The computer as we know it is dead. The monitor and keyboard will soon be replaced by Orwellian technology that watches, responds to, and even predicts our actions and needs.
This is the vision of the next step in computing painted by research groups such as Intel, Philips and BT. They predict that benign, hidden computer intelligence will soon be all around us in our homes, cars and workplaces, and have spent millions developing the concepts underlying this vision.
However, leading engineers have questioned whether technology can meet the ‘unrealistic’ visions of research laboratories, and have criticised the lack of focus on the engineering challenges.
The already abstract concept of an all-round computer environment, or ‘ubiquitous computing’, is further clouded by the many terms companies use to push their own brand of what is essentially the same technology.
The term ‘ambient intelligence’, coined by an EC advisory group on IT in the late 1990s and used by Philips and BT, is becoming the accepted European term for this technology; whereas US groups, such as Intel and IBM, talk of ‘ubiquitous’, ‘calm’ or ‘pervasive’ computing.
While their terms differ, US and European groups are both elaborating on an idea first put forward by scientist Mark Weiser of Xerox PARC, who in the early 1990s wrote a series of seminal papers sounding the death knell for desktop PCs, and predicting a ‘third wave’ of technology that would make computers retreat into the background.
But there are great technological challenges in creating that third wave, and Prof Andy Hopper, head of Cambridge University’s computer laboratory, believes companies have pushed aside engineering problems in favour of promoting way-out applications.
Hopper, who has led a team working on sentient computing and intelligentsensors since the late 1980s, said not enough work is focused on the backbone of ambient intelligence: the sensing, modelling and processing of the world.
‘The field is working on science fiction ideas but few people are concentrating on concrete, engineering-based projects that deliver something reliable, scaleable and dependable, and that interprets information correctly,’ he said. ‘Not many of the labs’ claims and suggested applications are realistic.’
One of the key aspects of ambient intelligence is ‘context-awareness’, so to achieve the technology’s full potential, cheap, embedded wireless sensors, actuators and displays will be needed. Today’s technology is simply not up to the job in terms of size and power.
As chips get ever smaller, clever engineering continues to follow Moore’s Law â€” that processing power will double every 18 months as new chips with twice the capacity replace their predecessors â€” but progress in this area is getting more and more expensive and difficult to maintain.
Sensor networks, for example, need vast improvements. Measuring temperature or vibration is straightforward, but tracking people in 3D in everyday life is difficult, because of the heavy processing needed to filter out unwanted information.
Mapping the world to integrate with sensor data is a major stumbling block, according to Giles Lane of London-based technology think-tank Proboscis. ‘Visions coming out of research labs see the world as a simple place,’ he said. ‘The only way to map the world is to simplify it; you just deal with the “home”, “street” and “park”. But that doesn’t correlate with the messy world we live in.’
Power is also a problem for an ‘always- on’ sensor and actuator network, because the nodes must either have enough battery power to operate until replaced, or be able to scavenge power from their suroundings. A number of groups have made innovations in this area, such as Intel’s work on tiny sensors known as ‘smart dust’, but a breakthrough is needed to make the technology cheap and reliable for everyday usage.
Wireless transmission is also far from perfect, with soundwaves, electrical cables or water pipes interfering with signals, especially in old buildings. Seamless crossover between networks such as Bluetooth and Wi-Fi is also essential for the true ambient applications to be achieved.
Prof Tom Rodden, director of the EPSRC-funded Equator group, consisting of eight research groups and UK universities examining interaction between the digital and physical world, said that technology will soon catch up with the advanced concepts.
‘Ambient intelligence raises fundamental technical and social challenges,’ said Rodden. ‘The more polemic visions are naive, but many researchers are more grounded.’ Rodden said that research is shifting towards integrating sensor networks, processors and actuators.
Dr Steve Wright, BT Exact’s head of long-term research, admitted that there are many futuristic visions, but innovative use of known technology in the shorter term will cause ambient computing to creep into the real world, leading to an intelligence breakthrough further down the line.
Clever use of simple sensors can provide a wealth of information, he added. BT’s recent ‘wellbeing’ healthcare project, for example, used cheap vibration sensors on water pipes to build up a picture of household activity, which it is hoped could be used to monitor and predict behaviour.
‘The first benefits of ambient computing will come by taking information from simple sensors using smarter data analysis,’ he said. ‘If we focus too much on long-term ambient “intelligence” we can lose a lot along the way. For now it is a question of trying to do what we can with simple information.’
Over the past few years groups like BT and Philips have focused most of their efforts on sociological studies investigating whether the public will accept and use ubiquitous computing. BT’s study concluded that its scenarios for ambient intelligent environments needed to be more grounded and applicable to the average person’s needs.
If ubiquitous computing is to become a reality, the same scrutiny will be needed to define the technological challenges, said Hopper. ‘The engineering underpinning the technology and dependability issues need more attention. Until that happens we will continue to have laboratory demos, not actual systems.’