Essay

Von Josef G. Boeck for The Human Side of Business GmbH

Many developers of digital intelligence base their software architectures on the capabilities of humans. This is why it is so easy for many to equate intelligence in computers and humans. Gradually, this idea is also emerging in the phenomenon of consciousness. The more capable the systems become and the closer they come to or even surpass the performance of humans, the more the question arises as to whether machine intelligence also gives rise to machine consciousness.

In this essay, I ponder over the question of how we can approach equipping computers with consciousness and what follows when experts develop consciousness for the digital tools we use. My specific question is whether and – if so – what knowledge from the natural sciences should be used in the design of intelligent and conscious computers

I choose the form of the essay. This genre allows being subjective and scientifically sound at the same time. This suits my topic because, even after many decades of research, we have gathered a lot of information about human brains, intelligence and consciousness, but we still don’t know how consciousness really comes about. We are all working on this topic with hypotheses, claims and opinions, which some are selling like proven knowledge.

Before writing this essay, I didn’t realise how I would answer the question of whether we should pursue machine consciousness as a development goal at all. However, due to my professional and personal interest in digitalisation and digital intelligence, I need an opinion on this question. In this text, I summarise everything that has helped me to formulate an attitude.

What is consciousness?

Human intelligence and human consciousness have always been the role models for computer developers, especially when working on digital intelligence (my term for artificial intelligence). I am not looking for a universally applicable definition of intelligence and consciousness here, but am focussing on aspects that have practical relevance to my questions.

Consciousness is an ability of brains that has developed through evolution. In conscious living beings, basic biological building blocks of the brain interconnect in a way that creates new properties and abilities that are not possible for the individual basic building blocks. Through consciousness, living beings can learn in different ways and develop a subjective idea of themselves and their environment. They can perceive their inner states as well as stimuli from the environment, evaluate them and convert them into intentional actions.

When thinking about and researching consciousness, humans have long been at the centre of interest and are still the most important reference today. Many experts now agree that consciousness should also be attributed to other living beings in varying degrees. In principle, this opens the door to recognising consciousness in other complex systems and raises the question of what is constitutive of consciousness and whether inanimate matter – which currently includes all computers – can also develop consciousness. Experts have very different answers to this question: Those who believe that consciousness can only occur in living beings give good reasons for this. Those who believe that all systems that fulfil the above criteria should be granted consciousness also argue convincingly for this.

We are still lacking a lot of information for the reproduction and diagnosis of consciousness in non-living systems. It has not yet been possible to measure consciousness. We do not know where and how the information that consciousness processes and the phenomena that consciousness produces are stored. Because this is the case, the pure “hardware” of the brain is not sufficient as an explanation for some thinkers. They introduce immaterial and metaphysical structures that make consciousness possible beyond the chemical, physical and biological components of living beings.

Until proven otherwise, I am convinced by experts who regard consciousness as a purely material phenomenon and who do not need any additional metaphysical systems simply because no one yet understands the capabilities of the “hardware”. I believe that one day we will be able to measure and understand consciousness.

In principle, as I understand it, consciousness is conceivable in non-trivial machines such as computers. If we want to teach machines consciousness via algorithms, we have to develop them. We are not helped by natural selection or the properties of basic biological building blocks. As we do not understand nature well enough at this point, we cannot claim that we are using it as a model. Of course, this may change. We don’t yet know when or how. As long as this is the case, we should adapt our definition of consciousness so that machine consciousness has a significance we can work with. Or, because of the difficulties involved, we should refrain from using the term for computers for the time being. In any case, experts will continue to work on simulating the observable phenomena of consciousness in living beings with computers. Perhaps something useful will emerge. In the past, we have learnt a lot about the objects we have tried to simulate, even if we have ultimately failed.

We should distinguish consciousness from the unconscious and subconscious in humans. I equate the two because the difference is of secondary importance for our context here. The subconscious is the world of metabolism, homeostasis and needs. The subconscious mind works partly in parallel and in addition to the conscious mind, but also gets in its way time and again. Like the conscious mind, the subconscious mind has its own agenda in every individual. Many of us use a lot of energy to bring the subconscious into the realm of the conscious in ourselves and others so that we are not too determined by the subconscious.

However, these statements do not yet sufficiently define the consciousness of living beings.

I start by stating that the brain and consciousness are not the result of calculation processes and do not calculate when making decisions. We do not know exactly how they arrive at their results. With computers, on the other hand, we are dealing exclusively with computational processes. If computer technology does not change here, there will always be an important difference between the two types of consciousness.

Then there is the finding that with practically identical genes in living organisms of the same species, consciousness is what gives the systems their individuality. In detail, no living being is like any other. This is the basis for the diversity of problem-solving strategies for survival and reproduction that can be observed in nature.

It has been stated repeatedly that the conscious and subconscious minds are bound to the living being in which they originated. They are always active and only stop working with death. The conscious mind may be temporarily shut down in some physical states (coma, unconsciousness, deep sleep), but at least the subconscious mind always seems to be at work in the background. The difference between a conscious and unconscious state can be measured in living beings using brain waves.

And finally, the conscious and subconscious minds are difficult to change from the outside. Both are subsystems that can only bring about change using their own resources. Of course, external impulses are of central importance for this, but consciousness and subconsciousness in living beings are resilient phenomena and contribute to the ability of one individual to evaluate another. This makes consciousness a factor in social interaction.

A necessary distinction is that between consciousness and intelligence. For our context, I follow those who define intelligence as the ability of systems to solve complex and constantly new problems within a certain time. The faster the solution, the higher the intelligence. Although there are correlations between consciousness and intelligence, they seem to arise and exist largely independently of each other. Highly intelligent creatures can have a low level of consciousness. Systems with a high level of consciousness, on the other hand, are often also more intelligent than average because they create a complex world in which they encounter complex problems. As with consciousness, intelligence in living beings is also characterised by the fact that results are not achieved through computational processes.

Consciousness and intelligence are therefore two different abilities of the brain in living systems that work together and interact with each other. Their interaction in turn forms a complex system that works differently in each individual. We do not know exactly how this works. However, we can observe the phenomena in everyday life.

In terms of the definition of intelligence used here, the question arises as to what this means for computers. So far, computers are only intelligent in a very narrow sense. The expertise of computers is currently limited to specific mathematical calculations; they cannot do anything else. As a rule, mathematical procedures cannot be applied to new problems without modification, and a computer can certainly not apply a mathematical procedure by its own choice to a new problem. This is one reason why experts have difficulty with the concept of computer intelligence. They therefore supplement the concept of intelligence with words such as “digital”, “artificial” or “machines”, because intelligence in computers is currently getting caught up in the media.

Experts agree that digital intelligence and machine consciousness are general purpose technologies (GPT). This means that they can be used across many platforms in many digital systems if they fulfil the necessary requirements. Of course, this reminds us of the fact that consciousness and intelligence of living beings are useful for an infinite number of problems.

Consciousness and evolution

The definitions of consciousness and intelligence make it clear how wide the range of interpretations is as long as we have so few hard facts at our disposal. Basically, we all ascribe to both terms what we want to see in them from our subjective perspective. To assess our question of whether we should develop machine consciousness, a clarifying definition is helpful, but not sufficient. We are more likely to reach a judgement if we look at the consequences that consciousness has for systems.

At birth, living beings begin to compete for the resources they need to survive, grow and thrive. The more consciously a living being can organise this competition, the fitter it is for life. Using humans as an example, we can see the unchallenged and dominant position in the world that results when consciousness becomes effective in beings with mobile and complex bodies.

Living beings with consciousness can dose their body’s own resources and activate them when it seems necessary. They can do this because they constantly perceive the outside world through their senses and evaluate its changes for themselves. They learn from their own and other people’s experiences. This learning leads to them learning to correlate the future consequences of decisions and actions. The conscious mind often even constructs cause and effect relationships. The higher the level of consciousness and the more it is supported by intelligence, the more individualised the problem-solving skills of individuals become. In social beings such as humans, subconsciousness and consciousness also seem to lead to reciprocal behaviour, i.e. an understanding that individuals in social groups should do something for others so that others will do something for them. Consciousness is therefore an important success factor in the development and implementation of survival strategies in social organisations.

I suggested that intelligence makes consciousness more successful in the competition for resources and in solving problems. This proves once again that intelligence and consciousness have very different qualities. It seems that the abilities add up, if not multiply.

What of this can be transferred to computers? How can we get closer to an answer to the initial question of whether we should develop consciousness in digital computers?

Living beings have evolved through evolution and the question of their purpose is a subject of research for philosophers. Computers are tools that humans have developed. Their purpose is clear: to help people cope with their tasks and problems. Consciousness and intelligence are resources for living beings, for tools they are properties.

With digital intelligence and especially in the form of robots, computers can become autonomous agents and replace humans in many areas. Computers solve resource problems for humans and other living beings. If consciousness and intelligence help living beings to compete for resources, consciousness should help computers to be more valuable tools for us humans. Intelligence and consciousness in computers have the character of “features” of hardware and software. We develop them into computers and can change them, delete them, transfer them to other systems and thus multiply them. We can even make them available as centralised services for a large number of systems at the same time.

Today, digital intelligence is defined as the ability of machines to learn. This capability is one of the advantages of consciousness in living beings. Computer learning by means of neural networks is modelled on the human brain. However, it currently functions entirely without consciousness. Computers with consciousness could probably learn more effectively and more quickly. They would need much less data than current digitally intelligent systems and could develop many more deductions from the correlations they can derive from statistics.

Computers would also benefit if they were directly connected to the environment via many simultaneously open channels and could process the information in an instance such as the consciousness does in living beings. Today, we have to create digital images of the environment in computers and compare them with the environment using cameras, sensors or microphones, for example. Computers are not yet able to work with the many other senses of living beings. Computer awareness would make this interaction with the environment much easier. This is why this aspect, which is a prerequisite for many other technical advances in computer development, is currently being worked on by a lot of researchers.

What applies to the input also applies to the output. Computer-generated texts or speech and computer-controlled actions following the occurrence of certain conditions are very sophisticated today. What is still missing is that computers are not only able to determine their output on the basis of statistics or trigger their actions according to “if … then”, but that they can also adapt to individual and dynamic contexts. This would increase the quality of outputs many times over and save resources – in terms of the data required and the energy consumed by the computers.

We can view the efforts of experts to realise the potential of digital tools as the evolution of machines. However, we can only speak of an evolution in the sense of nature when technical and biological systems work together. This is also currently being researched.

With computer consciousness, computers would quickly develop in different ways because their different inputs lead to different information and data, which in turn lead to different levels of consciousness. With consciousness, computers would only be exact copies of each other at the moment of delivery, but would then quickly diverge and develop different problem-solving skills. What I said earlier about these capabilities in nature also applies here – the greater the diversity of approaches, the more likely it is that solutions will be found for complex problems. We don’t yet know how this could work. What is clear, however, is that unlike humans, computers can be deprived of any problem-solving capacity by pressing the switch-off button or in the event of a power failure.

Consciousness and subjectivity

The consciousness of living beings does not only have advantages. If we ever manage to replicate it 1:1 technically, our tools will also develop a few disadvantages.

One major disadvantage may be that our consciousness does not provide rational decisions. Depending on our personal background and experience, decisions are made by our subjective consciousness, which is only partially rational. The decision-making process of the brain through consciousness is complex. So if we actually recreate human consciousness in computers, they lose the main strength of mathematically derived decisions – their rationality, which is based on mathematical rules.

This is linked to the lack of transparency of the decision-making process and the criteria used. This points to the current criticism of digital intelligence and the effect of neural networks, which many criticise for often failing to make their decisions traceable. However, trust in digital decisions stands and falls with this. This provides a foretaste of what it would be like if conscious technical systems were consistently modelled on humans.

The inability to find the right measure in the competition for resources is another weakness of human consciousness. We are extremely ill-equipped to find the right measure for our external resource consumption. As a result, we pollute or even destroy our environment more than necessary. Too many people pay too little attention to the impact of our actions on other living beings and our environment. In times when there were fewer people, this was not a problem. But now it is. Unfortunately, our consciousness has not adapted to the new situation.

We also see this excessiveness in the construction of tools such as digital intelligence. The vast amount of data required and the effort we have to make to process it technically and simulate the real world increases the demand for resources to such an extent that there are currently fears of environmental damage from the use of this technology if we do not change this. As our consciousness in the competition for survival does not have an effective switch-off per se, we are constantly in danger of harming ourselves in the long run.

Our minds need to be constantly challenged by the environment in order remain competent and up to date when solving problems. Whether it is the use of navigation systems, the acquisition of factual knowledge or the use of technology in general – digital technology reduces our overall competence in dealing with the environment. Our consciousness plays a trick on us in this respect through its desire to save energy.

So what do we get when we equip computers with a consciousness that is modelled on the human brain?

We are gaining a lot of new players who are replacing people in the processing of tasks for which there are not enough people available or which people do not like to deal with.

Machines with consciousness will lose their power to make purely rational decisions and be a rational tool for us. The consequences of this have certainly not yet been fully thought through.

Our lives are becoming considerably more complex thanks to conscious machines. We all already have many fellow human beings and other living creatures as an environment that is only predictable for us to a limited extent. We would then have unpredictable machines around us. This can frighten even optimists.

The combination of digital intelligence and machine awareness can of course become actively dangerous if developers and users use it to pursue goals that are not in the interest of the majority of people. We can already see this today in the world of cyberattacks on our systems and thus on our everyday lives. If we do not develop safeguards against such developments to the same extent as the possibilities of the technology, we will certainly ask ourselves whether the disadvantages do not outweigh the advantages in the long term.

Formulation of my position

What conclusions and what answer to the question posed at the beginning do these considerations lead me to?

When assessing the benefits we can derive from intelligent or even conscious machines, it doesn’t matter whether consciousness really exists as it does in humans or whether it is a simulation. What is important is whether it is useful for our technical tools and therefore for us.

We should not regard intelligent technical systems with machine consciousness as anything other than tools. They remain technical devices, even if we programme simulations of intelligence and consciousness into them. I therefore consider the question of whether computers should be given rights to be absurd. I see no reason to change this understanding of computers in the foreseeable future. The competition for resources would become many times more demanding for us humans.

With every further step in the development of computers, we need to ask ourselves for whose benefit it will be and in which applications the technology should be used. We need experts who can assess the consequences of technology and see it as a service for us all. This will then allow us to use technology ethically with the potential of digital intelligence and machine consciousness.

As a “general purpose technology”, we should continue to research both digital intelligence and machine consciousness because they can have enormous benefits. It would be ideal if we could switch consciousness on and off in machines. Then we would be able to decide when we need the rationality of computers and when we need a human-like behaviour.

As humans, we will benefit if the majority of our tools are intelligent. This will improve our quality of life in many areas. However, machine consciousness will not bring the added value for the mass of computerised tools that justifies the expense of development and implementation. Above all, however, the infinite mass of existing, non-intelligent tools must be disposed of or converted so that we can fully benefit from digital and networked tools. Even if we have the technology, this will take many decades and will proceed very differently in different regions of the world.

We will learn a lot about ourselves and other intelligent and conscious systems along the way if we continue to view them as tools. Collaboration between the natural sciences and engineering should become much more intensive. Just as many mechanical and civil engineering companies are currently becoming software companies, ambitious engineering organisations will hire natural scientists and humanities scholars.

Pre-thinkers / Sources for this essay

Bostrom, Nick; Chalmers, David; Damasio, Antonio; Darwin, Charles; Dennet, Daniel; Gabriel, Markus; Hinton, Geoffrey; Hoffman, Donald; Langhammer, Falk; Nagel, Thomas; Otte, Ralf; Searle, John; Sheldrake, Rupert; Tegmark, Max

To open and download the PDF of the English version of this essay klick here