Last Modified:                                                                                                

Homepage / Publications & Opinion /

Peter Cochrane's Uncommon Sense: Intelligent machines
What's the big problem - technology or human interaction with technology?

Every now and again the IT industry seems to hit a lull where nothing much exciting seems to be happening. I think we're in such a lull right now. Many consider our present range of devices positioned well for the future, with form factor and performance that will see us through the coming decade. I don't agree.

I think what we have is rapidly becoming boring. For sure, it is magical in its performance and abilities but nothing much fundamental has changed for the last five years. Yes, our devices have become smaller and displays are bigger, and interfaces have become a little friendlier but the reality is that the PC, PDA, mobile phone and laptop are pretty much the same devices as those we had five years ago. The question is: for how long will this continue?

In the past year I have seen chip technologies that will reduce power consumption ~10-fold and new system design techniques that will give further reductions by a factor ~x6. There are also new antenna technologies that are ~x4 more efficient, displays that require no back lighting and data storage devices that are ~x100 denser than today. Then there are fuel cells with ~x20 the energy density, and interfaces that are almost human.

So we should think of technology progress and change in several dimensions. We can only imagine the benefits devices that are 10 x 6 x 4 x 20 x 100 = x480,000 more powerful, or x100 smaller than today, for the same equivalent functionality.

The practicality of really small devices may of course be defeated, in some sense, by the size of our fingers, the wavelength of sound and the distance between ear and mouth but there is much more to come. Devices capable of storing every movie and music track ever recorded are not beyond the realms of physical possibility and will ultimately arrive.

If we factor in the long-term impact of exponential increases in chip performance and bit density, we are looking at more than x1bn over the next 30 years. At some point we will hit the physical limits currently designated as Moore's Wall. Here our semiconductor technologies come to the end of the road as we approach the physical limits dictated by the size of individual atoms and their subatomic components. Our best guess is Moore's Wall will occur sometime around 2020. But, even by 2006, Intel is predicting its microprocessors will be clocking at 10GHz - x3 faster than today in just three years.

So we can anticipate far greater rates by 2020 and memory capacities measured in petabytes - perhaps a time when the difference between RAM and hard drive will have been eliminated.

Beyond 2020 and Moore's Wall we see a lot of uncertainty but an even bigger range of options than today - and far more technology riches than we have enjoyed to date since in the invention of the thermionic tube by DeForest in 1915. I think it is reasonable to assume devices at least x100bn more powerful than today.

What would such devices bring us? I think they will be machines of great intelligence with location and state awareness, cognition and contextualisation, with the ability to continuously and automatically configure to a dynamic world. They will anticipate our needs and operate to our specific and individual benefit. And hopefully they will take away the tyranny of the GUI.

Today I travel the planet reconfiguring my mobile phone and laptop for different operating bands, carriers and regimes. By and large none of the functions are fully automatic and switching between hotel, home and company LAN, wireless LAN, dial-up modem and other modes is a pain. As I visit companies, universities and individuals nothing is easy.

If only it were all automated and I could just switch on and go. I really do not need to get involved in selecting the network and setting up the devices to establish communication. And if only all my files and support systems could be automatically updated as I travel. How much more efficient I would be.

By 2020 I expect all our devices to be making intelligent decisions about steering messages across a room or through a building instead of the dumb routings of today. I expect location-based activities and state to be subsumed into my devices so they can make sensible decisions about my travel, work and communication. I need far less overload and far more effectiveness and only intelligent machines can give me that. I need technology to augment my existence, to detect when I am tired, hungry or on a roll, so I can be automatically steered towards the right activities and decisions.

At a modest estimate my work output has been increasing ten-fold every decade and it has now started to stall. The principle reason is that the human aspects of technology are not advancing fast enough. Without intelligent machines I cannot increase my augmented intelligence and output.

Will intelligent machines sell? Look at the sales of intelligent vacuum cleaners, lawn mowers and robotic pets for the home. People empathise with animals and anything remotely intelligent, even if it is technology!

This column was dictated after an encounter with a really dumb ATM machine that refused to return my credit card for a full five minutes. My PA typed it a day later, I picked it up in a London coffee shop via a Wi-Fi link and, after revisions, it was on its way to via the same facility.