Software agents

A colleague recently referred me to an article about the remarkably prescient computer scientist Dr David Gelernter.

In his 1993 book Mirror Worlds he accurately predicted many of the internet technologies and services which we now take for granted - including blogs, search engines, and video streaming [1].

The last unrealised prediction in this book is about ubiquitous software agents.

These are software programs with a very clearly defined purpose, which collect data from a multitude of sources to augment our intelligence and decision making abilities.

An example is a medical software agent which collects data from an array of sensors on our bodies, and information relating to our lifestyle to provide early warning of the onset of medical conditions.

Another example is a software agent which crawls the internet for information about stocks, and provides hints to potential future market changes.

You can imagine each person having thousands, or even millions, of these software agents, each performing their task silently but diligently, providing us with hints and early warnings when required.

This is the sort of artificial intelligence which can be genuinely useful to us, augmenting our intelligence, and making up for our shortcomings as human beings - trawling through mountains of data, and spotting meaningful relationships.

This also ties into the idea around the semantic web, making the internet more machine readable, which would open up the possibility for increasingly powerful software agents.

In parallel to these developments is the constant reduction in the size and cost of sensors and micro-controllers which provide us with more information about ourselves and our surroundings.

There are services out there already such as Pachube, which allow the aggregation of data from a worldwide array of networked sensors.

As a final note, consider these humble, but nonetheless useful, applications these software agents could perform for us everyday.

  1. A software agent picks up our planned destination from a mapping application, trawls the internet for weather reports, and local sensors, and comes back to us with a suggestion to take an umbrella.
  2. Sensors in the toilet pick up trace amounts of chemicals and bacteria in our urine, and compare this to vast lookup tables online to provide early warning of the onset of infections.

And best of all, these networks of sensors could be almost ‘weightless’ and pervasive - so much so that they just tie into the material fabric of the world, creating a more intelligent, efficient, and pleasurable environment for us all.


  1. Seer of the mirror world, Economist