“To avoid the fate of alchemists, it is time we asked where we stand.  Now, before we invest more time and money on the information-processing level, we should ask whether the protocols of human subjects and the programs so far produced suggest that computer language is appropriate for analyzing human behavior:  Is an exhaustive analysis of human reason into rule-governed operations on discrete, determinate, context-free elements possible?  Is an approximation to this goal of artificial reason even probable?  The answer to both these questions appears to be, No.”

Hubert L. Dreyfus
“What Computers Can’t Do:  The Limits of Artificial Intelligence”

This chilling conclusion about the fate of artificial intelligence seems to put an end to the idea that we can automate innovation.  Since this book was first published in 1972, not much has changed, and the  field of artificial intelligence seems to be in decline.

For a machine to innovate, it would need to:

  1. Take a product or service and break it into its component parts
  2. Take a product or service and identify its attributes (color, weight,  etc)
  3. Apply a template of innovation to manipulate the product or service and change it into some abstract form
  4. Take the abstract form and find a way for humans to benefit from it

I like the odds of a machine being able to do the first two steps.  Imagine a computer that had the ability to “Google” a product or service to create a component list.  Try it yourself.  Search Google for “components of a garage door.”  You should be able to find several websites from which a component and attribute list could be developed.  There are lots of Web resources available to machines to derive lists such as patent filings, engineering specifications, instruction manuals, etc.

At Step Three, a computer could be programmed to spit out new embodiments of the original product that have been altered by templates.  For example, it could  apply a template like Division to the garage door.  It could create a matrix of internal and external attributes and spit out potential dependencies between them using Attribute Dependency.

Step Four is where machines struggle.  How would a computer take an abstract “solution” and work backwards to find novel and beneficial aspects of it?  What level of intelligence would it need to search the total human experience and match that solution to an unsolved problem of the human species?  Is it possible?  Not according to Dreyfus.

What if the machine could come close enough in Step Four?  Imagine a machine that could suggest some reasonably good guesses where to take the pre-inventive form to create a new product or service.  Invention Machine’s Goldfire, for example, pulls together information from multiple sources and leads people to find ideas.  It does the preparatory work, but you have to do the rest.   It does preparatory work, by the way, better than humans.  It gives humans an edge in innovating.

Humans are safe from machines taking over innovation.  But they are not safe from themselves.  Maybe we are approaching this the wrong way.  Instead of trying to make computers more human-like, perhaps we should focus on making humans more computer-like, more logical and systematic when innovating.  How can we help humans overcome their humanness to innovate more effectively?  By perfecting the use of innovation tools and processes in a disciplined, rigorous way.  That is a legitimate path to automated innovation.