Universal adapter

Nader Azarmi, head of intelligent systems at BTexact, is on a mission to free us from drowning in electronic data using proactive software.

Liberating workers from drowning in a deluge of e-mail and web-borne information is not most people’s first choice of philanthropic goals, but Nader Azarmi is pursuing this with a vengeance.

Ignoring eye-catching and media-friendly visions of a future where humans interact with hyper-intelligent robotic helpers, his vision is resolutely practical. He wants to create a symbiotic man-machine decision-making system by fitting artificial intelligence (AI) into existing programs and processes to make them more efficient and flexible for users.

‘Some people want to build a human/robot hybrid to replace mankind in the future,’ says Azarmi, chief AI technologist and head of the intelligent systems laboratory at BTexact Technologies, BT’s advanced research and technology business. ‘I believe in augmenting humans using artificial intelligence. Why build a robot when you can improve people by using computer systems? A lot of people need help with access to education on the internet. Intelligent computer systems can help with this as well as freeing humans from mundane tasks.’

But what inspired this goal? ‘The initial idea was always a dream of mine,’ says Azarmi, who began his career as a researcher into AI systems. ‘My own experiences of ploughing through texts meant I wanted something that could reduce the number of hours spent searching through information sources. I want to make employees’ lives easier.’ Growing animated, he adds: ‘With my work for BT I feel like I have done the field of AI a real service by proving that it can work, has real value and can change the world for the better.’

To achieve his dream, Azarmi has been combining the classic logic and form-based AI of the past 25 years with soft computing and fuzzy logic to build a system that can cope with the ambiguity of human needs. ‘Old-fashioned AI was very logic based,’ he says. ‘However, humans are anything but logical. The 0-1 system really isn’t sufficiently ambiguous, and we must incorporate something new if we are to aid interaction with machines and build a hybrid technology. I like to think of it as evolutionary computing.’

His quest for this goal has not been idle. Since graduating from Manchester University in 1982, Azarmi has won numerous industrial and innovation awards including, most recently, the Royal Academy of Engineering’s Silver Medal, as well as filing over 50 patents.

Following a background in scientific research, including two years as a visiting and senior lecturer at Kingston University’s Engineering and Computer Science department, Azarmi joined BT in 1989. Today he manages a team of 50 at BTexact. ‘The most important aim of my research is to develop an intelligent system and technology to radically improve group operations and mission-critical resources by looking at optimising technologies through AI and reasoning,’ he explains. ‘We are looking at new areas of artificial intelligence using tools such as fuzzy logic and neural networking that can cope with human behaviour.’

He and his team have already experienced high-profile success following the development of the Intelligent Personal Assistant, which won a 1998 British Computer Society award. Nicknamed the ‘electronic butler’, the system examines users’ interests through their computing activity and an analysis of key words, then uses this to search the internet for news items that might be of interest, before compiling a personalised newspaper on a daily or weekly basis.

A related tool, intelligent distributed information-management systems or IDIoMS, incorporates a contact finder, providing relevant names from within a community of users, such as a certain field of engineering research. It also contains an information finder function based on keyword entry and can pick out and suggest relevant information resources to complement a piece of work.

Meanwhile, Azarmi’s Intelligent Personal Assistant Framework, or iPAF, contains an e-mail tool that automatically classifies e-mails by priority by monitoring the way the user handles mail to adapt its future behaviour. ‘With the advent of the internet, people are suffering from information overload – broadband will only increase this,’ he explains. ‘What we need are systems that will help to deal with this and save workers’ time. There is no point in designing a one-size-fits-all solution – we need adaptive systems that can personalise information depending on the users’ needs, but also be capable of updating these as interests change. This update can even be made to take place automatically, based on an analysis of the user’s reading habits and the like.’

Inspired by the development of fruit flies, the BTexact team has been working towards incorporating biomimetics into computer network defence systems. During their development the flies’ cells must decide whether or not to make sensory bristles. This is done by sending signals to neighbouring cells and responding to their replies, to achieve the right bristle pattern without central control. Behaviour modelled on this self-assembly system has been used by Azarmi and Qinetiq to design a future battlefield communications system.

The system allows base stations in a mobile phone network to decide how available radio frequencies will be divided to meet the demand for calls without causing unacceptable interference. The network can continuously adapt to changes in demand for calls and can compensate if a base station fails without intervention from a central organiser. In the military system self-organisation allows communications to continue to work in the face of disruption such as unplanned movements and accidental or deliberate interference from other radio transmitters. These conditions are very difficult, if not impossible, to manage from central control.

With his working knowledge of the insect kingdom, Azarmi has now turned his attention to the human body in his quest for inspiration. He hopes to address the problems of network security breaches caused by the introduction of portable wireless devices such as laptops and personal digital assistants into the workplace, which can operate simultaneously inside and outside an organisation’s firewall, creating a back door for intruders. Devices equipped with new adaptive network perimeter technology, announced in July, can re-draw a system’s boundaries in real time as its topology changes. If a breach is detected, a chain reaction is triggered in localised nodes, resulting in a new defensive layer forming around the compromised area.

‘We looked at how the human immune system works in order to build security systems for networks,’ he says. ‘As anyone involved in this area knows, a firewall is not enough. If the human body senses an intruder, it singles this out and builds a defence around it to contain it. We are now mimicking this behaviour in software form.’

Another field of interest is the possibility of automatically generated computer programming. ‘The aim is to look at how to avoid writing a lot of basic code and the related debugging processes,’ says Azarmi. ‘If you can build on the same basic system for different programs it saves a lot of time and makes it easier to design a solution. People can tell us their problems and we will take this away and fit it into the program we produce. We can make working more enjoyable for the user by improving their interaction with systems.’

Proving a self-stated desire to produce technologies with good commercial prospects, over the past few years Azarmi has also been helping to exploit technologies outside BT through partnerships with other firms. Success stories include APSolve and its Taskforce software package, which uses AI to allow companies to optimise the scheduling of thousands of mobile workers in real time.

But when will we see a computer that can understand the mind of a sentient being? ‘AI is viewed by science as something of a Holy Grail and understanding human intelligence and mimicking it is a very interesting problem,’ says Azarmi.

‘We know what we want to achieve but not how to do this. It’s a very big challenge. No one has the illusion that it will be a reality in the next five to 10, or possibly even the next 20 years, but it will take a lot of time and effort.’

Given his record so far, while the ultimate system will take some time to achieve, it certainly looks as though there will be plenty of interesting discoveries on the way.

Sidebar: For the Record

In May Nader Azarmi was awarded the prestigious Silver Medal of the Royal Academy of Engineering for his work in Artificial Intelligence.

Before joining BT in 1989 he spent two years working as a visiting and senior lecturer at the University of Kingston while working on a master of philosophy in AI and logical databases. He also holds two MSc qualifications in telecommunications business and AI. He is currently completing a PhD in AI and intelligent scheduling at the University of London.

Among his awards are the British Computer Society IT medal for dynamic scheduling in 1997 and the BT Gordon Radley Fund Christopher Columbus first prize in 2000 for Intelligent Assistant software.

He has successfully applied Artificial Intelligence to a number of BT’s key operations, allowing the automation of routine tasks carried out by over 20,000 BT staff. These include employee scheduling and bill management systems, which have increased productivity and saved BT a significant amount of money. He has filed over 50 patents, some of which underpin software marketed by spin-off companies being nurtured by BT’s incubator, Brightstar.

He has designed intelligent tools that interpret workers’ computing actions and use information gathered by this to find news and research information for them, while automatically adapting to their changing needs as their behaviour changes with time.