Generating Intellegence

Simply creating intelligent machines isn't the only goal of scientists anymore, and it shouldn't be.

Scientists are finding an AI's ability to socially interact and learn from human contact is more important to the future of technology and mankind than its ability to serve.


Dr. Lola Cañamero, of the University of Hertfordshire's School of Computer Science explains in this 2008 Science Daily interview,

"If robots are to be truly integrated in humans' everyday lives as companions or carers, they cannot be just taken off the shelf and put into a real-life setting, they need to live and grow interacting with humans to adapt to their environment."

Basic intelligence is the ability to collect and assess data then adapt to changing situations.

From infancy, humans and other animals learn to assess situations through the reactions of others in their group. Negative and positive physical experience also influence the way humans and animals learn.

Training machines to read body language, facial expressions, and search for a desired reaction will greatly improve their ability to learn and adapt.

Let's call these early stages of programming "learning" because eventually these "learning" exercises can be used to develop a type of physical/emotional response in a machine. If AIs can assess our disappointment or joy in the tasks they complete, they will be able to adjust their action accordingly.

The videos below show how these early lessons in adaptation are being developed.

Leo "Learning" Social Referencing

The adorable Leo is "learning" how he should react to a toy the same way children learn from their caregivers.

This is called social referencing which contributes to the "evolution" of emotion.

Children learn to interpret the emotional responses of others.

Eventually, they form ideas about the appropriateness of emotional displays.

Then they learn how to respond emotionally to various situations.

But can the same be achieved when an instructor uses words instead of social referencing for an AIs evolution?

Teaching Leo To Adapt

MIT's Leo seems to be doing fine.

In the Adaption Video, Leo is learning about the concept of "all" as opposed to specific colors, and has adapted his response in this video.

Quite impressive.

Scientists are certainly going in the right direction.

Teaching  and learning may prove to be more productive in the evolution of artificial intelligence than programing.

Copyright (c) 2009 Sangay Glass All rights reserved.