Skip to content
  • Print

Artificial Imagination

In case you haven't been paying attention, we are in the midst of the "Artificial Intelligence (A.I.) Revolution." Just "ask Alexa," your Google Home assistant or soon-to-be six-year-old, Siri. Having lived through a number of technological revolutions, including very-large-scale integration (VLSI), the personal computer, the Internet, the World Wide Web, Google's PageRank algorithm, and now A.I., we're in exciting times.

Building on and integrating these previous innovations, A.I. has spectacular transformative potential, as tiny embedded machines, ubiquitous in the Internet of Things, learn to recognize, communicate and act on patterns they see in their environment. 

Einstein spoke about knowledge and imagination in the context of scientific research and creativity. However, one could argue (and I'm sure my colleagues in the philosophy department would argue back) that these two qualities are fundamental components of intelligence. How do knowledge and imagination turn up in artificial intelligence? One of the fundamental aspects is the new wifi-enabled, smartphone-connected and on-demand Information Age. It is composed of a vast amount of knowledge - facts, books, images, real time data and information of every possible kind at our fingertips - anytime, anywhere. 

In the late 1980's, less than one percent of the world's stored information was in digital form; by 2007, less than 30 years later, that fraction had grown to 94 percent. The ability to copiously reproduce digital information (text, sound, images, data, and now physical 3D objects), at low or no marginal cost, and to transfer it instantaneously virtually anywhere on Earth is transforming the world.

But the promise of the "Artificial Intelligence Revolution" is more about imagination than knowledge. When my daughter and I lie on the grassy hill near our home and find dragons, turtles and butterflies in the passing clouds, we are using our imagination to find patterns and matching them to items in our knowledge base. Artificial intelligence is about doing something similar, at scale.

Fans of the original Star Trek series will remember the Universal Translator, a device that listens to a speaker of one language, Klingon for example, and repeats the phrase in English. A similar tool is Google Translate, used by over 200 million people daily, in which users type a phrase in one language (or even let Google guess the typist's native tongue), and see the phrase translated into any of more than 100 other languages. The service was created in 2006 and first worked by coding linguistic knowledge into grammatical algorithms. Then it had set up equalities between the information contained in bilingual dictionaries. This knowledge-based approach worked reasonably well, but would often translate phrases such as "minister of agriculture" to "priest of farming." 

About a year ago, Google tried a different approach with computational (or artificial) neural networks. These networks are computer programs that try to simulate the connectivity and interaction of neurons in the brain. Google essentially lets deep learning, multi-layer computational neural networks look for patterns in the way people speak and write in one language, and then match those patterns to similar patterns in the way people speak in another. The result made Japanese Twitter accounts explode with awe as users wondered how Google had suddenly learned to translate Japanese so fluently.

If Einstein was right, and imagination is more important than knowledge, then perhaps we should reconsider the "A.I. Revolution" happening all around us, not as the Artificial Intelligence Revolution, but as the Artificial Imagination Revolution.

Dr. Hetrick is Assistant Director and Professor of the School of Engineering and Computer Science's Master of Science in Data Science program offered on the San Francisco and Sacramento campuses. More information can be found at