hosted by Central Networks & Technologies, along with Fujitsu, and taking place at Manchester’s Albert Square Chop House, the delegates sat down to enjoy a good meal and explore the myths, realities and potential applications of artificial intelligence (AI).

Delegates were clearly engaged with the topic. Conversation was wide-ranging, as they shared notes, asked questions, and discussed the implications.

Clearly, there was a lot of interest in technological solutions, both in a specific sense, and more generally in terms of the digital revolution’s ongoing potential.

For all the engagement with the topic, however, it ought to be noted that delegates demonstrated a clear-eyed perspective that technology is no automatic solution to life’s every little niggle. There was a healthy degree of scepticism, but not a dismissal of technological potential; no luddites here.

“Fundamentally, artificial intelligence is human-like behaviour from a machine and really what we are all grappling with is automation.” – Brian Moran

“There is a problem with the phrase ‘artificial intelligence’,” said Brian Moran. “It is a futuristic word and it has been futuristic since the 1950s… It was invented because an academic wanted to distinguish it from cybernetics.

“Fundamentally, artificial intelligence is human-like behaviour from a machine and really what we are all grappling with is automation. Now that is the right thing to grapple with, because there’s all sorts of automation we can do, and are doing right now, but not that much of it is AI based at all.”

He added: “I don’t think there are a lot of true uses of AI; I don’t think there are a lot of true uses of AI anywhere in the world actually, in the strong sense, as in a kind of generalised intelligence that can deal with anything.”

AI is certainly a term we hear much about. For many of us outside the sphere of experts who deal with this as their everyday working life, it is doubtless the realm of science fiction that tends to (mis)inform our thinking most.

As several of the delegates pointed out, there is much technology lumped under the AI banner that is more a case of clever programming, or a sophisticated algorithm, than any kind of machine intelligence.

But, “does it really matter?” asked Andrew Rafferty. “The definition of whatever camp [the technology] falls into, I’m not too worried about – is the outcome effective, does it add value, and is it affordable? That’s true, frankly, whether we’re talking about AI or a new network or IT system.”

A lot of ground was covered during the discussion, with delegates firing off examples of what they are doing now, or what they are looking to for the future. There was plenty of speculative talk about what may come about. There was also an examination of some of the ethical implications of technological interventions in tenants’ lives, such as using sensors to gather data on the performance of heating systems, for instance – where does monitoring become surveillance?

As you might expect, there was an interplay between the cost-benefit impact on the business and the social purpose: issues such as privacy, surveillance, GDPR, health and well-being, and more cropped up. Fair to say, there are no easy answers to such issues; society as a whole continues to wrestle with these questions.

Automation, rather than AI, was one of the key goals for the housing providers at the table; they saw in it an important route to smarter working. Several delegates, for example, were experimenting with chatbots for routine interaction with tenants; others were exploring the potential for fault diagnosis systems to assist with repairs calls. Another avenue was the use of apps for appointments.

Later the conversation turned to deeper matters; machine learning and the power of data, and it was here that delegates really felt there was scope for a major transformation in how the sector operates.

Martin Myers shared an example from the world of manufacturing (see case study below) and invited delegates to ponder how it might translate into their world.

Machine learning systems, whether based on algorithms or cutting-edge neural networks, are certainly fascinating developments, but they are not without their pitfalls, as Moran pointed out.

One of the issues, away from the tightly defined parameters of an industrial process setting, is the risk of perpetuating bias – racial or gender, for example. It’s an issue compounded by the supposed neutrality of a machine. At the end of the day, though, it can only learn from the data provided, and the criteria set for it by human creators. So, food for thought, there.

Data, however, remains a largely untapped – and powerful – resource for housing providers and their partners. This is true, whether it is fed into machine learning systems, or simply crunched by good ‘old-fashioned’ analytics. It can really make a difference to organisations and lives.

“In our sector, with the data that we have, we can predict with quite good accuracy where domestic violence is taking place in a property,” said Steve Rawson, by way of example. “It follows a pattern, in terms of anti-social behaviour, in problems with animals, with damage to property. We ran it with West Yorkshire Police, we analysed the data pulling out signs where we believed people were at risk, and the police were following up on that with really good results.

“Our biggest cost as a business is tenancy failure; that is what costs us money. Our biggest asset is the tenants. They are the ones who pay the money, and anything that changes those patterns and those behaviours, we’re pushing to make sure we know.

“We have masses of data across the organisation, and when you start linking that up with partners, such as police and social services… we do far better for the people we house.”

He added: “We need better data to get predictive. We do it very well reactively; we want to move from the reactive model to [solving] problems to a predictive model.”

Ultimately, all this talk of technology, at least in a housing context, is all about people; whether that’s making them more effective at their work or making lives easier for those it provides homes. And the first hurdles often start with senior executives.

“People still think this is technology that’s not been around long, [that] it’s from the Space Age,” said John Roe. “It’s been here years. Senior people don’t see that. Technology is probably not the challenge. It’s convincing your executives it’s the right way to go.”

Neil Pollitt said: “AI is a fascinating area… but if you are talking about chatbots and AI, it turns people off. If you talk about what money we are saving, or what’s the improvement in customer satisfaction that we are getting because of this, then it gets off the ground.”

But it’s all very well winning over executives, or even staff; at the end of the day much of the technological implementation will stand or fall with the tenants. “Customer experience – customers have got to value it,” as Pollitt added.

Rawson shared a similar sentiment: “A lot of people want technology, but they want technology that works for them, that actually makes their lives easier.”

# # #

CASE STUDY: A sharper eye for better blades

WHEN Siemens faced a bottleneck in the production of rotor blades for offshore wind turbines, it turned to Fujitsu UK for a potential solution, Martin Myers explained.

Martin Myers
Martin Myers, Fujitsu UK

As a case study, it’s miles away from housing, but Myers felt it represented a powerful example of how problem-specific machine learning can provide a transformative benefit for an organisation’s operations.

Siemens manufactures the blades at its state-of the-art factory in Hull. Each blade is 75 metres long and made in two halves, which are then chemically welded together. The join represents a point of potential weakness and therefore needs rigorous quality assurance (QA).

“They have a real fear of catastrophic failure of one turbine blade having a cascade effect through the rest of the farm,” said Myers. Offshore windfarms are expensive to maintain, but if a blade sheared and the fragments shattered neighbouring turbine rotors – that cascade affect – the costs would be astronomical.

Naturally, Siemens wants to do everything humanly possible to ensure the quality of the rotors. But the rigorous process was holding back its capacity to increase blade production.

“They have this ingenious ultrasound machine, which takes a scan of the entire seam, 75 metres one way and 75 metres back,” Myers said. “That effectively generates 900 meters of ultrasound scan per blade, which needs to be examined as part of the QA process.”

The problem was the company only had 10 QA technicians to study these ultrasound scans. While they had an accuracy of 94% it took eight hours to assess each scan, so short of a costly increase in human resource, there was little headway it could make.

“Our labs people were working on something called a convolutional neural network,” Myers explained. “This builds a computer model of how the human brain works at a very very simple level.”

Fujitsu was working on applying this neural network as a visual recognition system in several industrial applications, so they set about applying it to Siemen’s needs. They were given two weeks to work up a solution – and pulled it off.

The team were provided with 50 of the ultrasound scans, marked up to indicate flaws, and these were fed into the image recognition engine. It took half a day for this, then a day and a half for the engine to learn what blade defects looked like in the scans. After that it was put to work.

The results were impressive. The neural network scored 98% accuracy and only took one and a quarter hours to process each scan, potentially offering the scope to increase production capacity fourfold.

Myers went on to say that Fujitsu has taken this technology and used it in other cases, such as analysing railway tracks and roads for defects. “In a very short period of time you can take this raw engine and teach it to recognise just about anything,” he added.

# # #

Taking part were…


Liz Haworth, chief operations and transformation officer, Torus


Martin Brelsford, Business development manager, Central Networks & Technologies

Mike Stuart, Sales manager, Central Networks & Technologies

Danielle Masters, sales and marketing executive, Central Networks & Technologies

Gillian Ashworth, sales and marketing, Central Networks & Technologies


John Roe, Group head of ICT, Together Housing Group

Ailsa Dunn, head of insight and innovation, Prima Group

Martin Myers, chief technology officer, Fujitsu UK

Andrew Rafferty, director of technology and business improvement, Weaver Vale Housing Trust

Steve Rawson, director of resource, Wakefield & District Housing

Andy Webb, chairman, Alysium Consulting

Neil Pollitt, assistant director of business intelligence and insight, First Choice Homes Oldham

Brian Moran, deputy chief executive, Jigsaw Homes

Observing: Mark Cantrell, editor, Northern Housing, Crosby Associates Media


Roundtable: The myths and realities of AI
From left to right around the table: John Roe, Together Housing; Ailsa Dunn, Prima Group; Gillian Ashworth, Central Networks; chair Liz Haworth, Torus; Martin Myers, Fujitsu UK; Andrew Rafferty, Weaver Vale Housing; Steve Rawson, Wakefield & District Housing; Mark Cantrell, editor, Northern Housing; Danielle Masters, Central Networks; Brian Moran, Jigsaw Housing; Martin Brelsford, Central Networks; Neil Pollitt, First Choice Homes Oldham; Andy Webb, Alysium Consulting; Mike Stuart, Central Networks



%d bloggers like this: