Welcome to Foresight For Development

Artificial Intelligence II

The Future of Artificial Intelligence

The human-machine frontier

by Dr Anthon P Botha - TechnoScene

TechnoScene (Pty) Ltd
and
The Graduate School of Technology Management, Faculty of Engineering, Built Environment
and IT, University of Pretoria
South Africa

This email address is being protected from spambots. You need JavaScript enabled to view it.

 

In the beginning there was chaos…
And God created mankind…
…from the tree of knowledge of good and evil he shalt not eat…
Then he sinned and was doomed to toil hard…
So man created the machine…
…and the machine is getting knowledge of good and evil…

In this article we explore the human-machine frontier by looking at artificial intelligence (AI), its state of development and areas where the largest impact on life and work is expected.

 

What is artificial intelligence?

It is not a new unique technology, but the combination of sensing, comprehending and acting in machines (Purdy and Daugherty, 2016). Its theoretic and technological basis has been developed over a long time. What is unique to AI is that all three its basic capabilities are underpinned by the ability to learn from experience and to adapt over time. The term “Artificial Intelligence” was established in 1956, but the roots go back to Alan Turing’s ground-breaking paper in 1950 in which he poses the question “Can machines think?” (Turing, 1950). AI has been applied in industry for some time, but its appearance in our everyday lives, as it becomes part of technological convergence in smart devices, has made it a disruptive factor. It is now fulfilling functions that humans used to do with the assistance of technology. Sensing includes computer vision and audio processing, comprehending includes natural language processing and knowledge representation and acting involves machine learning and expert systems. This provides us with virtual agents, identity analytics, cognitive robotics, speech analysis, recommendation systems and complex data visualisation. It is, however, the ability that machines are acquiring to improve themselves and to develop algorithms and software that are so complicated that humans cannot understand it anymore, that both distinguishes machine intelligence from human intelligence and that is the cause of great concern among some that humanity may lose control over the world. The current drivers for progress and adoption of AI took shape since 2010. Three interrelated and supplementary factors are the availability of big data from a plethora of sources including e-commerce, businesses, social media, science, and government; which provided raw material for dramatically improved machine learning approaches and algorithms; which in turn relied on the capabilities of more powerful computers (Holdren, et al, 2016). AI is described as either being narrow, general or super-intelligent. Narrow AI addresses specific application areas such as playing strategic games, language translation, self-driving vehicles, and image recognition and normally underpins commercial services such as trip planning, shopper recommendation systems, and advertisement targeting, and is finding important applications in medical diagnosis, education, and scientific research. General AI, including deep learning, on the other hand, refers to future AI systems that exhibit intelligent behaviour, at least as advanced as humans, across the full range of cognitive tasks. Deep learning uses structures similar to the human brain, consisting of a set of units (or “neurons”), where in each unit multiple input and output values are generated. Facial recognition is an example of deep learning. General AI does not flow naturally from expanding narrow AI. Machine learning, where a body of data is used to derive rules by the machine itself, is different from expert systems, where humans provide the rules. It is evident that machine learning is more appropriate to the complexity regime where emergence is the rule. Artificial super-intelligence (Gurkaynak et al, 2016) involves AI much smarter than the best human brains in practically any field. This is the AI that is feared by some to be so goal-oriented and committed to achievement that it may destroy the world.

 

It is all in the metaphor

Artificial intelligence as a notion has been with us for a long time. To establish for how long, one has to look at the metaphor and worldview of what AI is. Carbonella et al, (2016) write that the aim of AI is to imitate human behaviour and that it has been present in many ancient mythologies and most religions. Engineers since the 12th century tried to create artificial humanoids. Leonardo Da Vinci made sketches in the late 1400s of a robot. The interest of people to create devices that replicate human behaviour fits the metaphor that man is the new creator. This worldview is shifting in modern times to a view that the machine may become the new creator. This has been driving science fiction film and its typical post-apocalyptic and dystopian scenarios of world doom. Another metaphor that emerged since the 1950s with the advent of the electronic computer is that computational systems are brains. This has led to similar views that the brain is a computer. Mental processes are now equated to computer algorithms. This merging of metaphors has led to the development of science that emulate the human brain (neural networks) and understanding of the human brain to build better computers, e.g. the Human Brain Project (HBP, 2016), a European Commission Future and Emerging Technologies Flagship, aiming to put in place a cutting-edge, ICT-based scientific research infrastructure for brain research, cognitive neuroscience and brain-inspired computing.

 

Human-machine partnerships in the workplace

Fear has always existed in debates of technology vs labour that humans will be replaced by machines in the workplace. This fear is now extended from blue collared workers to white collared workers. Will it soon be only steel collared workers that do the work? This notion is wrong, not only since robots will not be made of steel in future but of smart alloys and composites, but because humans have an important role to play. It is more about automatability of tasks than automation where humans are replaced by machines. Thus we should be speaking of human-machine partnerships in the workplace. The paradigms should shift from total task take-over as in automation to co-thinking, co-learning and co-working. In future machines will increasingly work and behave like humans. This means that creativity, intuition, motivation and ethics may be common to both people and machines. Will machines have a conscience? Human-machine algorithms will be developed and human-machine relationships will be challenging human resource management experts. Autonomous machine decision making will have to be trusted by humans. Human rights and machine rights may co-develop and require creative thinking from legal professionals. When will machines demand remuneration and time off? Relationships between humans and machines may call for a new industrial psychology.

 

In machines we trust

One aspect that will have to develop in the human race is to trust machines. If we entrust life-critical functions such as medical diagnosis and assistance devices and autonomous vehicles to artificial intelligence, we better be sure we are comfortable with them making the right decision. Machine decision making is still in early stages of development. It is largely rule-based where observation is tested against a database of rules. Even if the machine is self-learning, it builds experience in terms of previous decisions that have been right or wrong. In life supporting environments there is no place for error in learning. Humans, to a large extent, make decisions based on intuition. Machines will have to be smart enough to emulate that. This implies the inclusion of a conscience as well, to create a notion of machine emotion reflecting guilt or achievement. Decision-making based on the recognition of unusual behaviour should be part of the deep learning process in intelligent machines that are entrusted to life-critical tasks. Once these qualities are embedded in AI, humans will start to trust machines to the extent that they will allow them to provide independent guidance, without human interventions or approval. Hengstler et al (2016) state that the concept of trust provides a valid foundation for describing the relationship between humans and machines. It depends on trust in the technology, which in turn depends on its state of emergence or maturity, perceived risk in terms of uncertainty and impact, and trust in the firm innovating the machine and its domain knowledge and track record. But will machines be able to innovate?

 

Can a machine own IP?

Inventing machines have been demonstrated (Hantos, 2016). Already, governments are considering introducing new systems for protecting intellectual property that is generated by a non-human entity. This means that the “definition of ingenuity and innovation, which up to now has been the provenance of mankind, would be reshaped to accommodate these new inventors of the cyber world”. The question remains whether with the “open everything” movement and the increasing trend to open up research data for global use, patenting will still be the norm? Machine invention will be so fast that the patenting processes may fall behind (unless AI lawyers do that). The norm may become the same as what has been applied in the software industry for decades – do not patent, just be first to market. This may become an adopted principle in machine invention and turn the IP world on its head through radical disruption.

 

Humanitarian conflict resolution

Bob Dylan, who recently won the Nobel Prize for literature, sings in The Times They Are a-Changin':

“Come mothers and fathers
Throughout the land
And don't criticize
What you can't understand
Your sons and your daughters
Are beyond your command
Your old road is
Rapidly agin'
Please get out of the new one
If you can't lend your hand
For the times they are a-changin'.”

Olsher (2015) writes that “Truly understanding what others need and want, how they see the world, and how they feel are core prerequisites for successful conflict resolution and humanitarian response. Today, however, human cognitive limitations, insufficient expertise in the right hands, and difficulty in managing complex social, conflict, and real-world knowledge conspire to prevent us from reaching our ultimate potential.” He further states that AI systems are capable of understanding how people from different groups and beliefs view the world. AI can simulate their reactions, and combine this with knowledge of the real world, to persuade, find negotiation win-wins and enhance outcomes, avoid offensive behaviour and provide peacekeeping decision tools. In the past AI has often been entrusted only to do complex situational analysis, but not to intervene in negotiations. In a recent scenario developed by the author for the future of a metallurgical industry, he writes about a young person working in the industry in 2030: “She can now work for both South African and Chinese companies without fear of intellectual property infringement. The breakthrough came when the artificial intelligence lawyers agreed in a world wide web summit on quotas, target markets and role players”. AI has the potential to bring emotionless interaction, with calculated understanding of risk and consequence, and with no preoccupation or self-interest to geopolitical conflict resolution, peace-keeping, trade agreements and social advancement. These type of agreements would consider the regional, economic, digital and religious divides that exist among humans. Humans, especially those in countries that have been under colonial rule and influence for a long time, may start fearing colonialisation by machines. However, inclusive development of such algorithms will take cognisance of all kinds of fears and distrust and address them in an intelligent way. These systems will provide for conflict modelling and resolution; persuasion based on values, beliefs and religion; reconciliation, social media analysis where digital democracy is a strong trend; and early warning for conflict or crowd unrest.

 

It’s all about the data

AI is often described as a continuum of data plus algorithms plus computing capacity. Whilst this is true, it is the data science and conceptual interpretation of knowledge that drive AI. In the era of big data, AI depends on advanced data analysis, pattern recognition, deep learning algorithms and data governance. Big data is available at a variety of levels: Facebook can map faces and pictures on a large scale with more than 1.7 billion members; Google Earth data; massive research data repositories that are made openly available through science clouds, also for citizen science; consumer behaviour data; data used by governments for inland security; longitudinal health research data; environmental data; climate change data and many others. Often, answers to global challenges lie locked up in existing data. Artificial intelligence holds the key to mining and rediscovering those solutions in existing data, as well as in massive data streamed live. Venture capital firms are favouring start-up investments in data analytics. By adopting a holistic approach AI is not only used to unravel complex data into meaningful understanding, but also applied to form models of how the world works. A particular challenge the world is faced with is to deal with migration and urbanisation. Intelligent and smart cities will be the home for more than 75% of humanity after 2050. Supported by the Internet of Things, massive data streams will have to be monitored (Rathorea et al, 2016), interpreted and reacted to in these cities of the future. Intelligent sensors will be required to trigger AI action, receive feedback and to adjust traffic flow patterns, waste management, utility services, air quality, criminal activity, disaster management, health threats, urban agriculture and health services.

 

The end of white collar jobs?

Hantos (2015) writes that apps are available today that conduct routine legal research that junior lawyers normally would have done. These apps sift through large volumes of case law, looking for legal concepts and extract patterns. Although many consider this as a future threat to lawyers, the impact on the legal industry is very much a present issue (Abramowitz, 2016). Cutting edge search and practice management software employ AI and machine learning. AI is also providing opportunity for early stage start-ups and implementation in more mature law firms. Instead of seeing AI as a job limiting force, alternate approaches consider how AI can give law firms and their clients an edge. The same fear exists in auditing companies. Artificial intelligence applications continue to be taken up in both accounting research and accounting practice (Sutton, 2016). The importance of remaining on the leading edge of technology to provide leadership and competitiveness to the profession ensures that artificial intelligence capability and application is rapidly expanded. This has resulted in fears of the demise of an accounting profession. Reams (2016) states that this potential threat should be seen as an opportunity. Chief Financial Officer advice will soon be obtained from AI agents. However, time-consuming accountancy and auditing tasks that used to consume lots of time from highly qualified accounting professionals will now be taken over by artificial intelligence. This does, however, not eliminate the need for accountants, but should rather empower them. They can now have more direct interaction with clients and work in a more effective way. An important application of AI in accountancy is to combat cybercrime.

 

AI in your pocket

We all carry a significant amount of AI in our pockets and handbags. Smart phones have several AI supported functions (Brandon, 2016) and these are increased with each release of a new model. These include maps suggesting less congested routes; automated and intelligent text responses; facial recognition of photos; transcription of voice mail and recordings into text; the phone coming active when it is picked up; voice assistance (e.g. Siri on Apple devices); making automated videos of events, whilst identifying people and places, automatic grouping of news feed and notifications, and proximity awareness, where the phone reduces power when closer to the ear or face.

 

Machine consciousness

Will our intelligent machines ever feel for us? The question relates to whether there is a relationship between intelligence and consciousness. Consciousness is the state or quality of awareness, intelligence is the ability to acquire and apply knowledge. To answer this question, one has to look at the dissociation of consciousness and attention in humans. Haladjian (2016) is of the opinion that one can program ethical behaviour based on rules and machine learning, but not reproduce emotions or empathy, but at the same time admits that this may be possible through simulations. There is a deep relation between emotion and cognition in human intelligence. Emotions in machines will thus not be able to be programmed through control systems, but will have to be a learning and simulation issue. Haladjin argues that machines may develop access consciousness, but not phenomenal consciousness. Block (as quoted by Kriegel, 2006) makes a distinction by defining phenomenal consciousness as what it feels like to be conscious; whilst access consciousness deals with readiness (availability) to reason and control of action and speech. As Chris de Burgh sings: “It's the classical dilemma between the head and the heart”. The current state of AI is that it will act on the basis of rationality, but not empathy. Is there a relationship between consciousness and conscience? Conscience generally deals with an inner feeling or awareness of rightness or wrongness of behaviour, based on morality or value systems. Consciousness relates to an awareness of the environment. Machines can be very aware of where they are and in what context. The open question is whether they can distinguish between right and wrong? Does this mean that intelligent machines may be corruptible?

 

Intimacy with an AI robot?

A popular view is that the sex robot industry will drive the development of emotional and human attachment with machines. Some futurist views point out that human-machine weddings will be regular by 2050. A distinction should be made about a robot as a partner and one used as a sex toy (Eveleth, 2016). A real sex robot should be able to respond very quickly to a partner’s facial expressions, and predict and initiate actions. A partner must perform the emotional role that real partners do. Multiple disciplines are required to develop the ultimate machine partner. New technologies like artificial skin with different textures will have to be developed – silicone that is currently used is just not realistic. And then there is the question of battery life… AI in a partner robot will have to be developed so that the robot can engage with and learn from its human partner. The fact that machine emotional consciousness is in the early stages of being understood will result in non-ideal partner machines for many years to come. Eveleth writes: “Computers might be able to beat a human at chess, but sex is more like a dance; each partner has to predict and respond quickly to movement. And right now, artificial intelligence and natural language understanding is still a long way from being convincing”

 

Fears about AI

Like it is with so many new technologies and their impact on the world, AI has been introduced by science fiction movies. To name a few: 2001: Space Odyssey, The Terminator, The Matrix, Transcendence, Ex Machina. Most of these represent an end state where humans are not able to control their own creations, leading to an apocalyptic end to the civilisation as we know it. Much of this perception is fuelled by the AI in military systems in use. Recently well-known scientists and business leaders have joined up and expressed their concern about the proliferation of AI, mainly in smart weapons systems. Stephen Hawking, Elon Musk, Bill Gates and Steve Wozniak expressed concerns related to offensive autonomous weapons (Sainato, 2015 and Gibbs, 2015). This opinion was also supported by Google DeepMind chief executive Demis Hassabis. (DeepMind) is leading artificial intelligence research for positive impact. It is one of the research programs that are pushing the boundaries of AI. Sainato (2015) writes: “The threats enumerated by Hawking, Musk, and Gates are real and worthy of our immediate attention, despite the immense benefits artificial intelligence can potentially bring to humanity. As robot technology increases steadily toward the advancements necessary to facilitate widespread implementation, it is becoming clear that robots are going to be in situations that pose a number of courses of action. The ethical dilemma of bestowing moral responsibilities on robots calls for rigorous safety and preventative measures that are fail-safe, or the threats are too significant to risk”. Does this mean that we will have a United Machines Organisation alongside the United Nations soon? The introduction of artificial super-intelligence may at its extremes make us immortal or cause the extinction of humanity (Gurkaynak, 2016). Evolution has given humans their strongest instinct, that of survival. That is, for example, the sole reason why the field of medicine exists. Instinctively then, we will try to eliminate the threat that artificial super-intelligence has on us. The down-side of this lies in the fact that many of the good impacts of super-intelligent machines aimed at our survival will be eliminated as well. It is a matter of choice and risk.

 

A noble cause

AI specialists are of the opinion that super-intelligence may be with us by 2060 (Tegmark, 2015). This timing also coincides with Ray Kurzweil’s “Singularity”. The Singularity is an era in which intelligence in the world will become increasingly non-biological and trillions of times more powerful than it is today. It speaks of the dawning of a new civilisation that will enable us to transcend our biological limitations and amplify our creativity. It coincides with the point in time when a computer will be available at less that US$ 1 000 that will have the computing power of the entire human race. Tegmark states that AI can become dangerous when programmed to do so, or when programmed to do something beneficial, but develops a destructive method to achieve this goal. It is all about control – “People now control the planet, not because we’re the strongest, fastest or biggest, but because we’re the smartest. If we’re no longer the smartest, are we assured to remain in control?” Given the fears about AI, research on AI today is largely driven by a noble cause. The primary goal of AI is to build machines that are better at making decisions. But, as Stuart Russel (Russel, 2016) remarks: “Being better at making decisions is not the same as making better decisions”. It is about alignment of utility functions (goals) of machines with human values. Humans have the ability to learn what utility functions should be, as well as what is desirable. Intelligent machines will thus have to learn about value alignment as well as problem solution. By designing thinking machines that are value-aligned, the human race may have an opportunity to review its own view on values.

 

A new rule of law?

Legal systems are not yet accommodating AI. Cerka et al, (2015) writes: “The ability to accumulate experience and learn from it, as well as the ability to act independently and make individual decisions, creates preconditions for damage”. Neither national or international law recognises AI as a subject of law yet. The legal principle followed at present is that the human being that uses a machine as a tool or on whose behalf the tool has been programmed, is responsible for the damages caused by the tool. The notion of AI-as-a-tool is thus dominating liability thinking at present. The complication with AI, however is that it can programme itself to the extent that humans do not even have the ability to undo that programing. Furthermore, the daily application of AI in the many devices humans use is different from the institutional uses of AI. This will be complicated further with AI embedded in the Internet of Things, where humans may be out of the loop and AI will operate its own solution among “things” or objects. It is clear that we are presently in a situation that our future industries and daily lives will be determined to a large extent by a technological entity that is outside the current rule of law. The operating systems of AI are constantly evolving and changing and the software is dynamically improved by AI itself. This legal intervention should be universal, regardless of the changes and be constantly amended to create a dynamism for application. Furthermore, in the context of increasing globalisation, regulation and legislation of AI cannot be limited territorially. The emphasis may thus largely be on finding appropriate international law for AI governance. It will touch both civil law and common law. A large uncertainty currently exists on liability for damages. From a legal perspective, AI is viewed on the basis of the factor of a thinking human being and in terms of rational behaviour, thus systems that think and act like a human being; and systems that think and act rationally. These factors demonstrate that AI is different from conventional computer algorithms. AI comprises systems that are able to train themselves based on individual experience. Cerka et al (2015) state: “This unique feature enables AI to act differently in the same situations, depending on the actions performed before. This is very similar to the human experience. Cognitive modelling and rational thinking techniques give more flexibility and allow for creating programs that can ‘understand’, i.e. that have traits of a reasonable person (brain activity processes)”. At present it seems that AI can be treated as a source of danger, and the legal person or entity on whose behalf it acts as a manager could be held liable without fault. If the dangerous activities are profitable, the person who profits should be held responsible and compensate for damage caused to society from the profit gained. Such a legal person or entity should be required to insure against civil liability. New liability regimes will have to be developed for a situation that may be too complex for present law to deal with. Maybe the next generation AI lawyers will take this up as their first task…

 

Will AI make us safe?

In an interconnected world, living in interconnected cities, relying on our appliances that talk to each other on the Internet of Things, doing our financial transactions on interconnected banks and relying on driverless transport and intelligent medicine, communicating with our loved ones and friends on interconnected social media, we are becoming increasingly vulnerable to cyber attack. Just as we can get our ideas and creative products instantaneously distributed, we can all be attacked at the same time. In classical warfare, attack to defence ratios are typically 1:2; in cyber attack this can increase to 1:10. Counterattack can easily be predicted or sensed and counter-counter mechanisms can be devised on the fly. We are entering an era of cyber crime and cyber warfare where the expenditure on development of these new technologies are higher on the perpetrator side than on the security side. Command and control is not linear and based on cause and effect reasoning, but complex and emergent. AI has much better capabilities of thriving in these virtual battlefields of the future than the best military minds of today. Conflicting forces have access to the same technology that is applied both for committing and supressing crimes. CybeRisk states: “With statistics suggesting that assaults on individuals, corporations, and government bodies account for almost $400 billion in lost revenue annually, and some 90% of companies admitting to having been victim to some kind of attack – figures that translate to 18 individuals per second being affected by cyber-crime – countering these threats is a real and ongoing concern, for enterprises” (Cyber security, 2016). Although prevention of cyber threats and attack avoidance is ideal, rapid response after an attack and recovery is crucial. There is no time for an investigation, AI has to be employed to figure out the damage done, the source of the attack and countermeasures to avoid a repeat. The form of attack morphs and mutates all the time and fast learning about patterns of targeting and damage objectives are required almost instantaneously. Adaptive and machine learning algorithms must be integrated with transaction software and be able to predict possible attack styles for different transactions and build countermeasures. When humans commit a crime, the procedure is to arrest them, put them through trial, and once guilty, discourage them from future criminal activity and rehabilitate them. A moral and ethical environment kicks in where the rights of the criminal must also be kept in mind. What equivalents will we have in cyber crime? Once the AI recognises the offending intelligent machine, will there be legal ways of removing that system from the network through damaging counter-attacks or just isolate them through countermeasures and security walls? Just thinking this way, humans err by equating cyber justice to human justice. This shows the burden on legal and ethical aspects of AI that have to be developed. International treaties about cyber crime and cyber warfare are being discussed, but should be in place where both humans and machines agree to them. Although the future picture is not clear yet, it is likely that the words “crime”, “warfare” and “AI” will be used together to an increasing extent. We tend to think about humans fighting humans and machines having their own war, but the worst case scenarios, like those expressed by the doomsayers, may include machines fighting humans. Whilst in control, part of warfare has always been humans destroying machines. Will machines make us safe? Only when they do not adopt our weaknesses as a human race.

 

Our economies got stuck, AI to the rescue?

The factors of economic growth are changing. Capital and labour are no longer the sole contributors, but a new factor of production is emerging and it has the potential to transform the basis of economic growth for countries across the world. AI has the potential to overcome the limitations of capital and labour in the modern economy, which is a mix of many traditional economies, and open up new sources of value and growth (Purdy and Daugherty, 2016). In some developed economies, AI has the potential to double growth by 2035. AI is not simply another enabler of growth, but will do so by transforming our thinking about how economic growth is created. Several economic waves have determined economic growth of nations. It started with the agricultural economy, followed by the resource economy and the industrial economy. With the advent of computers in the 1950s, the information economy started to influence all the other economies and became an economic force in itself. A range of emerging economies are with us today. This includes the bio-economy, the nano-economy, the hydrogen economy and the neuro-economy. At the confluence of all these economic waves, we have the knowledge economy, which is supplemented with the algorithm economy. This latter economy is termed as such since it will be dominated by the mainstreaming of artificial intelligence. This will reinforce some of the historical economies, in particular the industrial economy, leading to the 4th industrial revolution (after steam, electricity and IT influences that represent the first three revolutions). AI will force us to look at economic growth in different ways. Intelligent automation supersedes traditional automation that just does repeat work very accurately, fast and efficiently. Complex tasks that require adaptability and agility to solve problems or come up with better solutions need intelligence. As birth rates slow and people age, fewer people will be available for the workforce. The workforce of the future will have to be augmented by intelligent machines. A new division of labour is essential. AI will not replace people, but work alongside them and free then to do what will add most value. For a long time, technology provided tools to augment human skill, be that physical or mental. Now AI can augment natural human intelligence by freeing the mind from tedious routine tasks. Capital efficiency can be enhanced by AI. Reduction of down-time, preventative maintenance, instrumentation health monitoring, productivity enhancement, dynamic systems management, workflow improvement and logistics enhancement all relate to better productivity of capital.

 

AI and developing economies

In developing economies, job creation is part of national development plans. Eradication of poverty through creating a healthy employment market is key to economic growth. How will AI impact these economies as globalisation spreads over rich and poor? The Bank of England is quoted (Arbess, 2016) to estimate that 48% of human workers will eventually be replaced by robotics and software automation. Up to 76 million jobs may disappear in the next two decades in the US alone. This is more than 10 times the jobs created in the Obama era. AI will affect low skilled and highly skilled people. The Chinese economic transition through the stages of industrialisation, urbanisation and finally consumption-driven growth, followed by exports, is typical of successful emerging economies. Arbess states that China is now the world’s largest economy after the US. They still have over 100 million people working in manufacturing jobs with only 36 robots for every 10 000 workers, versus only 12 million factory workers in the US, which has a robot penetration rate of 4.5 times China’s, 164 per 10 000 human workers. The question is paramount for development: How will displaced workers in emerging economies be affected? If they work shorter hours, how will they earn enough money to live? The financial rewards from the AI economy will primarily go to investors, entrepreneurs and shareholders. This may lead to destabilising social tension as the divide between rich and poor widens. In consumption-driven economics, this will have a severe effect on the fiscus and economic and political stability. To counter this dilemma, AI will bring higher productivity, less waste, improved logistics, no labour unrest and strikes, with the result of cost and price reduction, and customisation that prevents wasteful surplus of mass production. New opportunity for a different type of start-up that complement AI driven manufacturing may take up some of the job losses. But this is usually for skilled people. Where does the unskilled workforce go to, to earn money? First of all, AI diffusion is seen to be gradual and not a radical step change (Heath). Human-machine partnerships as a new way of working will be with us for a long time. It was pointed out that AI seems to find things that are simple to humans difficult, and difficult things simple (Moravec's Paradox). Manual roles, albeit low paid, will remain in demand, with the exception that workers will have to learn to do it alongside machines. Human creativity will still lead to the creation of art, music and content. The connected intelligent software era will mean that these can be distributed instantaneously to vast consumer numbers globally. Interpersonal skills will be required for a long time and will be expanded to inter-machine relationships and human-machine relationships. Motivating, comforting, caring, supporting, advising, guiding and rewarding, all excellent human capabilities, will be part of the workplace of the future. Heath states: “A more optimistic outcome than automation leading to mass unemployment is to see these technologies as a tool that will allow people to achieve more”. The role of upskilling and reskilling is clear. Education becomes critical. Having a global outlook is crucial, not getting stuck in cultural isolation. Activists for opportunity are needed more than proponents for resistance. Considering the way AI systems work today, it is difficult to conceive of a system that can replace a person's entire set of skills (Jennings quoted in Heath). AI software is normally very deep and very narrow; manual and professional jobs or tasks generally require a broader base of expertise and experience than AI is able to codify. The solution lies thus not in “against”, but “together”. There is a clear trend in history that technical change destroys some jobs, but creates others. It may not be without upheaval, as was the case in the transitions of the previous industrial revolutions. But through an adaptive education system, not scared of revolutionising itself, normality will be attained again after transitioning fully into the 4th industrial revolution. Creating work in developing economies that adopt AI may not lie in the ambit of industrial and economic development, but in national structures for education and social development. The ability of humans to adapt may be the breakthrough to forging the human-machine frontier. It is not going to be about more of the same old talk, but changing the conversation that will make developing economies thrive in AI.

 

Machine innovation?

So often one reads about the “innovation machine”, meaning how organisations have perfected innovation and drive their future competitiveness through innovative approaches and products. But will machines be able to innovate? As digitalisation and machine intelligence are rapidly emerging towards disruptive status, the question should be asked whether humans will or should remain in control of innovation? The 4th industrial revolution; the Internet of Things; smart everything, from wearables to cities; and artificial intelligence are future waves that will change forever the world of work, play, living and transaction, as we know it today. We tend to think that future “smartness” is hidden in the creative abilities that humans have to innovate. Innovation is dependent on the ability to recognise patterns, combine and integrate existing properties and knowledge to represent something that is perceived as new by the user, seeing the gap in the market and intuition. This is exactly what AI and intelligent machines do. David Autor is quoted in (Purdy and Daugherty, 2016): “Often people only think of AI boosting growth by substituting humans, but actually huge value is going to come from the new goods, services and innovations AI will enable”.

 

Are AI markets real?

In a recent market overview (Venture Scanner, 2016), 13 market sectors were identified, with a total of about 1 500 companies operating in them across 73 countries, with a total of $8.5 Billion invested. These market sectors are (quoted from the source):

Deep Learning/Machine Learning (Platforms): Companies that build computer algorithms that operate based on their learnings from existing data. Examples include predictive data models and software platforms that analyse behavioural data.

Deep Learning/Machine Learning (Applications): Companies that utilise computer algorithms that operate based on existing data in vertically specific use cases. Examples include using machine learning technology to detect banking fraud or to identify the top retail leads.

Natural Language Processing: Companies that build algorithms that process human language input and convert it into understandable representations. Examples include automated narrative generation and mining text into data.

Speech Recognition: Companies that process sound clips of human speech, identify the exact words, and derive meaning from them. Examples include software that detects voice commands and translates them into actionable data.

Computer Vision/Image Recognition (Platforms): Companies that build technology that process and analyse images to derive information and recognise objects from them. Examples include visual search platforms and image tagging APIs for developers.

Computer Vision/Image Recognition (Applications): Companies that utilise technology that process images in vertically specific use cases. Examples include software that recognises faces or enables one to search for a retail item by taking a picture.

Gesture Control: Companies that enable one to interact and communicate with computers through their gestures. Examples include software that enables one to control video game avatars through body motion, or to operate computers and television through hand gestures alone.

Virtual Assistants: Software agents that perform everyday tasks and services for an individual based on feedback and commands. Examples include customer service agents on websites and personal assistant apps that help one with managing calendar events, etc.

Smart Robots: Robots that can learn from their experience and act autonomously based on the conditions of their environment. Examples include home robots that could react to people’s emotions in their interactions and retail robots that help customers find items in stores.

Personalised Recommendation Engines: Software that predicts the preferences and interests of users for items such as movies or restaurants, and delivers personalised recommendations to them. Examples include music recommendation apps and restaurant recommendation websites that deliver their recommendations based on one’s past selections.

Context Aware Computing: Software that automatically becomes aware of its environment and its context of use, such as location, orientation, lighting, and adapts its behaviour accordingly. Examples include apps that light up when detecting darkness in the environment.

Speech to Speech Translation: Software which recognises and translates human speech in one language into another language automatically and instantly. Examples include software that translates video chats and webinars into multiple languages automatically and in real-time.

Video Automatic Content Recognition: Software that compares a sampling of video content with a source content file to identify the content through its unique characteristics. Examples include software that detects copyrighted material in user-uploaded videos by comparing them against copyrighted material.

The most used AI enterprise solutions include in order of high to low (Allidina, 2016): voice recognition and response, machine learning, virtual personal assistants, decision support, automated communications response, analytics, and robotics

 

Preparing for the future of AI

A report (Holdren, et al, 2016) issued in October 2016 by the Executive Office of the President of the USA, prepared by the National Science Technology Council, Committee on Technology, Subcommittee on Machine Learning and Artificial Intelligence suggests how to prepare for a future with AI. It states: “In recent years, machines have surpassed humans in the performance of certain specific tasks, such as some aspects of image recognition. Experts forecast that rapid progress in the field of specialized artificial intelligence will continue. Although it is very unlikely that machines will exhibit broadly-applicable intelligence comparable to or exceeding that of humans in the next 20 years, it is to be expected that machines will reach and exceed human performance on more and more tasks.” It further remarks: “In the coming years, AI will continue to contribute to economic growth and will be a valuable tool for improving the world, as long as industry, civil society, and government work together to develop the positive aspects of the technology, manage its risks and challenges, and ensure that everyone has the opportunity to help in building an AI-enhanced society and to participate in its benefits”.

 

What to expect next?

What AI based environments will be with us first? To a large extent they are here already, but keep an eye on (Ayers, 2016): automated transportation (drones and self-driving cars and self steering ships); cyborg technology (for disablement support as in prostheses, bionic implants or exoskeletons to enhance the physical strength of the human body); taking over dangerous jobs (defusing bombs, cleaning up hazardous spills, deep mining, etc.); solving climate change (big data and data analytics); having robots as work partners, friends, pets and companions (watch the Japanese lead with this); improved health care and supervision (for the aged, the robo-nanny, intelligent ICUs, etc.); and, of course, the workforce of the 4th industrial revolution. The next game changers in AI will include aspects of context awareness, improved dexterousness, and computers behaving more and more like humans (Russel, 2016). Everything good in life is the result of human intelligence. Adding AI could amplify this “goodness”. Brain-machine interfaces for the disabled, multiplication of the short term memory of humans and protecting humans from virtual crime and cyberwar may be noble outcomes of advanced AI. In many cases today, AI can do the difficult things easier than the easy things. Control algorithms are for, example, not good enough yet for simulating the human hand in robots. This is because of the complexity of the anatomy of the human hand. There is hope that 3D printing could open new opportunities for more complex robot systems that are more human-like in terms of mechanical motion. In general, AI developers adopt a philosophy of technology for humaneness. This ethic should be transferred to the making of super-intelligent machines. It is crucial that AI must be built by all demographic groups across the world and that the population of the planet must be represented. Young people will be the drivers and adopters of AI and create the culture of co-existence of humans and machines.

 

Let’s make it work

AI is a typical emerging technology spilling over the edge of disruption. Many of us associate disruption with discomfort and radical change and try to avoid it, but disruption has positive impacts as well, first mainly for the disruptor, but eventually also for the disrupted. What can we do to embrace AI and all the advantages it can bring humanity? Purdy and Daugherty (2016) suggest we prepare the next generation for the AI future. This may be easier than thought, since this is the generation that will not have the legacies of human-technology strife, but grew up with technology as an enabler for human action since babyhood. AI regulation will be a necessity. Laws will have to be revised to include machines that do tasks and take responsibility. All those lawyers that may be out of a job from AI incorporation into he legal practice may have a good creative challenge to come up with the new legislation that will deal with human rights and machine rights alongside each other and will have to set boundaries for machine ethics and values. For example, if machines become super-intelligent and adopt, through simulation, value systems, would they start demanding remuneration at some time in future? Can a machine be fired from its position? These may seem tongue-in-the-cheek for now, but are these notions so unrealistic? It will be necessary to advocate a code of ethics for AI as intelligent machines are rapidly moving into the previously exclusive human social domain. There are three broad areas of ethical and social concern that have to be addressed as pointed out by Lin et al (2011): safety and errors; liability laws; and social impact. Autonomous action by machines (e.g. driverless vehicles), will have to know when to give preference to life saving action or goal oriented task execution. Standards and best-practices in the development of AI will have to give recognition for such ethical codes. It will further be necessary to understand and manage the knock-on effect of AI. It is often not in the immediate vicinity of AI application that the largest effects will be felt, but in remote, higher order environments. Data accumulated by autonomous vehicles, for example, may have the largest effect on the insurance industry and not the automotive industry. Policy makers will have to embrace the benefits that AI can bring, but at the same time be sensitive to pre-empt the dramatic and potentially devastating effect of misusing AI. The essence is, though, that no national development plan, economic strategy or industry roadmap can afford to be without he consideration of AI anymore. Like the French tradition to taste the first young wine of the season in early November, announced by the slogan “le beaujolais nouveau est arrive” , we should all say with unbridled enthusiasm “AI has arrived!”

 

 

Dr Anthon Botha is Managing Director of TechnoScene (Pty) Ltd, a Future Thinking consultancy that operates in the realms of technology management, innovation management and knowledge management. He is also part time academic at the Graduate School of Technology Management at the University of Pretoria in South Africa. He can be contacted at This email address is being protected from spambots. You need JavaScript enabled to view it.. For more information on TechnoScene, visit www.technoscene.co.za

Read more about Anthon and his view on being a futurist

 

References

 

Share content with FFD

Foresight For Development - Funding for this uniquely African foresight site was generously provided by Rockefeller Foundation. Email Us | Creative Commons Deed | Terms of Conditions

Top Desktop version