Efforts to combine the two foundations of AI may help systems express themselves. Arthur Miller describes in SCIENTIFIC AMERICA how new avenues of research are pairing logic and learning.

The idea is not unlike how toddlers acquire skills: first, they learn, then they reason based on what they’ve learned.

All AI systems work in one of two fundamental ways:

  1. Using logic – preprogrammed computer code instructs operations step-by-step, known as symbolic computing.
  2. Using learning – giving computers many examples from which a system figures its next steps, known as machine learning.

Researchers want to know what happens when the certainty of logic-based systems interacts with the serendipity of learning systems. Might originality be the result? Someday, the outcomes might even be considered to be ‘imaginative.’

Miller is the author of The Artist in the Machine: The World of AI-Powered Creativity, and a Professor Emeritus at University College London.


  • News autoscribe systems work with logic-based systems. They use pre-determined data sources and rules to convert the data into formatted stories. Creative AI systems could one day help move beyond this automated journalism model, perhaps to synthetic journalism, narratives that more closely emulate free-flowing human composition.
  • It’s likely journalistic needs for accuracy and verification will slow the use of creative AI systems for news. Still, advances are important to track for the longer-term possibilities.


Creativity and AI: The Next Step
SCIENTIFIC AMERICA | October 1, 2019 | by Arthur I. Miller