Unfolding plans 24 – the true test of artificial intelligence is for it to deliver something unpredictable

It’s an interesting concept.  How do you test artificial intelligence?  It is a concept that is difficult to define yet we all can understand that human intelligence is different from so called artificial intelligence.  I’ve already said in a previous piece that I struggle with the definition of artificial intelligence, or machine intelligence as all technology has initially been developed by humans.  Everything that machines can do today is of a human construct.

It is clear though that there is a difference.  In the machine world logic prevails.  An action, providing the kit is working correctly, will lead to an outcome time and time again.  In the human world logic will often go out of the window. An action will lead to many possible outcomes depending upon the people engaged and the way that they are feeling.  Context prevails.

There was an interesting article published on the BBC’s website this morning about this very thing. Professor Gary Marcus from ‘Beyond the Turing Test Workshop’ said that the Turing Championships would host a series of events to test different parts of what defines intelligence.

‘We are trying to figure out a way of evaluating real progress towards artificial intelligence, not the kind of narrow progress where you build a computer programme that can do one thing.’

An example of a test that could be included is requiring a machine to assemble flat-pack furniture from a diagram to which some wit had added on Twitter that a definition of the human condition is to ignore flat-pack instructions and just dive in.  We’ve all been there.

Much insight can come from humour.  There is many a true word spoken in jest.  What was said was insightful though in that what we admire from the machine world is its consistency and its accuracy.  It always does what it is supposed to.  Yet what we admire most form the human world is our unpredictability, our creativity and our ability to think through different alternatives.

To be human is to break the rules.

In fiction our most loved characters are those that challenge authority, who live at the margins of society, who take risks and do the things that no one else will dare.  Our heroes are anarchic and push their way through the things that stand in their way to achieve what they know to be right.

I was at a conference today and Andy the technical director of the host company suggested that we should automate everything.  I agreed, it is a commercially logical approach but then it panders to a machine approach.  Automation means that the result is predictable.  The person with need can click a button and get the outcome that they want or at least what they think they want.  To automate everything assumes that every action should lead to a defined outcome yet this is not human.  Absolute predictability is not what we are made to do.  To err is to be human.

You cannot automate innovation.

Perhaps the true test of artificial intelligence is for it to deliver something unpredictable, creative and exciting.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

w

Connecting to %s