Humans Machines and Armageddon

I presented a shorter version of this essay to the Atheist Society in Melbourne, Australia, on Tuesday 14th of March 2017.

There is lot of speculation about the future of humanity. One expectation is that “transhumans” will live longer and be richer, much smarter and happier.

Several famous people, including Stephen Hawking, fear the worst. Their greatest fears are that information technology will destroy or subjugate us in some way or other. It is this pessimistic future that I will now discuss.

But how far into the future can we expect to see. There are pitfalls in forecasting even for futures as near as a decade or two. With new technologies, the speed of invention and innovation often depends on unforseen obstacles and breakthroughs. For a few years in the 1970s I was a member of a think tank that was looking into the future of telecommunications in Australia. One forecast relating to optic fibres was that by the year 2000 communication services would be provided by “one fibre per subscriber”. The National Broadband Network is unlikely to achieve a nationwide shared fibre by 2020. But on the other hand, no one then had dreamt that mobile telephones would ever do all the things that they now do. And nuclear fusion as a source of cheap clean electricity generation has been 30 years in the future for about 50 years. So futures are always uncertain.

Keeping this in mind, here are three possible adverse outcomes that people have talked about.

The first possibility is that algorithms will replace humans in many kinds of jobs.

An algorithm is like a recipe. It sets out a series of processes that will perform some specific task. Computer programs are algorithms. They perform all sorts of tasks, from the simple to the complex, such as calculating, collecting information, analysing information, making decisions, controlling production processes and controlling administrative processes. The apps on your smartphone are algorithms.

It is feared that widespread use of algorithms will replace almost every job that involves any kind of intellectual activity. The remaining jobs would mostly be in lower paying occupations.

The second possibility is that machines will take over.

Computers will become much more intelligent than humans in every way, will become self-sufficient, will supersede humans as the dominant intelligence on Earth, and, maybe, eliminate us

The third possibility is that humans will unintentionally or deliberately cause their powerful intelligent machines to severely damage or destroy the natural environment, the infrastructure and/or humanity.

Something like this has happened many times on a small scale, when the down side of some kind of technology began to take effect. The extent and the seriousness of this possibility have increased in step with the increasing power of the new technologies and with the increases in population. The situation is now looking dangerous and irreversible.

These three possibilities could all be happening at the same time.

Before continuing, I will repeat that I am discussing only the consequences of using our own technologies, not what the microorganisms, or the normal processes of the planet, or what the rest of the universe might do to us.

Significant Issues

The most significant issues in helping us to guess what might happen, at least in the short term, are probably:

the likely objectives and intentions of humans and of machines;
and the present and likely future abilities and limitations of humans and machines;

Objectives and Intentions

A human objective is a specific condition that a person wants to bring about. Fulfilment of a particular objective would be attempted through the formation of human intentions derived from plans and/or opportunities that may arise.

A human intention is a state of mind that the person should perform one or more specific acts, or not perform specific acts that otherwise would be performed by the person.

If all states of mind are the outcomes of the conditions and operations of the brain, then human intentions are physically determined.

Machine objectives and intentions are determined by algorithms that produce specific actions under specific circumstances.

Human objectives and intentions are personal, commercial, national or religious. They can be benevolent, cooperative, competitive, antagonistic, malicious and aggressive. Each new generation has its own slant on what is appropriate behaviour, and it produces increasingly powerful machines to help achieve its particular range of intentions.

Humans use machines to adapt the planet to serve our various purposes, mostly ignorant or careless of the impact this will have on other species or our own future wellbeing.

Humans use machines to provide goods and services, to provide entertainment and recreation, and to destroy people, animals, property and the biological environment.

Some machines already have simple kinds of intentions, related to the purposes for which they were designed. Some future machines might be provided with the equivalent of the full range of human intentions or the ability to decide their own intentions.

The obsolete F-111strike aircraft could fly at supersonic speeds while keeping to a fixed low-level height above the ground, irrespective of the terrain. This ability was provided by the computerised coordination of its radar, which surveyed the ground in front of and below the aircraft, with a mechanism that directed the aircraft to fly up or down. Its successors have automatic navigation and manoeuvring capabilities that are significantly more sophisticated.

Scientists in Queensland have developed a robot to reduce the population of the crown-of-thorns starfish, which have become a serious factor in the destruction of the Great Barrier Reef. The robot swims over the coral and is able to recognise the starfish and distinguish them from other organisms and objects. It injects a lethal dose of toxin into each starfish it comes across. (The toxin is the bile of cattle, previously a waste product of slaughterhouses. It has no adverse effect on mammals but is deadly when injected into sea stars and related organisms.)

The robot is not conscious but it doesn’t need to be. It is “single minded” and has only one function, and under some conditions it needs to be remote controlled. Enhancements to these robots’ intelligence could create robots that capable of more complex assignments requiring a large range of judgments. Such robots could be used for worthy or benign purposes, but they could also become armies of killer robots that could recognise and attack any kind of specified target, on land, at sea or in the air.

And driverless cars will have short-term objectives of getting safely to their assigned destinations, and long-term objectives of obeying all the rules of the road and performing all the courtesies of the ideal driver during their assignments.

However, none of these present machines have autonomous intentions. They have to be started up by some action that has been initiated by human action and they are able to do only those functions that they have been programmed to do.

Future machines will be programmed to have more intentions of our choosing. They might even be provided with the equivalent of the full range of human intentions. They might be given the ability to develop their own intentions Some machines already have simple kinds of intentions, related to the purposes for which they were designed. Some future machines might be provided with the equivalent of the full range of human intentions, or the ability to decide their own intentions without human control. So we will have to be careful what we do..

Present and Future Capabilities

The relevant basic differences between humans and machines are:

intelligence;
physical strength;
robustness, repairs and maintenance
production and reproduction;
and how many there are.

Intelligence

Present machines surpass human brains in many ways. Human brains surpass present machines in very many ways.

The intelligence of humans differs from that of machines in at least four ways:

speed of processing;
the amount of data that can be remembered and processed ;
kinds of information processed;
and physical structure that provides the intelligence.

The acquisition of information by brains, and its transfer between brains, is comparatively slow and incremental. Transfer and processing of information in computers is very fast.

Computers can contain and process very large quantities of information – sorting and analysing data, and performing huge complex mathematical calculations. On request, search engines such as Google very quickly select and display enormous lists of specific detailed information from a store of billions of items. This far surpasses what any individual human can do. These capabilities were, of course, put into the computers by humans and are continually revised by humans.

Computers can learn such things as the rules of games or the required steps of a process by analysing large numbers of details of actual performances. They also learn by trial and error. This, plus their much greater processing speed, enables them to defeat champion players at complex board games such as chess. But in addition to making decisions on the basis of recorded events, humans also understand particular situations as entities. Simply winning, which is what such computers are designed to do, is not necessarily the desired outcome.

Computers can learn to recognise individual people irrespective of the angle they are viewed from, and can distinguish between live people and pictures or models. Some can discern subtle changes of moods and other conditions.

Computers have composed music and painted pictures that some humans have admired. As far as I know, none have been ranked alongside the notable human composers or painters. In both music and painting, computers are given some kind of theme, some rules for processing it and some system or “embellishing” it, which is not the same as human inspiration.

Some people aspire to give machines all the intelligence and versatility of human brains. In theory, such computers could perform any intellectual or manipulative task and do it better than humans. In practice, this would be much more complex and difficult than most people imagine. Human brains are said to be “the most complex systems in the universe”. Scientists who are attempting to build a digital replica of the brain are finding that this may be no great exaggeration. And present computing technologies are nearing their physical limits of capability.

New technologies that use light instead of electrons may make computers faster, require less space and emit much less heat. Quantum computing, which is a fundamentally different technology, is expected to greatly increase computing capacity and intelligence. At present it must operate at very low temperatures, requiring a lot of additional energy. Large quantum computers would not be very mobile unless mounted on ships, trucks or aircraft. Some developers expect them to be in use in a couple of years. Others are much less optimistic.

Science fiction has featured robots that are indistinguishable from humans and are independently able to act like humans. They are controlled by computers inside their bodies and have their own power supply. They are depicted as friends, or servants or enemies of humans. One scenario is that they are the machines that are going to replace us. Science fiction has often been prophetic, but a lot of major breakthroughs would be needed before anything like this kind of robot could be developed.

The computers we are most familiar with, and which people use many times a day, are small – smartphones, tablets, laptops and desktop computers. Almost all of them are connected to big powerful remote machines via the Internet, and have access to enormous amounts of the information that the big machines contain, that is, to nearly everything that humanity thinks it knows. Future autonomous computers or drones might use this same body of information for their own purposes.

Most domestic computers contain a lot of their owners’ personal details, which some big machines already access, sometimes for unauthorised, subversive or aggressive purposes. Hackers sometimes access them.

It is hard to know precisely what the bigger and more complex computers are doing at any particular moment. These systems control very large intricate processes, such as trading on the stock exchange, selecting the answers in search engines, monitoring inputs to social media or making decisions about individual people. Sometimes such systems go inexplicably wrong.

Humans have very flexible and adaptive intellects. They can negotiate a variety of environments, both physical and social. They have ways of satisfying their likes and avoiding their dislikes. They have emotional relationships with people, animals, objects and places. They surpass present-day computers in all of these abilities. And they have curiosity, a sense of self and engage in self-analysis. . For most everyday purposes, the comparative slowness of human brains is not a significant disadvantage.

Humans have whims, which often make their thinking, their wishes and their actions insightful and unpredictable. Emotions can give humans an advantage over machines because emotions produce determination, resourcefulness and unpredictability.

Human brains continually assess their environment using inputs from sensory organs, predict possible outcomes, and send into consciousness whatever seems significant.

Emotions are the greatest influences on human thinking. Humans also have whims, which often make their thinking and their wishes and their insights and their actions unpredictable. There is no obvious advantage in making machines emotional or subject to whims, and it could interfere with their reliability. In some contexts, emotions could give humans an advantage over machines because emotions produce determination, resourcefulness and unpredictability.

Emotions can also lead humans to do things that reason tells them are unwise, and to neglect doing what reason tells should be done.

Humans are already being enhanced by the use of pacemakers and other devices that are attached to or implanted in their bodies. Defective organs can be replaced by organs artificially grown from cells from the body. These provide additional and enhanced physical and mental abilities, and such technologies are continually expanding.

Some scientists expect new developments will greatly increase human brainpower, physical power and health. But it would be mainly the richer minorities who had access to these enhancements.

These enhancements might not alleviate psychological issues, relating to differences of belief, socio-economic status, and preferences of lifestyles, that cause antagonism and disruption within and between human societies.

To some people all this might, on balance, look bad for us. But all of the kinds of organisms on Earth have managed to survive and reproduce without needing the huge abilities that computers have. By necessity, all organisms have an intrinsic urge to find nourishment, to secure their safety and to reproduce. All organisms are the products of a process that has adapted them to be able to live in a particular kind of environment.

And all computers and other machines are the products of humans, who use them for human purposes, good or bad.

Physical Strength and Dexterity

Some machines have enormously greater physical strength and capability than any living organism on earth. This relates to abilities to lift, push, pull, hit and move quickly, all of which humans can do, and to fly, which humans cannot do. Some machines are a lot harder to immobilise or destroy than humans. Some machines can perform much more precise tasks than the human hand.

But no single device can yet equal the ability of the human hand in performing the complex and diverse range of manipulative tasks that most of us do every day, or to devise and perform the appropriate manipulation for an unexpected task. Perhaps someday some computer might.

Robustness, Repairs and Maintenance

Humans and machines have different susceptibilities to damage and to healing.

Human have many self-healing capabilities, and access to many kinds of health professionals. They readily help each other, and can recognise when help is needed. And their injuries do not always disable them. The need to sleep could make them vulnerable, and they also need protection from environmental hazards.

For machines, regular maintenance, and detecting and repairing many kinds of damage, wear, and malfunction are performed by human specialists, who use specialised testing and repair machines and spare parts. It would be very unlikely that there would be autonomous machines that could undertake and coordinate these tasks without human control.

Humans can tolerate humidity and getting wet, but only a small temperature range. The computerised parts of machines are generally susceptible to humidity and water and to mechanical damage and to high operating temperatures – and their operation produces a lot of heat. They could be built to tolerate these things, but heat could be a problem.

Reproduction

The very complex process of reproduction of all organisms has been continuously achieved over the past two or three billion years.

Machines cannot reproduce themselves.

While algorithms can supervise complex management routines, it would not be feasible, even for extremely clever and capable machines, to be able to mine and refine the necessary minerals, process all the materials into forms necessary for assembly, design the assembly of every kind of component, and then assemble and test a new machine. There would need to be very many different kinds of computers that were already specifically programmed for each process, and they would need to have been tried out to ensure that the programs had no defects. Perhaps some thoughtless genius might try to make such machines, but the range and types of processes would be so enormous that it would be quite impracticable, particularly when the inevitable unexpected events happened.

Perhaps machines might enslave humans to do some of the work for them. But there would be comparatively very few super-smart machines, and there are so many humans, so sabotage should be able to stop hostile computers from manufacturing copies of themselves.

Energy requirements and supplies

Machines, and particularly computers, need a specific steady reliable power supply.

Humans need frequent nourishment, sleep, social contact and grooming. But they have an enormous choice of food from a very diverse range of sources of supply. But they do not have stringent requirements.

Large computers use very much more energy than humans for comparable processing of information. The energy used in getting rid of the heat makes computers even less efficient. Even desktop computers use more energy and space than human brains.

Energy and space requirements need not be an issue for machines that don’t move. Autonomous machines in vehicles would use the vehicles’ power sources, but might need large vehicles to hold them.

I think it will be a long time before the energy and space problems are solved for humanoid robots.

What Does All This Say about the Three Fears Concerning Machines and Humanity?

1 Algorithms

Algorithms have already been taking over jobs from humans. It is fairly certain that they will increasingly be used by individual humans and by organisations to take over many more intellectual functions such as:

routine management of production, and of most kinds of services, and transport and research;
complex decision-making, such as legal judgments, political decisions, and diagnostic analysis;
and surveillance of the actions and processes of humans and machines..

But these processes will be commissioned, designed and supervised by humans.

Using them will be cheaper, quicker and more reliable than using humans. Overall this would make the society economically richer, but it would cause massive losses of jobs in many fields of activity.

Algorithms will be used directly or indirectly to serve humans, who will pay for the services. But there will need to be enough people still in jobs to be able to pay.

Almost all new technology creates new jobs as well as job losses. Some of the new jobs will be in creating new algorithms, servicing the new algorithms, and correcting glitches in existing algorithms, some of which will go destructively wrong.

Also, human emotions, curiosity, impatience, aspirations and whims will make us want new and better things. There have always been new things for humans to devise and deliver to accommodate such demands. So there will be some jobs arising from this. There will also be jobs in inventing unexpected new kinds of things that will create new demands.

Every society, irrespective of the technology it uses, has to develop some kind of balance between the cost of providing goods and services and the ability of the users to pay for them. A crucial factor is the proportion of the economy that is displaced by the new technology, in comparison to the proportion created by it. The nature of the society and the new technology will determine whether some sections of the society are able to exploit any changes, at the expense of other citizens. In the case of new intelligent machines, their owners are likely to have increased power.

If incomes are distributed reasonably fairly, most people should be able to live comfortably and have a lot of leisure time, because of the total increased value of the new algorithms. Addressing this is a social issue that all societies have had to face whenever there have been significant changes in technology. Some societies will handle it better than others.

The use of algorithms for surveillance of the actions and processes of people and machines could enable cities to operate more efficiently and effectively and to become safer from accidents and crime. But it could also make it easier for governments to have greater control over their citizens, and for corrupt officials and hackers to access personal details for their own purposes. This is another issue that each society will have to deal with.

2. Machines Will Supplant Humans

Computers already have programs that provide something like intentions, and instructions for fulfilling those intentions. For computers to be able to generate their own intentions they would need to be able to identify new needs or threats or ambitions, and address them by generating plans and having a compulsion to act. We might wonder who would give machines intentions of controlling or displacing us. But don’t discount malevolent geniuses.

However, there would be comparatively very few super-smart machines, which could not reproduce themselves, and only a few of them would be able to take hostile action against humans. There might be heavily armed very intelligent autonomous self-driving cars or aircraft that could be fearsome opponents, but they would be on both sides of the conflict.

For highly intelligent machines to work together they would need to communicate with each other. This could make them susceptible to hacking and viruses that could make them innocuous or unworkable.

One suggested type of hostile machine is the intelligent self-replicating nanobot, or, as the scenario goes, multitudes of nanobots. (A nanobot is a very tiny robot, even smaller than microscopic.) These tiny but very intelligent and capable machines would dismantle any piece of material that they might encounter and use the bits to make replicas of themselves. This would progressively convert all material, including humans, into nanobots.

There are massive problems with this scenario, starting with putting the required complex hardware into something as small as the nanobots. But whatever their size, these robots must be able to move, and to identify the particular atoms they need, and to gather the required number and types of atoms, and to stop these atoms from reacting with each other or with the environment in the “wrong” way during the process, and to be able to stop collecting when they have the necessary numbers of each, and to assemble the atoms into components, and to assemble the components into a workable replicas of themselves, and have the right kind and quantity of energy source and tools to perform all these processes. There would be very few environments within reach of a nanobot that would have all of the chemical elements necessary to create another nanobot. And these vicious little machines must also be able to avoid being destroyed by whatever is in their environment, including other nanobots.

So that is one more suggested kind of machine that we don’t have to worry about

3 Actions of people

Optimistic futurists describe all of the wonders that new technologies will bring. Their futures will transform the world we live in and will create great enhancements to our bodies and minds, and also to our wealth.

Historical statistics show that humans have continually made advances in their wellbeing over at least the past few millenniums. These advances relate to caring for each other, to abilities in healing, to knowledge about the workings of nature and societies, to providing facilities that make life safer, more comfortable and more affluent, and to the depth and universality of education. Futurists also point to evidence that the proportionate amount of warfare and conflict throughout the world has been continually decreasing, despite the current presence of the international tension and conflict in many countries. They expect that warfare will continue to decrease.

Humanity, optimist say, will become more peaceful and more responsible about caring for its members and the environment. We might not think that is true, but we know how uncaring humanity has been in the past. Pessimists might not accept that the improved behaviour will necessarily be passed on to future generations.

The pessimistic futurists point to human irrationality, human greed, human jealously and passionate beliefs about religion and morality. There is no shortage of examples, past and present, of all of these human characteristics and their consequences.

Currently, humans are causing:

global warming;
proliferation and extinction of other organisms;
pollution;
reduction of the productivity of the land and the seas;
and deliberate destruction of the environment, infrastructure and people.

The pessimists worry that the increased powers of future machines, and the ignorant and uncaring habits of humans who use them, will greatly exacerbate the damage to the sustainability of life.

With good will, and with greater understanding of the issues, and careful planning, humans could start to improve the situation. But despite the continuing public declarations, there is as yet no sign of anything like adequate will, or understanding or planning. The main reason for this is that the way we live our lives is dependent on the sophisticated technologies we have built, which are also the causes of the degradation of the environment that we depend on for our existence. It will be extremely difficult to turn around our lifestyles sufficiently and in time to avoid some kind of catastrophe, and at the same time maintain a workable social and economic structure. Any efforts to succeed in this will be disrupted by increasing global warming, and pollution etc., to say nothing of vigorous political opposition.

There is also the likelihood that when more people become more knowledgeable, and machines become more powerful and intelligent, then hostile individuals and groups will have greater opportunities to cause severe damage to communities, irrespective of where the perpetrators and victims are located. Our increasingly interconnected society is vulnerable, with the internet providing the means for espionage, the recruitment and coordination of dissidents, and the disruption of processes. In a world already stressed by a damaged environment, this could be the final straw.

Human actions that degrade the conditions that sustain human life will probably have less impact on machines, which will be designed to cope with their environments.

What is the likelihood of the various expectations?

History supports the optimists. There is a saying that each generation tries to make life impossible for the following generation, but it never succeeds. Perhaps no generation has deliberately tried to do this, but humanity has continually introduced new technologies that at first seem successful but later show their downsides.

But we have continued to proliferate, grow richer and live longer. Will human ingenuity save us again, or has the luck have run out?

Humanity will decide how it handles the issue of algorithms that displace humans from their jobs. Algorithms make societies more productive, so we should expect richer lives. At the worst, algorithms will cause the same kind of displacement that previous innovations have caused, which would be disruptive but survivable.

Some machines will become more intelligent than humans in many respects. But the limitations of such machines would prevent them from replacing or destroying humanity. Sheer intelligence is not enough; it is not the same as wisdom. The world’s greatest geniuses are not also the leaders of the world. And if the machines were so smart, they should value what they get from human help.

There are so many more humans, many of whom would be vigilantly watching all computer developments.

So I don’t share the fears of Hawking and the others. I think we would be quite capable of stopping hostile machines from wiping us out.

If machines were to bring about a total calamity, I think the most likely cause would be the direct or indirect use by humans of their technologies, both deliberately and thoughtlessly. The collapse of societies has already happened this way many times at a local scale, often in conjunction with local environmental changes and/or warfare.

Weapons continually become increasingly powerful, with greater range and intelligence. Warfare and colonisation cause famine by disrupting the environment. And it is easy to forget how completely we depend on the environment. The environment is the ultimate source of all of our food, no matter how indirect the process might be between the original production and the eating.

Currently there is increasing inequality, contested global power, climate change and other global degradation, with elites intent on increasing their power for their own purposes.

The situation would be made worse by the unrest caused by jobs being taken over by algorithms, and by hostile super-intelligent machines, and by other things humanity is doing.

Despite all this, I don’t think it is inevitable that humanity will be entirely destroyed by its machines, particularly now that there are so many of us, and we are very diverse and adaptable and have so many smart computers to help us.

But it is impossible to know what humans are going to do next.

bring about a total calamity, I think the most likely cause would be the direct or indirect use by humans of their technologies, both deliberately and thoughtlessly. The collapse of societies has already happened this way many times at a local scale, often in conjunction with local environmental changes and/or warfare.