For the longest time, robots with artificial intelligence have been for the most part limited to goal-seeking robots that at best might find a charging station. There have been chatbots, the winners of the Loebner Prize, that for 30 seconds at a time have fooled the judges, but these have not to any extent been hooked up to the hardware of a robot. Even the mighty Asimo, Honda's $100,000,000 (that's a HUNDRED MILLION DOLLARS!) publicity stunt, while it can climb stairs, walk in circles and wave at the audience, has no intelligence to speak of. Sure it can detect movement and direction of objects, voice commands, responds to its name, and recognizes faces, but so can a dog. For the most part Asimo is run by a legion of hidden technicians. "Pay no attention to the geeks behind the curtain." As a matter of a fact, an open-source robot project called "Leaf" can, excluding walking, do nearly as much as Asimo. For a cost of less than $2000, one can build a "Leaf" robot that is just as smart, and perhaps smarter than Asimo. While leaf does not have a full staff of dedicated engineers, leaf DOES have an international group of about a dozen part-time, but very active programmers, engineers, and researchers building what might just become the world's most advanced robot. It is certainly now the world's most advanced home robot.
In this article, Robots with Artificial Intelligence, I will explore AI, that is, (robots and) artificial intelligence as it was, is, and will be integrated with robots. You see, I have built a "Leaf" robot called "Robot Betty9" and I'm building new AI software routines into her every day. Still, she is less than a chatbot. More about this later.
ARTIFICIAL INTELLIGENCE AND CONSCIOUSNESS IN ROBOTS
For decades, artificial intelligence researchers and scientists have been laboring to create truly intelligent computers. While progress has been achieved in some specialized areas, the goal of creating conscious machines seems to have remained elusive. Though robots today are truly remarkable in their technology, the creation of intelligent machines like those that have populated so many science fiction books and films since the early fifties is still very much a pipe dream. One clear and important reason for this is that the nature of the human mind is still only partially understood, and this mysterious human mind must serve as the model for robots with artificial intelligence.
“Consciousness is perhaps the last remaining mystery in understanding what it is to be human,” Owen Holland, a scientist at Essex University said. “By attempting to build physical systems which can produce a form of artificial consciousness, we hope to learn more about the nature of consciousness.” Holland is only one of many scientists whose dream is to build a “conscious robot.” For him, succeeding in this work will not only bring intelligent, self-aware machines a step closer to reality, it could also provide valuable clues on how human consciousness develops.
The reason modern robots today are not completely regarded as truly intelligent is because although many are amazing for sure, they are still only capable of following instructions. They do not have the capability to “think” for themselves. Holland's idea is that a conscious robot would have to build up internal models: one for the robot’s “self” and another for the world around it. These models would then be the basis from which the robot would build experience. Much like a human, the robot will learn by experimenting with how its own body reacts with the world around it, figuring out for itself what is beneficial to it and what is not. Robots with artificial intelligence have a long way to go.
But how can we verify that an intelligent being is actually self-conscious? And more importantly what is consciousness for machines? In 1950, computer science pioneer Alan Turing posed a similar problem concerning intelligence and proposed a famous test, known as the Turing test. He wanted to find out if a machine could be considered as intelligent as a human being. Two separate keyboards are connected, one to a computer, the other to a person. An examiner types in questions on any topic he likes, and both the computer and the human type back responses that appear on the respective computer screens. If the examiner cannot reliably determine which answer came from the person and which from the machine, then we say the machine has passed the Turing test. Today, no computer can pass the Turing test, unless the interaction is limited to topics that can yield very specific responses, like chess. Robots with artificial intelligence it seems have a long way to go.
Although most advanced robotic systems intend to model humans, most of the time they did not, like humans, acquire their knowledge through interaction with the real world. Holland and his team of researchers plan to build a robot which will be presented with a succession of increasingly complex environments which will eventually contain moving elements and things that will respond to the robot's movements. Holland said that the robot will have to model and predict what they're going to do and these things will have either positive or negative values for the robot. Although creating consciousness in a machine seems to simply mean producing better robots at some level, conscious robots would be able to react dynamically to their environment by adjusting their behavior on any information they could gather. This means seeing robots that don’t blindly follow their programming, no matter how intelligently it’s written, e.g. Deep Blue, the first ever “intelligent” computer to beat a world chess champion. But chess does not demonstrate robots with artificial intelligence, and perhaps any intelligence at all.
At present in the field of robots with artificial intelligence, we have a scope of researchers working on programs that aim to make robots learn and adapt. At the University of Tokyo, researchers are using a learning methodology they call interactive teaching to give robots the ability to drive their own development. The robot uses Bayesian Networks to map sensor evidence to behavior, and then assigns each mapping a confidence rating. With practice, the robot requires less intervention from the human trainers and eventually a task can be completed autonomously. Scientists have also begun to use attentional mechanisms to focus humanoid robots on visual areas of interest. Learning systems that allow a robot to autonomously recognize and categorize the environmental elements it extracts have begun to be developed. These robots are equipped to recognize perceptual clues such as sound, movement, color intensity or human body language (pointing, gazing and so on), but are they really robots with artificial intelligence?
We have robots like NOMAD (Neurally Organized Mobile Adaptive Device), developed by researchers at the Neurosciences Institute in La Jolla, who is a “real thinking robot who starts naive and learns from experience.” NOMAD’s behavior is controlled by the activity of simulated brain cells, and has already allowed researchers a “unique window to better understanding how the human brain works.” NOMAD can interact with its environment by sensing light and taste and by moving around and grabbing play blocks with striped or spotted patterns. And then we have robots like Adam, whose authors from Aberystwyth University in Wales report could “reason, formulate theories and discover scientific knowledge on their own.” At Cornell University in New York, researchers have reportedly designed a computer program that can uncover the fundamental physical laws behind a swinging double pendulum, and test out Issac Newton's laws of motion as it does so. NOMAD is a step forward in the search for robots with artificial intelligence.
To be sure, these modern robots are “intelligent” but it is simply not possible to verify the presence of consciousness in another brain, whether human or artificial. This is a property that can only be verified by his possessor, and since we cannot enter in another being’s mind, we can never make assumptions about his consciousness. We could, however, follow Turing’s approach and say that a being can be considered self-conscious if he is able to convince us that he is human. Of course this is precisely the purpose of chatbots (or chatterbots), which are designed to simulate intelligent conversations with one or more human users via auditory or textual methods.
So how do we know if we have finally succeeded in making truly conscious and thinking robots with artificial intelligence, such as HAL 9000, the main character in the 1968 epic film “2001: A Space Odyssey?” Believe me, we’ll just know.