Robots - Employees of the Future

From New Media Business Blog

Jump to: navigation, search

Traditionally, when we hear of Artificial Intelligence (AI), movies like the Minority Report, the sensationalist idea of robots taking over the world, and maybe its gaming applications come to mind.

With the help of Hollywood, mainstream media, and research, AI has gained tremendous popularity over the years, allowing people to better understand what it truly is, inclusive of its benefits, applications, and limitations. Quite simply, as a branch of Computer Sciences, AI is meant to emulate what is arguably the most complex and powerful object in the world: The Human Brain.

Contents

Artificial AI Overview

AI is the evolution of an ongoing study that attempts to understand how a “computer mind” can be created and developed such that it mimics the way a human thinks [1]. Just as humans learn by receiving information, through studying and experiences, AI technology has the ability to “learn” from the data it collects. The more data that is available to collect and analyze through carefully crafted algorithms, the better it can become at making predictions.

Advancements in AI have further paved the way for intelligent agents that are capable of "perceiving their environments and taking actions to maximize the chances of success" [2]. This includes using algorithms within business applications to determine how best to serve customers, analyze data and provide diagnoses to effectively treat patients, and even incorporating robots into manufacturing workflows to automate single-task processes[3].

Extent of AI

Artificial Intelligence is Born

The idea of machines possessing intelligence was first introduced in 1950 by the British mathematician Alan Turing, who created the Turing Test. The test involved an examiner sitting behind a divider and engaging in written exchanges with hidden entities, either a human or computer. If the examiner failed to distinguish between the responses received, the computer would successfully pass the test, deeming machines to be intelligent[4]. Although recent times have seen various machines come close, most critics agree that the test is yet to have been passed[5][6].

After the Turing Test, it wasn’t until 1955 when the term “artificial intelligence” was coined by American computer scientist John McCarthy at the 1956 Dartmouth Conference. He described AI as "every aspect of learning or any other feature of intelligence [that] can in principle be so precisely described that a machine can be made to simulate it”[7]. This pivotal moment in history is what eventually kickstarted developments into the field of AI.

Neural Networks and the Brief Rise of "Robots"

In the late 50s and early 60s, the research community began to try and understand what could be done to make McCarthy’s vision a reality. Through researching pattern recognition and working theories of the brain at the time, the idea of artificial neural networks was conceptualized. The concept of artificial neural networks (ANN) were inspired by what was known about how the human nervous system responded to information stimulus [8]. Using algorithms, these early ANNs were designed to recognize patterns by accepting inputs of data, organizing and classifying them through layers, after which further processing would produce an output[9].

The easiest example can be demonstrated by understanding a Tic-Tac-Toe algorithm. By following a preprogrammed algorithm, the AI player reads the choice of the opposing player, either X or O, analyzes all possible options through its classification layers, and responds with the best output that it thinks will beat the initial input[10].

This discovery saw the rise of some of the worlds first robots, namely Unimate, an industrial robot that replaced humans on the assembly line at General Motors; Eliza, a chatbot that could hold conversations with humans; and Shakey, the world’s first "general-purpose" mobile robot. This understandably exciting time created a feeding frenzy throughout industries, with companies placing a new emphasis into research and development in hopes of creating the next big thing.

Unfortunately, this excitement was short-lived, due to repeated failures, which eventually led to a substantial reduction in funding [11]. Reasons for this included limited computational power, inefficient systems to process the available data, and being too costly[12]. As such, further developments into ANNs stagnated and thus came the period, known today as the AI winter, which lasted until the early 90s.

Developments in Machine Learning

In the early 1990s, AI reignited the attention of onlookers, due to the synthesis of Computer Science and Statistics, allowing for models to incorporate uncertainty parameters when analyzing the data[13]. This method still initially required the use of algorithms being fed preprogrammed instructions[14]. However, the breakthrough allowed scientists to build intelligent systems that could analyze, learn, and even make predictions from large amounts of inputted data [15].

IBM's Deep Blue

Back in 1959, Arthur Samuel pioneered the idea of machine learning[16], defining the phrase as “the ability to learn without being explicitly programmed.”[17]. Less than half a decade later, AI was slowly starting to do just that.

IBM's Deep Blue

On May 11, 1997, IBM’s Deep Blue, a chess-playing computer, defeated chess grandmaster, Garry Kasparov, making history as the first machine to ever beat a human expert of the board game[18]. A year prior saw the IBM computer face-off against Kasparov, in which Kasparov won[19]. During this time, Deep Blue’s designers worked to improve the computer’s capabilities, through practice games, fixing errors, and increasing its ability to analyze more options[20]. As a result, Deep Blue succeeded by using algorithms to analyze approximately 200 million possible chess positions per second, followed by choosing the strongest move to best its opponent[21].

Although Deep Blue utilized certain machine learning approaches, it still heavily relied on preprogrammed data, which became a limiting factor[22]. As such, although the computer was excellent at playing chess, its use cases could not be expanded to other applications, or even other games [23].

The evolution of Deep Learning

Deep Learning with Neural Networks

The advancements in Deep Blue and machine learning, albeit limited in scale during the 90s, resulted in significant contributions to the field of AI. The onset of the 21st century saw vast advancements in science and technology, which benefited AI tremendously[24]. As computers and hardware became more powerful and capable of handling more data, considerable research into neural networks reconvened as well[25].

Eventually, it was found that neural networks could be layered on top of one another and connected through algorithms, such that data passing through would allow for the system to learn and be trained at unprecedented levels[26]. This deeper, more evolved version of machine learning created the core foundation of what is known today as deep learning.

As deep neural networks have multiple layers, this allows for processing different features of analyzed information. For example, in the case of a self-driving car, whereas one layer would detect the edges of a road for steering, another may detect the lane lines to keep the vehicle straight[27].

Google's Alpha Go

Lee Sedol vs. AlphaGo

Similar in concept to IBM’s Deep Blue, Google’s AlphaGo was a computer program trained to play the game, Go. Go is a very complex ancient Chinese board game, with nearly infinite possible moves. This poses a problem for traditional computing in AI, as standard machines aren't capable of determining all of the possible combinations. In the case of Deep Blue and chess, where the computer used brute force to determine the best options to succeed, this method would not work in Go[28].

Yet, AlphaGo managed to win, twice, against the legendary Go champion, Lee Sedol. As such, it has been argued that AlphaGo exhibited the very signs of intuition than one would find in human thinking, in order to win[29]. Also, AlphaGo was not pre-programmed, but rather used a concept known as reinforcement learning to accomplish the feat. Google experts trained the computer by showing it games played by renowned Go players, after which it played itself 30 million times, reducing errors and improving its odds of winning after each generation. Moreover, rather than requiring structure or defined rules, AlphaGo learns on its own, through experience[30].

Types and Limitations of AI

Computers like AlphaGo signify how far AI has come, especially in the last 20 years. Having the ability to seemingly think for itself is historic, especially without the need for much human input[31]. Nevertheless, AlphaGo and all other AI applications currently in existence have their limitations.

Narrow AI vs. General AI

Narrow AI vs. General AI

Narrow AI

Narrow AI, also referred to as weak AI is what all AI applications are currently characterized as. Although many examples depict machines seemingly superseding the intelligence of humans, this has only been accomplished for a particular task[32]. While Deep Blue and AlphaGo outperformed their counterparts within their respective games, these same computers would not be able to apply themselves past the games of chess or Go, on their own. Moreover, machines may have reached a point where they can process data faster than humans, but their ability to understand different contexts, think abstractly, or plan and solve general problems is limited.[33]

General AI

General AI, is the theory that AI systems will reach a point whereby they can reason for themselves within an environment, and perform a variety of unfamiliar intellectual tasks, to produce a solution, like a human would. This would involve a machine being able to do the following[34]:

  1. Learn autonomously, in real time, and from single examples
  2. Understand language, engage in meaningful conversations, and reasoning "contextually, logically, and abstractly"
  3. The ability to recall recent events
  4. Use of existing knowledge and skills
  5. Understand how to manage "multiple... conflicting goals"
  6. Have emotional intelligence

MarI/O: Machine Learning in Video Games

MarI/O - Machine Learning in Video Games















Applications

I Am AI: GTC 2018 Kickoff


AI Applications

Facebook

What is DeepFace? 'Human-Level' Face Matching, Explained

Facebook uses a sophisticated facial recognition system that employs a nine-layer deep neural network to identify faces within images posted on Facebook. Facebook developed this system and it has proven to be 97% accurate [1] (which exceeds the FBI’s 85% accuracy rate [2]) and began its roll out in early 2015. Due to privacy laws in the EU, this system is not employed in most of Europe. In 2017, a feature was implemented that would notify you if your face was posted in an image, even if you were not tagged[3]. Additionally, Facebook has submitted a patent for facial recognition payments[4].

finn.ai

Conversational AI is a result of a variety of different technological systems, such as Natural Language Processing (NLP), working together to understand and interpret natural language. finn.ai delivers conversational AI for banking, allowing users to ask questions to an AI that would normally be reserved to human employees. The financial assistant[5] system can be rolled out across a variety of platforms and languages[6].

Google.ai

Google has a been an industry leader in the field of Artificial Intelligence (AI) and Machine Learning (ML) implementations for many years. The firm provides open-source resources to developers to learn more about the field, such as access to a machine learning kit for mobile developers[7], a neural network example that produces psychedelic images[8], an open source JavaScript library for training machine learning models[9], and many more platforms for developers and students to learn and practice their AI and ML skills[10].

Google Maps & News

AI has been used by Google to redesign both Maps and News[11], two of their oldest apps. The addition and improvement of self-learning algorithms help provide users with more relevant and recent information. For many years, Google has employed these services to blur faces and license plates on Google Maps[12]. In 2018, the platform will receive an augmented reality addition that will employ the use of the camera, street view maps, and AI algorithms to help you navigate[13]. Google News now employs machine learning and AI to sort and categorize news headlines in meaningful ways to provide readers with Storylines and the opportunity to read about a single topic from a variety of sources in a simple and efficient manner[14].

Netflix

How Netflix Implements Big Data Is All about You

Most commonly known, Netflix employs AI to recommend what to watch[1]. The firm has been using this sort of system in one form or another for a number of years. Similar systems are used by YouTube[2] and Spotify[3]. More recently, Netflix has begun to use AI to improve the viewing experience for when internet is slow[4]. Different scenes require different qualities of stream – cartoon require far less than a vibrant fight scene. The Dynamic Optimizer algorithm identifies the quality required by a scene and buffers accordingly, resulting in better image quality and stream reliability[5].

PayPal

PayPal is an international online payment platform employs the use of AI to bolster security across the globe[6]. With millions of transactions occurring a day, the firm could not possible have every transactions be meaningfully analyzed by a security specialist. Instead, a system is used to process transactions in real-time, bringing potentially fraudulent activity to the attention of a security specialist. When fraudulent activity is confirmed, it is added to the system’s database of fraudulent activity and used in the future to spot similar purchases[7]. In May of 2018, PayPal purchased Jetlore, a California start-up that focuses on AI in retail[8].

Self-driving Vehicles

The state of self-driving cars: 2018

AI is used in various self-driving vehicle applications. Drive.ai is a third-party, unaffiliated tech startup that strives to provide the solution to autonomous driving. Ford and Dominoes have teamed up to attempt to combine pizza delivery with autonomous driving[1]. Cadillac uses an implementation to dramatically enhance the standard cruise control function through their Super Cruise[2]. The technology behind the self-driving vehicles is LIDAR, a surveying method that collects information about the surrounding environment, which is then interpreted by the AI of the onboard computer.

Sunspring

Sunspring

Through the use of a long short term memory recurrent neural network, a group of computer engineers taught AI to write a film script. The script was turned into a 9-minute short film starring Thomas Middleditch, the star of the popular sitcom Silicon Valley. The AI wrote the film after analyzing a library of ‘80s and ‘90s sci-fi screenplays[1]. The short film has received mixed reviews[2].

Robot Applications

Boston Dynamics' SpotMini

SpotMini Autonomous Navigation

The SpotMini is a “nimble robot that handles objects, climbs stairs, and will operate in offices, homes and outdoors"[1]. It is essentially a small, fully electric dog-like robot that can be used for a wide range of commercial purposes. In the example video, SpotMini was allowed the opportunity to learn its path prior to departing, so we can see that the technology is not fully autonomous yet[2]. Commercial production has not yet begun but is intended for sale in 2019. We have yet to see what the commercial implementations of such a robot can be.

Robots in Hospitality

Inside the Japanese Hotel Staffed by Robots

A Japanese hotel, the Henn na Hotel, employs robots to replace a majority of its workforce. A robot clerk handles check-ins and check-outs, there is a robot cloak room, concierge and a personal assistant in every room. Alternatively, the Hilton Hotels are piloting the use of a robot concierge, Connie, in some of the hotels[1]. A variety of other hotels are working on robot implementations[2].

While robots are believed to be eliminating millions of jobs in the coming years[3], the quantity of jobs created to support the robot industry will significantly offset this job loss. Hotels using the Relay Robot will have to hire additional staff to support the robot usage. As well, as robots are employed to increase hotel efficiency, cost realizations can lead to lower prices and an increase in occupancy rates, which will have additional job creation effects. Furthermore, the vast amount of data collected by robots will result in additional job creation relating to big data management and interpretation.

Robots in the Home

Sony's new Aibo is a very good robot dog

The International Federation of Robotics has reported that the industry for service robots will exceed $5B in 2017 and exceed $11B in revenue by 2020[1]. Robots have a wide variety of applications in the home, such as cleaning[2], entertainment[3], personal assistance[4] and security[5].

Robots in Medicine

A combination of artificial intelligence and robotics will be employed in the future to aid medical professionals and to alleviate the discomfort of the disabled. AI and Machine Learning is being applied to 5 hospitals in America to help reduce costs and increase the quality of healthcare[6]. Matia Robotics has created the Tek RMD, a robotic assistance device that increases mobility for those bound to a wheelchair[7]. A combination of robotics and AI can be used to centralize aids for disabled individuals, allowing them to communicative and coordinate care at a higher level than previously capable[8].

Softbank Robots

Softbank currently offers a number of different robots for various purposes. This past summer, Pepper the Robot was demonstrated at SFU to introduce 24 high school students to the world of computer science and artificial intelligence[9]. Connie the Robot, an implementation of Softbank’s Nao, can be found greeting hotel guests at various Hilton hotels[10]. Other variations of the Nao robot can be found in some Japanese banks[11].

Tesla Manufacturing

National Geographic Megafactories - Tesla

Tesla prides itself on nearly fully automated manufacturing of its vehicles. Without training, the machinery used in the manufacturing process would be useless. Each piece of machinery had to be programmed by hand. In total, the Fremont, California factory (that currently produces the bulk of Tesla’s vehicles) contains over 160 fully autonomous machines used in manufacturing[1]. Tesla is seeking to use AI to support its manufacturing revolution in the future[2].

Effects of AI on Wages

For people in the labour market, a secondary concern to job replacement, is changes to wages. As AI changes the entire job landscape, wages will likely not remain the same. As automation increases, either partial or complete, it has the potential to increase economic inequality. Those in highly skilled jobs that are less affected by use of robots, machines, and AI will most likely see an increase in wages, and those who are in middle paying jobs that are more affected by the increased use of automation could see wage stagnation [3]. The International Monetary Fund also has reported that introducing robots and AI into the workforce will be bad for equality and projects that skilled wages will increase while low-skilled wages will decrease. Furthermore, in general those who have access to AI and robotic technology and education will largely be better off than those who don't. Even the best case scenario is not good for equality in regards to wages or societal status. Though there are ways to try to prevent the worst effects as mentioned later in the wiki page, however, it is hard to prevent all ill effects [4] [5].


On a more positive note, the jobs that AI creates will largely see higher wages due to the increase in skill and education that the positions will require. An example on the higher-end sees AI experts earning over $300,000 per year. This can pose a threat to how democratized this technology can be, as not all companies or organizations can afford to pay their employees such high wages, which largely means top talent will be working in private companies which usually are able t pay more [6]. However, when looking at AI jobs in a broader sense (not just AI experts or researchers), AI related jobs have a more moderate salary, of around $86,000 per year in Canada. Again, there is quite a large range based on the position, with positions such as programmers making considerably less than higher up positions [7].

Responses to AI

Regulation

Global Regulation

Regulation of AI is currently difficult for lawmakers and regulatory bodies to agree upon. With technology evolving and changing so quickly, but without a widespread adoption of AI in workplaces, regulation of AI is stuck in a difficult place. There is the need for preventative measures but it is at a stage where knowing what preventative measures to take will actually be effective. Most view the need for global regulations in order to effectively regulate [8]. Most regulation thus far has been focused on human rights as the main consideration and compliance with international humanitarian law. Additionally, it sees both users and developers of AI as holding responsibility for using discretion to use AI ethically [9]. Other existing regulation isn’t specific to AI, but does have an affect on AI because it regulates data protection, privacy, trade practices, and health and safety [10].


One set of proposed universal rules includes “The Rules for AI” which are an adapted version of the famous Rules for Robots and have largely focused on three main tenants that AI should be following:

  1. “An A.I. system must be subject to the full gamut of laws that apply to its human operator
  2. An A.I. system must clearly disclose that it is not human
  3. An A.I. system cannot retain or disclose confidential information without explicit approval from the source of that information" - (https://dzone.com/articles/regulating-artificial-intelligence-definitely-mayb).

An organization that is attempting to bring together people from around the world with knowledge and expertise of AI to do research in promotion of AI regulation and safety is OpenAI. This research in turn will help to guide policymakers and creation of regulation. The organizations has a strong focus on keeping the humanity and advocating for AI regulation that would be positive for humanity [11]. Beyond conducting and publishing their own AI research, they also have open-source software tools that allow others to contribute to AI research as well [12].


Formal global organizations also have started to address the need for AI regulation. In June’s G7 meeting AI was discussed and an agreement that the G7 nations would work together to create a vision for the future of AI was made [13]. Some of the main focuses are: research and development, education and training, ensuring equal opportunities for all people to become involved in the industry, ensuring a wide variety of stakeholders are involved in discussing the future of AI and policy regarding AI, create policies to boost labour market needs for workers in AI, support AI security and related security, privacy and data protection issues[14].

Government Regulation

Asia

Singapore

Singapore has been touted as the leader in AI thus far because of their Research Programme on the Governance of Artificial Intelligence and Data Use. This programme focuses on:

  1. "… promot[ing] cutting edge thinking and practices
  2. inform[ing] AI and data policy regulation
  3. establish[ing] Singapore as a global thought leader in AI and data policies regulations” -(https://theaseanpost.com/article/regulating-artificial-intelligence-adoption-southeast-asia-0)

Additionally, Singapore has created a Smart Nation initiative which sees a move into digital and smart technologies, of which AI could potentially be a part of. Already, the Road Traffic Act has been amended to recognize self-driving vehicles in preparation for self-driving vehicles to be commonly used which would support their use of public transport (a part of the Smart Nation initiative). Singapore has expressed interest in autonomous vehicle technology and has supported testing within Singapore. Various government bodies have been created to monitor the progress of autonomous vehicles such as the Committee on Autonomous Road Transport for Singapore, Intelligent Transportation Society Singapore, and Singapore Autonomous Vehicle Initiative [15].

Another area Singapore sees the use of AI is in their financial market. The Monetary Authority of Singapore will be working with agencies to research and develop AI to be used in finance [16]. The Monetary Authority has provided a $27 million grant to invest in AI and data analytics and is working with universities to grow a workforce for financial AI [17]. They are also trying to bring AI technology into use within the government run financial area. To test out new products and services, the Monetary Authority of Singapore has created a FinTech Regulatory Sandbox. They can test products and services in the sandbox to see which Singapore’s financial sector should use[18].

China

Of course, China is also getting in on AI, and the State Council has issues “A Next Generation Artificial Intelligence Development Plan”. This plan has set three dates to reach their goals in stages with the end goal set for 2030, being the global leader in AI. To be the global leader in AI, China will be focusing on creating an AI industry that will be used in areas such as health care, agriculture, manufacturing, and more. They also have set goals to introduce laws and regulations to ensure a proper implementation of AI into Chinese society [19].


Beyond this, China has also enacted a new Personal Information Protection Standard to be used by regulators who perform audits on Chinese companies. Protecting people’s personal information will have an effect on information collected by AI, giving more rights to the people to protect their own data and more responsibility to companies for safeguarding the information that people do give them [20].


China is also moving to use machine learning to help develop foreign policy. They have been clear that though the government is hopeful to use AI to help create policies, humans will still be responsible for finalizing any policies. AI can simply help speed up the research into different policy options. One of the possible pros is that it would be subjective, however there can also be the argument that morality and emotions often are needed when creating policies and laws that can have a great affect on people. It is expected that the use of AI for foreign policy would put China’s strategy ahead of other countries who don't have the technology but it is projected that other countries will start to utilize AI in their foreign policy decisions too [21].

North America

Canada

Thus far, Canada has little-to-no governance for AI. The main laws that regulate the use of AI is the Charter of Rights and Freedoms and the Personal Information Protection and Electronic Documents Act. These basic protections for citizens do not provide enough guidance for use of AI within Canada. Experts on AI believe there needs to be a more action to create regulation and policies. There is a large movement into the research and development of AI by the government but there needs to be a concurrent move to implement laws and regulation [22]. Whatever policies are introduced must make sure the future of AI is inclusive of all Canadians and has proper collaboration between all areas that AI could be affecting within Canadian governance [23] [24].


To bring together these different areas, Canada has a Chief Information Officer (CIO) Strategy Council thats objective is to create the discussion among all stakeholders in an effort to create one common idea of what AI ethics should look like in Canada and in future policies [25]. The CIO Strategy Council includes different areas of government as members, such as Canadian National Defence, Health Canada, and Treasury Board of Canada Secretariat [26]. They have a specific Standards Policy Committee with the objective of creating standardization for new technologies including AI [27]. Other steps Canada is taking towards regulation and policies involves Global Affairs Canada, who is working with universities to research AI and human rights, and the Treasury Board Secretariat of Canada opening the discussion up online to get thoughts and opinions on ethical and fair use of AI, and of course Canada’s involvement in international coordination through the G7. The main areas Canada is focusing on is security, changes to the labour market, and keeping AI a competitive market [28].

United States

Much like Canada, the US is being slow to enact any formal policies or regulation for AI. There is a “Select Committee on Artificial Intelligence” which is under the National Science and Technology Council. However, this committee has many vacant seats for positions of key importance. While these positions remain vacant, the President, Donald Trump is effectively going to be advising himself on AI. Ultimately, this means that the committee is ineffective at providing any real suggestions to the administration regarding AI [29].


There is however, a group of legislators who are hoping to enact real change within the US. A Future of AI Act has been created with the intention of being presented to Congress. A main component of the proposed Act is to institute an advisory committee for the Department of Commerce that will be made of experts in the field that will then recommend how the US should proceed and what strategy is needed for AI for the US to be a leader in AI[30].


Despite these efforts to create a more homogenized AI plan for the US, there are some who call for limited government intervention when it comes to AI. Some of the AI companies working out of Silicon Valley have voiced their preference of limited policy and regulation around AI research and development. This is likely because it could give them an advantage when competing against other countries developing AI that are subject to more regulation. It is suggested that a balance of being given support from the government while maintaining a free market approach is preferable[31].

Europe

Europe has made a lot of headlines for their General Data Protection Regulations (GDPR). GDPR has caused concern for AI advancement because of the amount of personal information AI often collects and uses to be effective AI. It has been predicted that GDPR will increase costs of using AI, and reduce accuracy because of the more limited data collection and use. However, GDPR isn’t all bad and has provided a set standard of regulations that have a broad effect as all EU member nations must comply which is a step towards having international standards [32].


GDPR does not directly address AI, but the European Commission did set out a communication titled, “Artificial Intelligence for Europe” which gives a plan of what the EU’s response to AI will be. The communication lists investing in research and development of AI, providing opportunities for use of AI to smaller companies, promoting education and training for AI jobs, creation of a European AI Alliance to execute the goals and strategies listed in the communication, and sets out to review existing related regulations and make changes if necessary, as well as create AI ethics regulations [33].

Germany

Germany became a leader in Europe around AI especially when looking at autonomous cars. Housing car manufacturers giants such as BMW and Mercedes-Benz, there was the need to develop rules around autonomous car development and use [34]. German government passed guidelines for self driving cars with a huge emphasis on human life being the number one priority [35].


More legislation that Germany has passed affecting their AI use and development is the Network Enforcement Act. This requires all online platforms to remove hate speech. This affects AI because social media companies such as Facebook and Twitter use automated systems to flag content that goes against their website’s terms of use, but will now be under stricter rules to remove hate speech content quickly or face fines of up to 50 million euros [36]. This act may bolster the need for a more culturally aware AI to be developed that understands different languages and context of words based on the language and culture. Current AI is not smart enough to entirely be responsible for content removal, and so social media companies have hired thousands more personnel for content monitoring and removal [37].

France

Emmanuel Macron, France’s current president, has been a strong advocate for move into AI. Under his administration, the French government is allocating 1.5 billion euros towards their AI strategy. Macron has shown a great personal interest in AI and belief that France can only benefit from being open to AI and executing initiatives to support the move into AI [38].


France is looking to be the top European country in AI and has put out a comprehensive strategy to challenge other European countries. The main focuses of the strategy involve policies that support AI and safe data collection and privacy protection, investment into AI research and training programs to create workers in AI, support AI that helps the environment, ensuring AI is developed and used ethically in the long term, and that AI will benefit all people [39].

Within the scope of AI use within government functions, France believes they can utilize AI better than other countries in areas such as healthcare as they currently have robust data collected under the responsibility of the government already. The move to AI using this existing infrastructure would mean a smoother transition into government use of AI with controls already in place to protect sensitive and private data [40].

Universal Basic Income & Social Security

Define

Basic Income is a form of government assistance that involves "a periodic cash payment unconditionally delivered to all on an individual basis, without means-test or work requirement[41]". For example, a full basic income program would provide citizens with an amount of income that would reasonably allow a citizen to meet their basic needs. A partial system would be an amount less than this. Basic income is seen to be a far simpler and transparent system that traditional welfare systems[42]. Many believe basic income to be the solution to the automation woes that our society will likely face in the coming decades [43][44][45].

Examples

Finland is a major proponent for universal basic income[46]. The country’s study began in January of 2017 with 2,000 participants between the ages of 25 and 58 who were previously receiving unemployment benefits. The participants were randomly selected. The purpose of the study, which concludes in December of 2018, is to study the effects of basic income on labour participation[47].

The Ontario provincial government began a three year, 4,000 participant pilot of universal basic income in the spring of 2017[48]. Participants were between the ages of 18 and 64 and were already receiving unemployment benefits. While the amount received is not contingent on whether or not the participant is looking for work, it is reduced by 50% of the earned income of an individual. This program was cancelled in July of 2018 [49].

Funding

There have been far more discussions regarding the potential impact of basic income than the actual feasibility. Many proponents believe that cost savings in the elimination of a traditional welfare program (and the associated bureaucracy and administrative costs) will pay for a large portion of the expense. Additionally, there is belief that a basic income will reduce federal medical costs associated with diseases of poverty[50]. In 2017, The Guardian posted an Editorial that suggested taxing big-data giants that monetize citizen information in order to pay for part of a basic income program [51].

Societal Responses to AI

Public Opinion

AI and robots have been a prominent piece in popular media for the nearly past century[52]. The word robot was introduced to the English language in 1921 from a science fiction play Rossumovi Univerzální Roboti (Rossum's Universal Robots). As previously discussed, human responses to robots will have long-lasting future policy implications.

AI can have positive social implications too. For example, DeepMind, a division within Google, has been used to reduce Google’s Data Center cooling bill by 40%, reducing electricity usage[53]. Visabot is an AI chatbot that helps foreigners understand the process behind applying for US immigration.

As far as society is concerned, AI will both harm and benefit us [54].

Expert Opinion

Astro Teller, Google’s Artificial Intelligence Chief, believes that society should not fear AI, as it will change the lives of millions of people, such as by providing a safe ride to a job or a doctor’s appointment for those who can’t drive themselves[55]. Elon Musk, on the other hand, believes that AI could be the downfall of our civilization if it is not treated appropriately[56]. Various other experts have just as conflicting opinions[57].

Ethics

As the extent of the digital age of AI and robotics-enabled automation becomes more apparent, so does the need for ethical guidelines to moderate such technology. With AI slowly starting to take over decisions normally made by humans, new layers of ethical complexities will begin to exist[58]. Similar to how we have ethical initiatives in place for food, aviation, and traffic, the thought is that similar institutions must be put in place for AI.

An issue of particular concern is the idea of enforcing accountability for the actions of a machine. Unlike humans, who can provide explanations for why they engaged in a particular decision, a machine’s decision-making capability is directly linked to complex algorithms, pathways, and networks. Not only is it difficult to determine which pathway was utilized to arrive at a decision, AI is also typically incapable of explaining itself or its internal processes in a manner easily understood. As the integration of AI into society is growing, generating methods to reduce this informational gap is necessary[59].

A potential solution proposed by Maaike Harbers, a research professor at Rotterdam University in the Netherlands is to limit the autonomy of machines, thereby allowing humans to exercise control within three critical stages: "data input, processing and reasoning, and the output or action" [60]. However, critics argue that such regulation can stunt the growth of AI, proposing the need to instead teach these systems "human ethics and values"[61]. It is evident that these views are at odds with one another; whereas the former is fearful of the advent of AI, the latter sees hope in the machines of tomorrow.

Challenges facing AI expansion and adoption

Automation Readiness Index

Automation Index Map

Today, two schools of thought exist on how automation will affect economies and workforces. Some believe that the incoming wave of technology will be similar to industrial revolutions of the past and that the inclusion of AI, machine learning, and advanced robotics will naturally rebalance the job market. More controversial then, is the idea that this emerging wave of automation is distinguished from ones previous, possibly displacing current workforces, given that AI and robotics are “starting to automate higher order, non-routine tasks”[62].

If we are to believe the latter opinion, it becomes evident that countries must become proactive in developing strategies and policies that allow for effective transitioning into an automated future. This includes creating policies to help individuals and businesses take advantage of the new opportunities automation will provide while supporting workforces that will have jobs displaced due to AI[63].

Providing context to this problem is the Automation Readiness Index, which compares countries based on their level of preparedness for the new age of automation. Countries are examined on policies and strategies that exist pertaining to areas of "innovation, education, and the labour market"[64].

Countries near the top include South Korea, Germany, and Singapore, due to how they have embraced education and curriculum reform, occupational training, and flexibility. Moreover, these countries have also increased funding for organizations to further research AI and robotics implications.

The report concludes that countries must make stronger attempts to implement policies that can handle the impending challenges automation will pose. This requires input from multiple-stakeholders, inclusive of governments, businesses, educators, labour unions, and civil society organizations. This currently poses a problem, due to low engagement levels between these groups. A lack of cohesion between all systems can otherwise result in governments developing and implementing policies separately, which can be troublesome as the approach is bound to alienate some groups.

Quality Assurance of AI

Bias and Discrimination

If AI worked as perfectly as the developers would like, AI would eliminate bias and discrimination. AI has been offered as a bias and discrimination free way to evaluate candidates for jobs. Although AI is currently helpful in eliminating some basic biases, algorithms are not perfect yet. The data that is used has a huge impact on the results being biased or not. The data must be clean and accurate data that in itself is free from bias. Additionally, the algorithm must not always follow previous decisions through machine learning, it must also learn from previous bias mistakes so that future decisions have less bias and do not continue on a set pattern. Additionally, using algorithms may reduce diversity that is an important factor to bringing in new ideas and views [65].


There are a number of companies trying to tackle the bias and discrimination problem. Pymetrics advertises themselves as being a bias-free way of identifying candidates for jobs. They use neuroscience games to determine behaviours and personalities of a candidate, and machine learning and AI to match candidates to roles. Using different algorithms for each company and updating algorithms continuously through their algorithm auditing process, they claim to bias-free selection [66]. More established companies, such as Accenture, are also trying to solve this issue. Accenture has introduced the AI Fairness Tool largely to combat racial, sex, and age biases which are common biases. AI Fairness Tool is supposed to run multiple tests to recognize even the slightest bias patterns that can show up that will then be removed from the algorithm. Accenture will be implementing this tool with some of the organizations and companies they work with [67].


The important thing to keep in mind for companies who are using AI in some capacity to help remove bias, is that whatever tool they use should be extensively tested, audited, and ideally developed by a diverse team to mitigate bias and discrimination within their AI[68].

Projected Growth and Trends in AI

Future Job Landscape

The rate of jobs requiring AI skill has grown dramatically in the recent years [69] These jobs are exciting, high paying and provide lucrative benefits [70]. Machine learning, deep learning, and language processing are becoming high-demand skills, as most of the jobs created are related to the creation and development of AI technologies [71]. The rapid expansion of the robot market will continue to both eliminate and create jobs in a variety of fields [72].

Authors

Graham McLean Rebeka Thompson Rahim Rajani
Beedie School of Business
Simon Fraser University
Burnaby, BC, Canada
Beedie School of Business
Simon Fraser University
Burnaby, BC, Canada
Beedie School of Business
Simon Fraser University
Burnaby, BC, Canada
grmclean@sfu.ca rebekat@sfu.ca rrajani@sfu.ca

References

  1. https://www.youtube.com/watch?v=8_lfxPI5ObM
  2. https://medium.com/protopiablog/tesla-wants-to-use-ai-to-create-a-manufacturing-revolution-11bf6b5b9d8
  3. https://www.mckinsey.com/featured-insights/future-of-organizations-and-work/ai-automation-and-the-future-of-work-ten-things-to-solve-for
  4. http://www.abc.net.au/news/2018-05-23/we-should-fear-robots-says-imf/9786460
  5. https://www.artificialintelligence-news.com/2018/01/11/ai-wealth-inequality/
  6. https://www.nytimes.com/2018/04/19/technology/artificial-intelligence-salaries-openai.html?smid=tw-share
  7. https://neuvoo.ca/salary/artificial-intelligence/
  8. http://www.aies-conference.com/wp-content/papers/main/AIES_2018_paper_13.pdf
  9. https://www.swp-berlin.org/fileadmin/contents/products/comments/2018C23_dkw_job.pdf
  10. http://thehill.com/opinion/technology/375606-should-the-government-regulate-artificial-intelligence-it-already-is
  11. https://blog.openai.com/openai-charter/
  12. https://openai.com/about/
  13. https://pm.gc.ca/eng/news/2018/06/10/prime-minister-concludes-successful-g7-summit-focused-creating-economic-growth
  14. https://g7.gc.ca/en/official-documents/charlevoix-common-vision-future-artificial-intelligence
  15. https://aitech.law/publications/av-singapore
  16. https://www.finews.asia/finance/26851-mas-pushes-ai-ecosystem
  17. http://fintechnews.sg/19617/fintech/ai-finance-ecosystem-in-singapore-mas/
  18. http://www.mas.gov.sg/Singapore-Financial-Centre/Smart-Financial-Centre/FinTech-Regulatory-Sandbox.aspx
  19. https://www.natlawreview.com/article/china-s-vision-next-generation-artificial-intelligence
  20. https://www.insideprivacy.com/international/china/china-issues-new-personal-information-protection-standard/
  21. https://www.scmp.com/news/china/society/article/2157223/artificial-intelligence-immune-fear-or-favour-helping-make-chinas
  22. https://www.cigionline.org/articles/policy-deficit-behind-canadian-artificial-intelligence
  23. https://www.cigionline.org/articles/policy-deficit-behind-canadian-artificial-intelligence
  24. http://snip.ly/6wilr#http://policyoptions.irpp.org/magazines/april-2018/toward-a-national-policy-on-ai/
  25. https://www.itworldcanada.com/article/holding-ai-accountable-public-leaders-organize-to-ensure-algorithms-influencing-government-are-ethical/404793
  26. https://ciostrategycouncil.com/about-us/
  27. https://ciostrategycouncil.com/standards/
  28. https://www.globalpolicyjournal.com/blog/09/06/2018/regulating-robots-canadas-potential-leading-way-responsible-ai
  29. https://thenextweb.com/artificial-intelligence/2018/05/14/the-white-houses-new-ai-panel-is-a-joke-trump-is-its-top-adviser/
  30. https://www.wired.com/story/the-us-needs-an-ai-strategy/
  31. https://venturebeat.com/2018/05/11/silicon-valley-to-trump-write-checks-and-a-gameplan-for-ai-but-not-regulations/
  32. https://www.roboticsbusinessreview.com/ai/will-european-privacy-rules-hinder-ai-development/
  33. https://www.covcompetition.com/2018/05/covington-artificial-intelligence-update-european-commission-publishes-communication-on-artificial-intelligence-for-europe/
  34. http://www.futurecar.com/1599/German-Automakers-Suppliers-are-Leading-Self-Driving-Car-Patents
  35. https://www.roboticsbusinessreview.com/unmanned/germany-creates-ethics-rules-autonomous-vehicles/
  36. https://www.theguardian.com/world/2018/jan/05/tough-new-german-law-puts-tech-firms-and-free-speech-in-spotlight
  37. https://www.digitaltrends.com/social-media/network-enforcement-act-germany/
  38. https://www.wired.com/story/emmanuel-macron-talks-to-wired-about-frances-ai-strategy/
  39. https://analyticsindiamag.com/breakdown-of-frances-ai-strategy/
  40. https://www.politico.eu/article/macron-aims-to-drag-france-into-the-age-of-artificial-intelligence/
  41. https://basicincome.org/basic-income/
  42. https://www.degruyter.com/view/j/bis.2008.3.1/bis.2008.3.1.1106/bis.2008.3.1.1106.xml?form=
  43. https://futurism.com/images/universal-basic-income-answer-automation/
  44. https://www.cnbc.com/2018/02/20/richard-branson-a-i-will-make-universal-basic-income-necessary.html
  45. https://qz.com/911968/bill-gates-the-robot-that-takes-your-job-should-pay-taxes/
  46. https://www.kela.fi/web/en/basic-income-experiment-2017-2018
  47. https://www.youtube.com/watch?v=mkF-Lsy-SlM
  48. https://www.ontario.ca/page/ontario-basic-income-pilot
  49. https://www.huffingtonpost.ca/2018/07/31/ontario-basic-income-pilot-cancelled-by-minister-lisa-macleod_a_23493366/
  50. https://www.youtube.com/watch?v=cQyr9l22fLE
  51. https://www.theguardian.com/commentisfree/2017/sep/15/the-guardian-view-on-universal-basic-income-tax-data-giants-to-pay-for-it
  52. http://www.vulture.com/2015/03/15-best-robot-movies-of-all-time.html
  53. https://www.forbes.com/sites/fridapolli/2018/01/30/ai-for-social-impact/#15a23a43343d
  54. https://www.forbes.com/sites/forbestechcouncil/2018/03/01/14-ways-ai-will-benefit-or-harm-society/#7f06fbcd4ef0
  55. https://www.popularmechanics.com/technology/robots/news/a28645/googles-alphabet-astro-teller-ai/
  56. https://www.vanityfair.com/news/2017/03/elon-musk-billion-dollar-crusade-to-stop-ai-space-x
  57. https://www.beencrypted.com/future-artificial-intelligence-expert-opinions/
  58. https://diginomica.com/2018/01/29/big-question-responsible-ethical-ai/
  59. https://www.washingtonpost.com/outlook/ai-is-more-powerful-than-ever-how-do-we-hold-it-accountable/2018/03/20/e867b98a-2705-11e8-bc72-077aa4dab9ef_story.html?noredirect=on&utm_term=.96e58f6a1087
  60. https://m.phys.org/news/2018-07-digital-age-desperately-ethical-legal.html
  61. https://theconversation.com/does-regulating-artificial-intelligence-save-humanity-or-just-stifle-innovation-85718
  62. http://www.automationreadiness.eiu.com
  63. https://www.businesswire.com/news/home/20180423005873/en/ABB-Economist-Launch-Automation-Readiness-Index-Global
  64. http://www.automationreadiness.eiu.com/static/download/PDF.pdf
  65. https://www.cnbc.com/2018/05/30/silicon-valley-is-stumped-even-a-i-cannot-remove-bias-from-hiring.html
  66. https://www.pymetrics.com/science/
  67. https://business.financialpost.com/technology/ai-might-be-biased-against-you-but-accenture-wants-to-give-business-a-tool-to-fix-that
  68. https://www.businessnewsdaily.com/10956-stop-hiring-bias-artificial-intelligence.html
  69. https://www.forbes.com/sites/louiscolumbus/2018/01/12/10-charts-that-will-change-your-perspective-on-artificial-intelligences-growth/
  70. https://www.techrepublic.com/article/the-6-most-in-demand-ai-jobs-and-how-to-get-them/
  71. https://www.inc.com/jessica-stillman/21-future-jobs-robots-are-actually-creating.html
  72. https://www.computerweekly.com/news/450431755/Robots-will-create-more-jobs-than-they-eradicate-by-2020


Personal tools