Applications with a human touch

Future watch: Synthetic characters

Synthetic characters, together with complementary technologies, will result in software agents and robots capable of learning how to serve us better, or how to provide us with exactly the information we need, based on their experiences of interacting with humans. Equally important, a synthetic agent could be the only software that we must program only once, after which it would essentially update itselfSome people are easier to get along with than others, we all know that. Their character makes the difference - that indefinable mix of qualities that can inspire repulsion or appeal. But what has this got to do with computers?

Researchers at Massachusetts Institute of Technology and elsewhere are developing synthetic characters, software agents, and robots that not only can solve problems in the traditional way through programmable rules but can adapt to new inputs, learn new behaviours, and even display emotions. Think automated customer service with an electronic good attitude. Or cyberdoctors with impeccable bedside manners. Or actors who don't merely play Hamlet or Ophelia but "are" those characters. And what character would you like for your synthetic butler, maid, or gardener?

One thing computers are good at is answering questions that are precisely defined and belong to a carefully etched context. For example, today's software agents can find price information from an online catalogue or select a range of flights to your destination. Generally speaking, computers work well in situations where a question can be easily translated into a database query, according to a predefined structure. This means that given the current level of technology, computers can respond only to questions that are anticipated by their programmers, with answers that are prebuilt into the system.

There is clearly a wide range of situations in which more interaction is needed, such as helping customers find a product that better satisfies their need, diagnosing a patient, or providing interactive training to users.

For these relatively open-ended situations, today's computers are woefully inadequate. Consequently, whenever scripted interaction is impossible or inappropriate, there's no substitute for good, old-fashioned human dialogue.

Synthetic characters aim to deliver such human-like helpers, but without the potential moodiness and inevitable coffee breaks. One of the most fascinating and ambitious new technologies, synthetic characters encompasses and builds on the results of several other disciplines, including natural-language processing, psychology, and physiology, to name a few. It is also the foundation for developing intelligent and autonomous robots.

Although thought-provoking, current commercial applications of synthetic characters still suffers from limited capabilities to understand context or to adjust to sudden switches in context and to make associations, skills that are crucial to human interactions. Technological breakthroughs in these areas are necessary for artificial intelligence to even approximate human thinking. Some experts believe that a major advance will result only from a new generation of computers capable of thinking processes that simulate the connections among neurons in the human brain.

Although software agents capable of humanlike learning and behaviour are in the works, they probably won't reach the market for another six to 10 years. So put away wishful thinking and get to work!

Join the newsletter!

Error: Please check your email address.

More about Massachusetts Institute of Technology

Show Comments

Market Place