In popular culture we see artificial intelligence in many movies and novels such as "The Terminator" and Phillip K. Dick's novel, "Do Androids Dream of Electric Sheep"; but our dreams of AI to that scale are still a long way off. In this article we look at what driving forces have led to humankind's desire for AI and what advancements we've made over recent years as we edge ever closer towards our goal.
Artificial, not natural; imitating nature
Artificial Intelligence, (use of) computers replacing human intelligence
Intelligence, (noun) intellect; quickness of understanding; collecting of information. esp. secretly for military or political purposes; information so collected; people employed in this
Since the beginning of time, nature has evolved naturally through the survival of the fittest. This has always been so, and also every muscle movement of every animal has to be learnt through practice and trial and error. This can be seen in young children as they are learning to walk or to read and write; they need to practice. Some activities though are not a learned process but take place through our own instincts - something which all animals have; we don't need to teach people how to breathe or their heart how to beat - this is part of living.
When looking at neurones inside the brain it is evident that new connections are made between them as you learn something; this is an important fact that comes is essential to how we understand learning and how we consider the construction of an AI.
Artificial intelligence used to be a thing of science fiction until more recent years when more and more people have taken interest in the "evolution" of AIs. An example of this is RoboPike, a robot fish that was developed at MIT after scientists realised the inefficiency of submarines and so started to look at nature for clues on how this could be improved. Real pikes create a force to the side of them with their tail, so that they can move themselves sideways to propel them forwards. Exact measurements were taken of the pike and the data on how the pike moves was fed into the RoboPike's processor. Once it is switched on and in the water it will navigate the water itself with no human intervention. It demonstrates a simple form of AI, one which uses sensors and logic to determine the use of motors, but nevertheless it is still artificial intelligence.
Here at the MIT Towing Tank we are building robotic fish. The aim of these robots is to help us learn more about the complex fluid mechanics that fish use to propel themselves. In the long run, propelling small autonomous vehicles via fish-like swimming could have enormous energy savings, and increase the length of time a small machine could swim. In the mean time, by studying and building these robots, we hope to answer Gray's paradox, which is that fish don't seem to have enough muscle power to propel themselves at the speeds that they do.
-- John Muir Kumph, MIT
A similar example is a Mars lander which whilst being worked on was tested in hostile conditions such as Antarctica. It was designed to withstand the cold and hospitable conditions one would find on Mars, but this isn't why it has an AI. It was, or had, an AI because a signal coming from Earth directing it where to go and how to react would take approximately 20 seconds to reach the lander - by this time the lander would have moved further away from where the mission control would intend it to go and could even be lost if they saw a danger too late. That is why they decided to develop and AI system for the lander so that it can analyse it's surroundings to explore by itself. If it detects from onboard sensors that it has reached a sudden drop, i.e. the infra-red beam it sends takes considerably longer to reflect back from the front sensor than it does from the rear sensor then it is safe for the AI to assume that there is a drop in front of it and will know not to proceed in that direction. Whilst the lander explores the Martian surface it takes samples from the soil and if it finds anything the AI perceives as "interesting" then it will stop to examine the soil and rock samples, and possibly even drill if the situation demands it.
Human beings make decisions in response to their environment. How do you encapsulate this behavior into a rover, or a robot, sitting on a planet millions of miles away? That's what we are working on. We want to put software on rovers to give them the capability to be artificially intelligent.
-- Kanna Rajan, NASA
When people say that computers are "clever" they are referring to their ability to perform calculations with great speed. In truth, the computer's ability to do this has to come from the engineer's ability to design an efficient processor. This processor is using "switches" in order to perform these calculations which is what we call binary arithmetic. So, to a certain degree of truth you can say that a computer is "clever" however it would be more appropriate to say that it is fast. When it comes to recognition the computer is very poor at calculating what different objects are, this is where Neural Networks come in.
