Excerpt
Table of Contents
Table of figures
1 Introduction
2 Artificial Intelligence
2.1 What is intelligence?
2.1.1 What is memory?
2.1.2 How do we learn?
2.2 Can machines think? Turing Test
2.3 The Winograd Challenge exuberates the thinking of machines
2.4 Comparison of artificial intelligence and biological intelligence
2.5 Current State of Artificial Intelligence
2.5.1 Artificial General Intelligence
2.5.2 Society’s perception of artificial intelligence
2.6 Ethical problems of artificial intelligence
3 Consciousness
3.1 What is consciousness? An exposition of the philosophic debate through the course of time
3.2 What is consciousness? A scientific approach to the revelation of the mystery
3.3 The importance of the intersection of philosophy and science in today’s time
3.4 Does artificial consciousness exist?
3.5 What importance does artificial consciousness have?
4 Future work
4.1 Definition of the term
4.2 The role of human beings in the prospective work sphere
4.3 Productivity augmentation due to the introduction of machines and computers
4.3.1 Why can we augment productivity?
4.3.2 Company example of how humans and machines work together
4.4 Future work under the consideration of business ethics
5 Conclusion
Appendix
Table of figures
Figure 1: Comparison of a feed-forward and a recurrent network
Figure 2: Neuronal Correlates of Consciousness
Figure 3: Complexity of questions about consciousness
1 Introduction
The purpose of this thesis is to discover in what way human beings will experience a transformation of their own position as creators in the world of tomorrow.
Digital transformation, the era of life 3.0, automation, Internet of Things, Big Data and Artificial Intelligence are major frontier technologies, that today’s business world is confronted with. The idea of obtaining the highest possible amount of efficiency in our society and meanwhile reaching the greatest point of human capacity, through widening our own horizon and creating inorganic mechanisms, that know and continue to learn how to function on a level of an organic life are the most recent and thought provoking aspects of our current time.
This manner of research is significant, because humankind is only at the beginning of incorporating inorganic life into their work sphere, which leads many corporations to creating trend hubs, hiring the best available technology experts to consult the company and prepare it for future incidents. The conversation covers predominantly the technology itself, but not what will happen to the person being replaced, which is a decisive element in people refusing opportunities. They are not aware of what will happen to them and how their role changes
2 Artificial Intelligence
The devotion of Artificial Intelligence (AI) is to endow a machine with intelligence. AI is a sub part of computer science. It includes the study and design of intelligent agents that have the ability to analyze the environments and produce actions which maximize success. AI research uses tools and insights from distinct fields, includingcomputer science,psychology,philosophy, neuroscience, cognitive science, linguistics, operations research, economics, control theory, probability, optimization and logic. AI research also overlaps with tasks such as robotics, control systems, scheduling, data mining, logistics, speech recognition, facial recognition and others.1
2.1 What is intelligence?
Defining intelligence is controversial, because there is no explicit, unequivocal definition of the term. As R.J Sternberg states it: “Viewed narrowly, there seem to be almost as many definitions of intelligence as there were experts asked to define it.”
Therefore, a comprehensive definition will be given, that incorporates the largest possible aspects.
Intelligence is the capacity to achieve complex objectives.2
This definition is comprehensive enough in order to incorporate different concepts, that are understood as aspects of intelligence. For instance, understanding, consciousness, problem solving and learning can all be examples for complex objectives.
Since various forms of objectives exist, we observe a diversity of manners of intelligence. It is difficult to compare intelligence to one another. Taking the example of computer programs. Is a program that only knows how to play Go more intelligent than one that only knows to how to play chess? There is no reasonable answer for that context. We can only say that a third program would be more intelligent, than the mentioned ones, if it performs at least comparably as well as the other two in achieving their objectives. This brings us to the classification of intelligence, since the differentiation between limited and encompassing intelligence exists.3
Human beings are capable to achieve a variety of aspirations, which is why our intelligence is more encompassing than the intelligence of computers. This argument is grounded on the fact that we are able to execute a certain job with successful goal achievement, have a variety of hobbies, we perform well in and speak multiple languages at the same time. Human beings have shown to be multi-skilled.4
Machines on the other hand are designed to achieve only one specific goal successfully at a time. For instance, IBMs chess computer Deep Blue managed to win against the chess world champion Garry Kasparow5, but lost against a five-year-old at Tic-Tac-Toe.6
Thus, these information lead us to the conclusion that humans can have strong skills in a variety of domains, whereas machines can surpass our capabilities in different sectors decisively.7
Another conclusion from this theory is that Artificial General Intelligence (AGI) is aspiratory. The term has predominantly been coined by Ben Goertzel and signifies that machines come to achieve every objective at least as well as humans do.8
In certain areas, machines have shown to already perform tremendously better, than we do. Taking the example of a calculator, especially in theorem proving, we cannot keep pace with computers, owing to the way our neuronal networks function.9
Concluding, intelligence is an information processing mechanism that is outperformed by particles, or to be more precise, neuronal networks, moving around according to the laws of physics. In addition to this, there is no physical evidence to the fact that a machine cannot be better than human beings in all possible perspectives.10 A following conclusion to this is that prospective capabilities will widen the horizon of mechanical possibilities.
2.1.1 What is memory?
The key word neuronal networks brings us to the question on how our memory works and how it has to be differentiated to inorganic mechanisms. Human beings use a whole spectrum of tools in order to save information. This spectrum starts from books over brains to hard disks. It is irrelevant on what matter information is saved, because the information will stay the same.
From this fact, we can derive that information have an independent existence and are substrate independent. This also demonstrates how something intangible like intelligence can be substantive. Since memory is independent of its substrate, engineers have managed to enhance computer’s memory capacity, without changing their software.11
Human brains have a memory of 10GB, which rapidly shows that our biological systems defeat against the ever-evolving capacity to save information on computers.12 Apart from this, costs of memorizing information have drastically decreased over time. Counting from 1955 until today, hard disks have become 100 million times cheaper.13
Not only the amount of storage capacity is what distinguishes organic beings from machines, but also its execution and the procedure of its application are differentiating aspects. The most tremendous difference will find itself in the fact, that to find required information, a computer needs to know where information is hidden and human beings need to know what information needs to be encountered.14
Computers save their memory in a binary number. In other words; they translate information in numbers of 0 and 1. Thus, within searching information, a computer searches information like an address. Biological brains function rather like search engines. A given information is associated with other information. This storage system is called auto-associative system.15 Familiar patterns are indicated and auto-associative memories are fed with the output of each neuron and impose an input.16
An example for this phenomenon is the phrase “to be or not to be”. The associative system will, if the phrase is familiar, instantly continue with “that is the question”. Through given patterns, we remember and associate. This is the main difference between the human and the technical saving mechanism.17
Our memory recalls information by association. The mechanism of how we learn is to be divided into several steps. The first aspect is the filter in our brain, selecting what information are meaningful and need to be processed. Otherwise our brain would be overloaded with information and would not function as it does. The next aspect is the differentiation of short-term and long-term memory, whose storage capacity is determined by the repetitiveness of memory recall.18
In computers, information needs to be saved only one time and it will be invocable under the condition of enough storage space. Concluding, human beings have a more complex and more elaborate process of information assimilation and memorization. Computers are fed with information by human beings or are able to learn, by deriving provided information.19
2.1.2 How do we learn?
Computers are able to learn, because of the existence of neuronal networks. They are a set of neurons, organized in layers. Mathematical operations are the base for neuronal existence. Taking input, multiplying it by its weight and consequently passing the sum to the activation function towards other neurons is the programming method. Neuronal networks are supposed to classify information as a result of previous information intake.
Neuronal networks learn if they are put repeatedly into the same condition. They will learn the state and be able to come back to it from an adjoining condition. Neuronal networks have had a significant influence on artificial and biological intelligence.
Every neuron is connected with approximately one thousand other neurons. The junctions are called synapses. It is the power of one hundred trillion synaptic connections that encode the biggest part of information in our brain. 20
It is still being researched on how neurons influence other neurons with electric activity.
Neuronal networks are very powerful and independent of substantial machine-oriented details of their construction, which is an indicator for their substrate independence. 21
Evolution has created neuronal networks in a way, that they are universal, which means that any function can be operated by simply increasing the synaptic power. Our human neuronal networks are complicated, because evolution does not reward effective simplicity. The conception of our biological systems have shown to be efficient, which they prolifically proved over the precedent centuries. 22
A method of operation of this matter is portrayed in two adjacent neurons, that are active simultaneously (“fire”), resulting in the reinforcement of their synaptic connection.
Hence, they learn to trigger the “fire” among each other. Donald Hebb is responsible for these research results and framed the famous slogan “fire together, wire together”, which was derived from the previously explained context. In other words: neuronal networks are able to learn when they are repeatedly confronted with the same situation or with the same information. This kind of confrontation is also called “training”, when we talk about artificial neuronal networks. In the human relationship, we call it “studying”, “education” or “experience”.23 With regards to our biological intelligence, there is still enough data to be researched on. Moreover, the issue of learning is far more complex in a wide range of cases. Research is still required in order to discover, what mechanisms our brains use to learn. Nevertheless, there is no scientific evidence of a violation of the laws of nature.24
Artificial neuronal networks, that set our AI systems up, are based on the Hebbian deterministic theory, that synapses are updated over time. Following, networks are able to carry out complicated calculations, if they provided with enough data.
Artificial and biological neuronal networks have their differentiated distribution of tasks in common. In contrast to feed forward networks, recurrent networks can organize information in order to stream them into different directions.25
The following figure demonstrates the differentiated approach of feed-forward to recurrent networks:
Abbildung in dieser Leseprobe nicht enthalten
Figure 1: Comparison of a feed-forward and a recurrent network26
It is important to understand the way neuronal networks work, because theyrecreate the computing potential of the brain, which is a premise for further technical development.
The derivation of these facts is that matter can be arranged under the condition of natural principles, with the objective to learn, calculate and remember, without the requirement of being biological.27
2.2 Can machines think? Turing Test
“It seems probable that once the machine thinking method had started, it would not take long to outstrip our feeble powers… They would be able to converse with each other to sharpen their wits. At some stage therefore, we should have to expect the machines to take control.”28 – Alan Turing
Alan Turing believed that computers can execute intelligent behavior, with the premise of being designed and educated accordingly. His vision was that humans and machinery would be indistinguishable in their behavior. Therefore, he introduced the Turing test. Turing intended to test whether a machine has obtained intelligence on a human level. The procedure of the test starts with a human being, being an interviewer, having an interrogation with two unknown interlocutors, by means of a keyboard and a screen. One interlocutor is a computer and the other one is a human being. The test person does not have any visual or hearing contact with the other two players. If the interviewer does not achieve to recognize who the machine is, the machine passes the test. Consequently, the machine obtains the status of disposing human intellectual power.29 Nevertheless, the so-called imitation game has been criticized for rather testing human credulity, than real artificial intelligence.30
2.3 The Winograd Challenge exuberates the thinking of machines
The Winograd challenge is a competition to the Turing test. While Turing has been criticized for testing human credulity, the Winograd challenge tests common sense, that is widely missing in deep learning systems.
A typical Winograd question looks like the following:
The town of councilors refused to give the demonstrators a permit, because they feared violence. Who feared violence?
Answer 0: the town councilors
Answer 1: the demonstrators31
The challenge typically consists of two halves and discusses common sense topics such as convenience, conditions, or weight. Human beings routinely adopt their knowledge about reality when they analyze a sentence.32
The consequence of the Winograd challenge is the discovery of common sense in machines. Both Turing and Winograd aim to discover the human component in the cybernetic matter.
The derivation of these tests provide insights to the inner life of a technology. They function as ascertaining method to ensure the thinking ability of a computer. In following chapters, it will be discussed what premises are required in order for a computer to be able to connect an overall context.
Since the development of computerization has only initiated, it will be engaging to see what further development the progress in engineering will bring. Answering the question whether machines can think or not, the answer depends on the viewpoint. The ascribed human-like capabilities of machines are perceived differently by various parts of society. As a result of information provision, computers obtain the possibility of being knowledgeable. This concept implies the dependence on an external influence. Thereby, the controversy arises whether a machine is responsible for the accomplished inner processes, or whether their entirety can be assigned as the product of human intelligence.33
Despite the tentativeness of this circumstance, it is meaningful to execute tests on computers, robots or machines. Considering occurred astonishing moments in the sphere of science, where computers outperformed or self-improved themselves drastically, it will remain significant to record the progresses a machine enforces.34
2.4 Comparison of artificial intelligence and biological intelligence
The conclusion of the Turing Test and the Winograd challenge is the demonstration of lacking capabilities of computational systems in comparison to human capabilities, referring to non-calculational matter.
As it has been portrayed in the previous chapter, conspicuous differences between human and artificial intelligence is the broadness, on which expertise is enabled to take place. Human beings constitute a unique setting proposition in the world, due to their ability to outperform value-adding tasks in a self-organized society and the competence to extrapolate uncommon terrain. This aspect is aligned to the conception of self-created scopes, the intrinsic and self-established conduction of hypotheses in connection of contrivance and contemplation of non-existing phenomena. This affair is the most distinctive aspect of human beings in comparison to any other familiar living creature, the reason of our peak positioning and the present most distinctive coefficient, differentiating artificial and biological intelligence. This state of affairs is not only given, due to the arrangement of our intelligence, but also and more importantly by the inevitable virtue of the existence of our consciousness. The topic of consciousness will be discussed in the following chapters. As to be reported for this chapter, human consciousness has given us the proficiency and intuition to grasp the inexistent, the yet- to be discoverable and the possibility of wrongness of verified information. Mistrust and everlasting curiosity have led us to our current state and artificial intelligence is just one beneath different various by-products of human intelligence.35
In this area, it is reasonable to address creativity as well. Creativity is ought to be one substantial aspect, shaping human experience and driving progress. It is often alluded to portray the human soul. Music is the science of putting tones and sounds in sequence, whose combination is expected to exhibit a certain harmony. This composition is driven and carried out by human creativity. However, the creation of EMI (Experiments in Musical Intelligence) aims to create symphonies that are indistinguishable from man-made creations. Submitted rules are processed by Machine Learning systems and consequently translated into new compositions. When pianists and EMI dueled in a showdown, without the audience knowing what part was composed by “classical music by a computer “ and which was the human part, spectators praised the machine for its soulfulness and emotional effect.36
Following the interviewed expert Dr. Jan Sölter, he appeals to the boundaries of when creativity can be called likewise. The intelligent reproduction of rules or positive outcomes as a result of numerous trial and error endeavors cannot be called creativity on human standard. Notwithstanding, this does not deny that we are approaching the state of indistinction of the exposed duality.
2.5 Current State of Artificial Intelligence
Artificial Intelligence is a scientific multidisciplinary field whose goal is to create intelligent machines. Accordingly, artificial intelligence has to be differed into four subsequent categories. A subset of artificial intelligence is machine learning (ML). Statistical techniques are used; thus, machines can learn. This process is common as training, using an algorithm that continuously enhances its performance on a specific objective.37
Another sub area of AI is Reinforcement Learning (RL), which deals with software agents that are supposed to exert goal-oriented behavior. It is confronted with rewards and penalties by a trial and error environment in order to move closer to achieving a specific goal.
Besides, Cognitive Analytics (CA) deals with cognitive behavior that resembles human thinking. A cognitive system puts concepts and relationships in context to one another and improves its performance by two distinguishing methods. The first one is by following conversations from human beings and consequently using the given assessment. The other method is the usage of background information, which is mainly required for the determination of language. The Winograd challenge tests the mechanical understanding of context, as we learnt previously.
As a last aspect, Deep Learning is ought to be mentioned. It is created in the manner of patterns of neuronal networks. The name is referred to the layers of neurons that support the learning of data with the goal to achieve enhanced performance previously.38
Due to these automated processes, different industrial sectors can profit by adapting artificial intelligence into their method of operating and achieving productivity augmentation. Hence, certain tasks can be automated. For instance, recruiting suitable talents or adjusting marketing towards specific customer needs are techniques that adopted Artificial Intelligence make an advantageous entity towards achieving productivity enhancement.39
Referring to the current state of artificial intelligence, significant developments have been achieved in recent years.
A result of the significant development of Artificial Intelligence is the emergence of topics, such as security, data protection and concerns regarding the correct usage of new technologies, approaching the center of attention in the international debate.
Organizations such as the Future of Life Institute (FLI) have initiated summits with the aim to appeal to the importance of data security and to achieve a consensus about sensitive issues, requiring a distinction of right and wrong in the moral and ethical context. Famous personalities such as Elon Musk, Stephen Hawking or Bill Gates and other significant experts in the field of science and innovation have shown the willingness to progress in cybernetic questions to contribute to more awareness and attentiveness, by adhering with scientific research or monetary support. Significant advancements have been achieved in AI. It provides beneficial contributions in health care, music production, warfare, creditworthiness, marketing etc. The sub parts of AI, that were described in this chapter are used in almost every industrial field, where technology is required. The biggest challenge of the utilization is the application of AI to process change management.40
2.5.1 Artificial General Intelligence
“The first ultra-intelligent machine is the last invention that a man need ever make, provided that the machine is docile enough to tell us how to keep it under control.” – Irving J. Good 1965 41
With regards to Artificial Intelligence, the term Artificial General Intelligence (AGI) has acquired rising significance. Artificial General Intelligence means intelligence on human level. In the world of AI science, AGI is also coined as singularity, implying the uncontrol-lable, irreversible influence of technological growth on human society. It is a predominant objective of AI- scientists to achieve the creation of AGI. Nonetheless, 42 the world’s most important AI experts do not come to terms whether Artificial General Intelligence will ever be attainable. It is difficult to predict a forecast in uncertain subject matters. Certain ele-ments are still to be exceeded. The first determinant aspect is the adequate computational power that meets the standard of the human brain.43
Hans Moravec, a significant scientist in the research and development of artificial intelli-gence has conducted tests on image processing tasks between computers and human beings. He aimed to discover what commitment our brain executes by processing images. He ob-served the undertaken mechanisms of a retina before sending results via optic nerves to the brain. The result has shown that a computer would need a billion FLOPS and that the brain executes calculations multiplied by ten thousand times more than an artificial retina.44
FLOPS are a measurement procedure to express and indicate the performance capability of a supercomputer. The acronym is an abbreviation of floating-point operations per second. The quantum computer is the only one owning the capability of surpassing the supercom-puter. For instance, Google’s 56 qubit quantum computer, named Sycamore owns a compu-tational power of 200 PetaFLOPS and would outperform a calculation in 200 seconds, which a supercomputer would need ten thousand years for.45
The human brain disposes a calculation capacity of 10¹³ FLOPS, which is equilibrating the optimized capacity of a computer of present days. In other words, this implies that Artificial General Intelligence can surpass the human level of capacity. Artificial General Intelligence can become true in the upcoming decades. With rising technological innovation, humankind can inaugurate the possibility of surpassing organic with mechanic 46 intelligence. There is no scientific evidence of unfeasibility. Former arguments against the attainability of AGI crea-tion have been the large amount of expenses or insufficient hardware, which are now refut-ed. Certain data have contributed to the awareness of the rapid advancement. Moreover, a worldwide community of AI-scientist work on existing challenges together, which implies that progress is made and will become more common in the upcoming times, when present-able outcomes are discovered.47
2.5.2 Society’s perception of artificial intelligence
Concerning the nowadays world in reference to its perception of artificial intelligence, the significance and importance of state-of-the-art technology differentiates vigorously. Not only scientists argue about future outcomes that artificial intelligence will deliver to our every-day lives. Especially the human civilization holds distinctive assessments about the interference of inorganic intelligence in our workspace and our private sphere. The holistic spectrum of concerns with regards to AI is exceedingly broad, which is ascribable to the wide range of interferences, artificial 48 intelligence can have. For instance, it can influence various manners of industrial sectors. Thus, it either holds potential for efficiency and effec-tiveness for human beings and can be seen as an opportunity. On the other hand, artificial intelligence can be perceived as a threat, owing to its capability of replacing a human being in his value adding contribution to society as a whole or simply to his or her ascribed work-place. 49
Additionally to the concerns and opportunities in the corporate sphere, Homo Sapiens can experience a transformation of its own significance in the holistic hierarchy of beings in the world and could lose its position of the “apex predator”, the highest entity of the food chain, when losing the position of being the most intelligent entity in our present community on this planet. Still, certain fractions of civilization discern the digital transformation as a positive aspect, characterized by chances and positive outcomes or the opposite.50
According to Max Tegmark, the debate of how we chose to handle artificial intelligence, is the most important one of our time and we need to ask ourselves how we need to act in or-der to create the future we strive for.
The approach to artificial intelligence is able to be differentiated by distinguishing three different groups of thinkers. These groups are relevant, because they do not only represent the judgement of society, but also the conceptions of leading experts conducting the de-bate. Hence, the distinction is illustrated between digital utopians, techno-sceptics and the Pro-Cautionary defenders.
Starting with the first group, digital utopians support the viewpoint of digital life being the natural upcoming step in the evolutionary advancement. According to this group, the emancipation of digital life would have a positive outcome, under the premise of conceding intellectual freedom to this entity instead of trying to subjugate it. Hence their futuristic motivation is grounded in the immersing of innovative technologies and welcoming them as a life enhancing possibility, leading to the disclosure of unaffected terrain
The second category, portraying a long-ranging group of our society are techno sceptics. They do not fear technological defiances, attributable to their disbelief of high-tech su-premacy. This skepticism extends to the denunciation of the contentious singularity and superintelligence, also referred as to Artificial General Intelligence (AGI).
The main consequences of demurral towards AI certainly leads to a derogation of progres-sion.51
Lastly, the Pro-Cautionary defenders are to be disclosed. This circle has become popular under the name of the Beneficial AI Movement (BAI), representing renown personalities, such as Stuart Russell, Geoffrey Hinton or Peter Norvig. The core statement encompasses the objective of redefining artificial intelligence concerning its purpose. Thus, AI is de-signed to be beneficial for humanity, including intellectual, moral and ethical parameters. The BAI understands the emergence of AI as a permanent responsibility towards the crea-tion of new lifeforms, that affects all parts of the world in their subsequent circumstances.
[...]
1 c.f. Nilsson 2009: .5
2 Goertzel, 2007: 21
3 c.f. Van Duin/ Bakhshi 2017
4 c.f. Davies 2014: 61
5 c.f. Specht 2019: 1
6 c.f. Murphy 2016: 3-4
7 c.f.Anderson et al. 2018: 83
8 c.f. Humanity+ Press Language 2016: 119
9 c.f. Bridge 2010: 15
10 c.f. Tegmark 2017:
11 c.f. Tegmark 2015: 242
12 c.f. Wahlsten 2019: 36
13 c.f. McCallum 2019
14 c.f. Wilson/ Daugherty 2018: 116
15 c.f. Hoppfield 1995: 809
16 c.f. Pennartz 2015: 96
17 c.f. Kuvich 2017: 149
18 c.f. McGaugh 2012: 135
19 c.f. Null/Lobur 2014: 760
20 c.f. Hebb 2005: 175
21 c.f. O'Connell 201: 183
22 c.f. Blumenfeld 2014: 388
23 c.f. Aggarwal 2018: 25
24 c.f. Sanz 2013: 396
25 c.f. Kumar 2017: 82
26 c.f. Pekel/ Soner Kara 2017: 159
27 c.f. Zador 2019: 1-7
28 c.f. Copeland et al. 2017: 400
29 c.f. Levesque 2017:129
30 c.f. Epstein/ Roberts/ Beber 2008: 37
31 c.f. Davis/ Morgenstern/ Ortiz 2017: 16
32 c.f. Levesque, 2011: 534
33 c.f. Khosrow-Pour 2018: 517
34 c.f. Banik 2019: 19
35 c.f. Sarangi 2018: 82-90
36 c.f. Song 2017: 1520-1540
37 c.f. Hogarth/ Benaich 2019: 37-40
38 c.f. Kim 2017: 2-17
39 c.f. Derwin 2019: 28
40 c.f. Guida 2018: 2-19
41 c.f. Russel 2019: 105
42 c.f. Ford 2018: 528-530
43 c.f. Childers 2019: 24
44 c.f. Maitra 2020: 320-326
45 c.f. Anderson/ Rainie/ Luchsinger 2018: 73
46 c.f. Sabanovic 2010: 439–450
47 c.f. Tegmark 2017: 286
48 c.f. Lee 2019: 305
49 c.f. Springer Science 2017: 26
50 c.f. Guenduez 2016: 7-8
51 c.f. Dutton 2018