Artificial Intelligence: The Basics

Kevin Warwick

• Editorial: Routledge
• Año: 2013
• ISBN:
  9781136629822
• Edición: 1
• Categoría: Philosophy
• Categoría: Philosophy History & Theory

'if AI is outside your field, or you know something of the subject and would like to know more then Artificial Intelligence: The Basics is a brilliant primer.' - Nick Smith, Engineering and Technology Magazine November 2011

Artificial Intelligence: The Basics is a concise and cutting-edge introduction to the fast moving world of AI. The author Kevin Warwick, a pioneer in the field, examines issues of what it means to be man or machine and looks at advances in robotics which have blurred the boundaries. Topics covered include:

• how intelligence can be defined
• whether machines can 'think'
• sensory input in machine systems
• the nature of consciousness
• the controversial culturing of human neurons.

• 
  
  

Exploring issues at the heart of the subject, this book is suitable for anyone interested in AI, and provides an illuminating and accessible introduction to this fascinating subject.

WHAT IS INTELLIGENCE?

SYNOPSIS

Before embarking on a tour of an artificial form of intelligence, here we take a look at what intelligence actually is in humans, animals and machines. The important aspects of mental make-up are considered, some myths chopped down to size and comparisons are made between intelligence in the different entities. For example, what is the intelligence of a spider? What does it mean for a machine to be intelligent? How would human intelligence be regarded by an alien? Clearly the subjective nature of intelligence is important.

DEFINING INTELLIGENCE: AN IMPOSSIBLE TASK?

It is important, before looking into ‘artificial’ intelligence, to try to understand what exactly intelligence is in the first place. What do we mean when we say a person, animal or thing is intelligent? In fact, everyone has a different concept based on their own experiences and views, dependent on what they think is important and what is not. This can easily change – what may be deemed to be intelligent at one time and place may not be so deemed later or elsewhere.

As an example, in the New English Dictionary of 1932, intelligence was defined as: ‘The exercise of understanding: intellectual power: acquired knowledge: quickness of intellect.’ Clearly, at that time an emphasis was placed on knowledge and mental speed, with a leaning towards human intelligence. More recently, the Macmillan Encyclopedia of 1995 stated that ‘Intelligence is the ability to reason and to profit by experience. An individual's level of intelligence is determined by a complex interaction between their heredity and environment.’

In the 1900s, Binet (the inventor of the IQ test) picked on judgement, common sense, initiative and adaptability as ‘essential ingredients of intelligence’. Recently, intelligence has even been linked with spiritual awareness or emotions. Clearly, intelligence in humans is important but it is not the only example of intelligence and we must not let it override all else. If we are comparing intellectual ability between humans, then standard tests of one type or another are useful. However, we need here to consider intelligence in a much broader sense, particularly if we are to investigate intelligence in machines.

ANIMAL INTELLIGENCE

It is well worth considering intelligence in creatures other than humans in order to open our minds to different possibilities. Here, we will look at a few to consider aspects of intelligence such as communication, planning and some of the terms just defined, such as initiative, reasoning and quickness of intellect.

Bees exhibit individual behavioural characteristics within a tightly knit society. They appear to communicate with each other by means of a complex dance routine. When one bee returns from a pollen collection expedition, it performs a dance at the hive entrance, wiggling its bottom and moving forward in a straight line. The distance moved is proportional to the distance of the pollen source and the angle moved indicates the angle between the source and the sun. In this way, other bees can learn which is a good direction to fly.

There are over 30,000 different species of spider, each with its own speciality. Water spiders, for example, live in ponds and build an air-filled diving bell out of silk. They then wait underwater for passing prey such as shrimps. At the right moment the spider pounces to deliver a fatal bite, pulling the prey back into its lair before devouring it.

Many creatures have been witnessed exhibiting learning abilities. A good example of this is the octopus. By training one octopus to choose between objects of different colour, experiments have shown how a second octopus who has watched the first through a glass partition can then carry out the exact same decision-making process.

Many creatures use tools. An unusual example of this is the green heron. Herons have been seen dropping morsels of food into water where fish are expected to be. When the fish swims to take the bait, the heron catches it.

Because of their genetic links to humans, chimpanzees are the most widely studied non-human animal. They can: communicate (even with humans); plan hunting trips; use a variety of tools in sequenced order for food collection or climbing; play; put the blame on others; and even use devious ploys to gain sexual favours – this on top of exhibiting basic learning skills. But perhaps it is easier to measure such abilities when they are close to those of humans. The capabilities of creatures such as spiders, whales or slugs can be extremely difficult to give value to if they are meaningless to humans.

BRAIN SIZE AND PERFORMANCE

It could be argued that one way in which direct comparisons can be made is in terms of brain size, relative numbers of brain cells (neurons) and complexity. Comparing a human brain of approximately 100 billion neurons with a sea slug consisting of 8–9 neurons appears to make a good start. However, between species brain size, neuron size and connectivity all vary tremendously. Even between humans there can be large variations. In the past this was used to ‘prove’ all sorts of results.

In Germany in 1911 the minimum requirement for a professor was a head circumference of 52 centimetres. This was used to discriminate against women; Bayerthal, a leading medical physicist of the time stated: ‘We do not have to ask for the head circumference of women of genius – they do not exist.’ At the same time, Gustave Le Bon, a French scientist of note pointed out that, on average, women had brains which were closer in size to gorillas than they were to those of men!

These serve as good examples of trying to use some sort of measure to come to the conclusion that was wanted (in this case by some men) in the first place. This is something that must be avoided at all costs in studying intelligence, yet it is one that has appeared time and again in studies. That said, it is also inappropriate to overlook observable differences simply because they are deemed to be not politically correct.

One issue with brain size and a count of neurons is the definition of what exactly constitutes a brain. For an individual creature this might be answered quite simply in terms of the main group of central neural-type cells (in a creature's head). In humans, approximately 99% of neurons are in the skull, with the other 1% in the nervous system. In many insects the divide is more like 50–50 due to their dependence on rapid processing of sensory input. In machines, however, the brain is often networked – leading to a conclusion that the effective brain size is the total number of neural-type cells in the network, rather than merely those in one central repository.

A pure count of brain cells is extremely problematic, even in humans. As an example, consider a person who has had a stroke such that their neuron count is significantly reduced due to neural death over a section of the brain. Yet they may still be able to perform in many ways much better than many ‘normal’ individuals.

Perhaps energy usage would be a better start point. Brains are highly expensive in this regard. Human brain metabolism accounts for as much as 22% of total body requirements. In a chimpanzee this figure drops to 9%, and in insects is lower still. In machines that do not move, apart from cooling fans and indicating lights, not far short of 100% of its energy requirements are used for information processing.

SENSING AND MOVEMENT

Intelligence is an important part of an individual's make-up. However, this depends not on their brain alone, but also on how it senses and activates things in the world around it. How the world is perceived by that individual depends on the functioning of their brain, their senses and their actuators (e.g. muscles).

Humans have five senses: vision, hearing, taste, touch and smell. This gives us a limited range of inputs. We cannot sense many signal frequencies; for example, we do not have ultraviolet, ultrasonic or X-ray sensory input. Our perception of the world is therefore quite limited – there is a lot going on around us that we have no idea about because we cannot sense it.

At the same time, another creature or a machine with different senses could be witnessing a major event which a human would know nothing about. A being's senses need to be taken into account when considering intelligence. Just because a being is not the same as a human – for example, if it senses the world in a different way – this does not necessarily make it better or worse, merely different.

The success of a being depends on it performing well, or at least adequately, in its own environment. Intelligence plays a critical part in this success. Different creatures and machines succeed in their own way. We should not consider that humans are the only intelligent beings on Earth; rather, we need to have an open concept of intelligence to include a breadth of human and non-human possibilities.

The story is much the same in terms of movement. Humans are able to manipulate the world in various ways and to move around within it. Each being has different abilities in this respect, depending on what their life role is. It is not appropriate to say something is not (or less) intelligent because it cannot do some specific task. For example, it would be wrong to say that a creature or machine is stupid because it cannot make a cup of tea – this is a very human task. Only in comparing humans should such a task even be considered as some form of measure.

Based on this broadening discussion, a more general definition of intelligence might be: ‘The variety of information-processing processes that collectively enable a being to autonomously pursue its survival.’

With this as a basis, not only can intelligence in animals and machines be respected and studied for what it is, but also intelligence in humans can be put into perspective in terms of merely serving as one example. Clearly, this definition is open to much criticism, but it is felt to be a substantial improvement on those given at the start of the chapter, which have far too strong a human bias to them. It could be argued that the earlier definitions are not explaining intelligence in general, but only human intelligence.

ALIEN VIEW

An interesting way to consider the problem of intelligence is to think of yourself as an alien from another planet, inspecting Earth from afar. What would you consider the intelligent life forms on Earth to be? Could they be vehicles, networks, water, clouds, animals, bacteria, televisions? Presumably you would apply some tests based on your own concepts of life form and intelligence. So, if you are living on a planet for which the main sensory input is a type of infrared signal, then your view of Earth may well not include humans as an intelligent life form.

Even considering what we as humans define as being the basics of life could lead to apparently strange conclusions. From basic biology we could consider the following as indications: nutrition, excretion, movement, growth, irritability, respiration, production (production rather than reproduction as humans produce, they do not ‘reproduce’ other than through cloning, which is ethically questionable).

From an alien standpoint, even a telephone exchange or communications network satisfies these qualities of life – perhaps much more obviously than humans do – merely in terms of electrical pulses rather than chemical. From an alien viewpoint it could be concluded (even now) that a complex global networked intelligence on Earth was being served by small drone-like simpler beings – humans.

SUBJECTIVE INTELLIGENCE

Intelligence is an extremely complex, multi-faceted entity. In each being it consists of many different aspects. Intelligence is also subjective in terms of the group by which it is being viewed and the group being viewed. For any particular group that is considering intelligence, what are and what are not regarded as intelligent acts are dependent on the views of that group and are steeped in the social and cultural trappings of its members.

When a puppy walks by the side of a person, this could be considered to be an intelligent thing to do or simply as the puppy satisfying a trivial programmed goal. When a human is able to rapidly calculate answers to mathematical questions or accurately remember a series of facts on a particular topic, these could be regarded as intelligent acts – indeed the person could be called a ‘mastermind’ – or they could be regarded as a mere entertainment exercise.

With differences between species the problem is exacerbated due to their different mental and physical capabilities and requirements. For humans studying different species (I include machines here) it is therefore important to try to recognise aspects of intelligence for what they are worth within that species rather than merely in terms of how they compare to aspects of human intelligence.

Between humans we need to try and retain a scientific basis for our analysis of intelligence rather than to pamper to social stereotypes. For example, why is it that knowledge about politics, classical music or fine art is seen by some to be more indicative of intelligence than knowledge about football, pop music or pornography? Why is it that playing music by Mozart to a baby while still in the womb is considered, by some, to make the baby more intelligent, whereas playing music by the Rolling Stones is considered to be dangerous? Is there any scientific basis at all for such conclusions? I think not. Where are the conclusive scientific studies that have shown these things to be so? There are none.

Unfortunately, we can quickly run into the problem previously mentioned, in that we already have a conclusion and we try to fit certain observations to match that conclusion and ignore others that do not match. If you wish to succeed at school or university, it is better (I take these merely as examples) to learn about fine art or classical music rather than football or pop music as these latter subjects can be seen as disruptive or a complete waste of time. From those who succeed in these areas of education will come the teachers and professors of the future who, in turn, because of the subjective nature of intelligence, will value those who toe the line and follow the lead of learning about fine art or classical music – those who perform well in the areas considered to be proper by the teachers themselves, who define the subject areas. And so it goes on.

A strong social bias runs through such human educational systems and this can result in completely different values associated with subject areas. An individual can be regarded by others as being stupid simply because they do not know particular facts, cannot carry out specific mathematical calculations or deal with some aspect of everyday life. Clearly, this is merely representative of one aspect of their intelligence – nothing more and nothing less.

Despite this, humans often tend to use the same approach to make comparisons with other creatures or machines. Sometimes we do not give value to non-human abilities, partly because we do not understand them. Conversely, we give value to animals copying some aspect of human abilities – for example, some consider dolphins to be intelligent simply because they do some tricks and are friendly to humans, whereas sharks are sometimes regarded as mindless killing machines because humans do not necessarily have the same mind set and values as a shark.

Each individual has their own concept of intelligence with which they can measure others, both human and non-human, in order to make comparisons – often to come to the conclusion that one individual is more or less intelligent than another. A group's view of intelligence arises from a consensus between individuals who hold similar social and cultural beliefs and share common assumptions. Everyone's concept also partly reflects their own personal qualities.

When assessing the intelligence of a non-human, possibly a machine, if we wish to put it down and claim in some way that it is not as good as a human, then we can certainly make comparisons of the non-human's abilities in a field in which humans perform well. We can, of course, compare human abilities with a non-human in a field in which the non-human performs well – however, the result would not be so good for humans, so we don't tend to do such a thing.

In assessing the intelligence of an individual we really need to get to grips with the physical make-up of that individual, their mental make-up, their social requirements (if any) and the environment in which they live and perform.

IQ TESTS

It is a basic feature of human nature to compare and compete. Indeed, our...