Despite some experts’ developing theories on the impossibility of our ever creating “strong A.I.” — that is, the kind of robot intelligence that we need to worry about getting away from us and eliminating us as threats to itself (ahem Skynet ahem) scientists out there are still plugging away at this fascinating issue.
One way to potentially solve the problem of achieving human-like consciousness is to overhaul the way machines learn, making it more like the method used by human babies and children. At the moment, many machines learn rigidly, systematically testing new input against a vast amount of information already stored. Flexibility in learning, however leads to very fast gains in intelligence, as anyone who’s ever observed a child grow from birth to age four would know! Researchers are now quantifying this human process — a statistical evaluation called Bayesian learning — and applying it to A.I., attempting to reduce the mass of data and time required to gain the same knowledge about the world.
“The new AI program can recognize a handwritten character just about as accurately as a human can after seeing just one example. Using a Bayesian program learning framework, the software is able to generate a unique program for every handwritten character it’s seen at least once before. But it’s when the machine is confronted with an unfamiliar character that the algorithm’s unique capabilities come into play. It switches from searching through its data to find a match, to employing a probabilistic program to test its hypothesis by combining parts and subparts of characters it has already seen before to create a new character — just how babies learn rich concepts from limited data when they’re confronted with a character or object they’ve never seen before.”
The University of Auckland’s Bioengineering Institute is taking this trend in a startling direction, as it works with “BabyX.” BabyX is an A.I. interface that is a 3D animated blonde baby, who can interact with researchers through a screen, demonstrating the real thinking and learning process of the machine intelligence behind it. The interface is essentially one big metaphor for the learning machine, with a bundle of fibre-optic cables as a “spinal cord,” connecting outside input to its “brain.” So BabyX learns by responding to its user as a real baby would to a parent.
“‘BabyX learns through association between the user’s actions and Baby’s actions,” says [project leader and Academy Award-winning animator Mark] Sagar. ‘In one form of learning, babbling causes BabyX to explore her motor space, moving her face or arms. If the user responds similarly, then neurons representing BabyX’s actions begin to associate with neurons responding to the user’s action through a process called Hebbian learning. Neurons that fire together, wire together.’”
All this work really goes to show that something so natural and seemingly simple — the infant human learning process — is actually really complicated, and hard to replicate for a machine. It will be very interesting to see how BabyX, and this new kind of A.I., “grows up” with us.