Government policy for regulating artificial intelligence technology

Policy Drafting Exercise
Table of Contents
Introduction 1
The Direction of AI 2
Benefits 2
Issues 3
Security, Legal and Ethical Issues 3
Policy Needed 6
Why the Government Should Establish Policy 7
Basic Principles 8
Conclusion 9
References 10



Introduction
The rise of automation has led to the expansion of numerous industries getting involved in automated machines. Uber, Lyft and Google have all dabbled in automated automotive engineering, with self-driving cars predicted to be the future (Duhigg, 2018; Kosoff, 2018). The rise of artificial intelligence has also change the way people think about managing their own lives. Google Assistant, Amazon’s Alexa, and Apple’s Siri are all listening on all devices, waiting for their owner to mention their name so that they can respond (Navarro, 2018; Pesce, 2018). The technology is not perfect yet, however. Forsberg (2003) has noted that speech recognition software can be faulty and that has led to some people having their entire lives recorded by their devices without their even realizing it (Chokshi, 2018). With Amazon’s Echo, one is supposed to say, “Hey, Echo,” and the device will activate—but as some consumers have learned, the device may be activated unwittingly and it may even end up sending private information to people who should not have it (Chokshi, 2018). This has obviously led to security and privacy concerns as well, which is what this policy paper intends to deal with.
The Direction of AI
Artificial Intelligence (AI) has allowed for some significant achievements in terms of the direction of technology in recent years. Self-driving cars are now being manufactured, and devices that connect people’s homes and machines have been produced and sold. More and more workings seem to be leaving the hands of individuals and being put into the control of machines, run by software engineered to have AI capabilities. But what is AI? AI is really nothing more than algorithmic code that is programmed to “learn” from various input sources—but this learning is not comparable to human learning or understanding. While some codes may be programmed to conduct certain tasks at alarmingly fast rates, they are highly specialized and can only perform a narrow range of functions. AI is nowhere near as comprehensive as human intelligence, and to expect it to be as good or as efficient is to expect far too much at this point in time. That means more concern and awareness has to be given to how devices that use AI are being entrusted to the public. Are they safe? Is privacy protected? Are there threats to being hacked? Can it really be trusted?
Benefits
The benefits of AI algorithmic decision-making are that the machines can work quickly to help people perform functions that used to take longer. For instance, in market trading, algorithms have been written to help traders read the news headlines, interpret messages on public social media profiles and scan the Internet for all relevant data before making a trade—and all this performed within the blink of an eye (Yip, 2018). This kind of AI usage obviously has its benefits for traders who are looking to profit off information before human users have time to process information and put on a trade. However, it has its limitations. Algorithms can cause flash crashes in the market because they fail to interpret moves or information correctly and they all pile into trades that make no logical sense and that human traders can see and profit from. Thus, the traders or funds that were using all AI software lose while humans win. However, there are other benefits—such as the autonomous car that uses AI and allows the user to get from point A to point B without any hassle.
Issues
Issues are that nothing is perfect and anything designed by man will have flaws and limitations. This is true of AI as well: the autonomous car has not yet been perfected, as the numerous reports of Tesla vehicles failing to recognize a change in driving conditions and caused a wreck continue to pour in. There is also the problem of devices that use algorithms to recognize voice inputs picking up and sharing information that was never intended to be recorded or shared. These problems are legitimate and have to be considered because if too much reliance on AI is used across a variety of industries, it could lead to ugly ramifications for society. Privacy could be breached, lives could be threatened, and infrastructures could be made vulnerable to enemy attack or to internal weaknesses in the software itself.
Security, Legal and Ethical Issues
Security is a number one issue for devices that use AI and that connect via the Internet of things. If all devices are linked together, there is significant risk of one device being hacked somehow and all devices being breached. This means that whatever personal data is stored on one platform is likely to be stolen. Even something as simple and basic as cloud computing is not without its risks. AI technology has been promoted as the next big thing in technology, yet it is a very unproven technology that features a great deal of risks to people’s privacy and to people’s lives. Their personal information could be at risk if AI algorithms fail to understand what is required of them; information could be collected and distributed in a manner that is not proper. Even students who are learning at school sometimes make mistakes, so it should not be a surprise to find that AI-driven machines also will make mistakes. But those mistakes should not be allowed in markets where so much can go wrong. Cars driven by AI can crash and kill. Devices controlled by AI can be breeched or can breech themselves.
One of the most problematic characteristics of the digital cyber space today is that the space is essentially considered a battlefield by intelligence communities and they themselves are trying to collect as much data as possible. The gathering of Big Data has made AI a necessary function in government and intelligence communities have to have a way of interpreting information using AI—but the technology itself may not be as safe or as effective as governments think. Today, cyber warfare is a reality. Taking advantage of other nations’ faults through their weak AI can be disastrous. Because so many nations’ technological infrastructure is now dependent upon cyber technology, acts of war can be committed in cyber space and, without launching single missile, a nation can be taken offline and off the grid. But that is not all: The problem of too much AI may be more precarious than anyone yet has imagined: it is believed that “by the year 2045 the quantity of nonbiological intelligence on the planet will substantially exceed that of the entire human” (Shanahan, 2015, 157).
As more and more data is extrapolated, more and more tests run, more and more knowledge gained, more and more dollars are invested into the realm of technology, including the realm of the AI space, or into the arena of cyber technology that uses AI, or into the realm of meta-cognition, where machines are expected to learn on their own, the world becomes a more complicated place and a less stable one—because machines can be broken, they can be hacked, they can be destroyed, and people who rely upon them to much can lose everything. Moreover, Nance and Ryan (2011) maintain that “research into digital evidence is likely to become increasingly challenging as infrastructure as a service (IaaS) and software as a service (SaaS) through the cloud become increasingly common” (p. 5). As everything transitions to AI and to the Internet of Things, the problems can increase: too much dependency upon one type of technology makes the entire infrastructure more vulnerable. This is not just a problem for people and privacy but rather a problem for entire networks, companies, businesses and governments: the entire way of life, where everything has become digitized, could be threatened.
Ethical considerations have to be made especially in times when the law fails to keep apace with the changing times of the technological world. Even from a simple utilitarian ethical framework, the problem of not knowing wherein lies the greater good makes it problematic for the field of consumer products that rely on AI as a feature to allow homes to become smart. But what if foreign entities can hack into homes and steal information or eavesdrop on people? What if one’s own government begins to do as much?
The digital age is rapidly changing the way the world works today, which is problematic for the laws which were enacted to govern the world of yesterday. Today’s laws must be updated to address the needs of concerns of people in cyberspace. Businesses, people’s data, privacy, and information are all at risk. Legislation needs to catch up to where the digital age has taken it. The digital world has altered so many things about the modern world already—and there is still much more likely to be changed. Legislation therefore has to reflect the legal questions that have arisen.
Policy Needed
Policy is needed that will address these security, legal and ethical considerations before too much dependency on AI—an unproven technology—occurs in the nation. While industries have been tasked with monitoring themselves because of the demands of the market, their ability to self-regulated has come into serious question in recent months and years. Tesla, for instance, is driven by a CEO whose actions have routinely raised eyebrows. He is hailed as a visionary but in practical terms he has failed as an efficient manager of the company. He has promised to redefine the way engineering is achieved but he has not succeeded in creating the AI-driven robotics to build cars quickly and efficiently that he promised. Tesla’s manufacturing subsystem is a highly complex system involving AI, robotics, and trained employees both in the manufacturing plant and the Gigafactory where Tesla’s batteries are produced. The need for change within this subsystem is based on the fact that Tesla has not reach its production goals, as the factory is too reliant on AI and robotics, which were envisioned by Tesla CEO Elon Musk as being the future of auto manufacturing. Musk recently admitted that this overreliance on robotics was an error (Matousek, 2018). Moreover, the automobiles built by Tesla that use AI to read roads and engage the auto-pilot have not shown themselves to be fully efficient. Yet no one is really regulating this industry. Every crash results in an investigation, but as of yet nothing has come of these investigations.
This signals that more strident policy is required and that the government should be the one to get involved in this process. Policy should focus on overseeing the development and testing of AI products just as new drugs are overseen and tested before they hit markets. Of course, the downside to this is that testing can sometimes take years, which means that technology that was expected now could be pushed back a decade, which could in turn make competitors in other parts of the world all the more eager to jump in and gain market share in markets where companies here sought to dominate prior to the introduction of oversight.
Why the Government Should Establish Policy
The government is responsible for protecting the lives of the community. The government also has the systems in place to oversee and regulate industries. Companies that use AI to enhance people’s lives, whether through the Echo, Siri, other Amazon devices, Google, Tesla or any other machines, should be subject to government oversight so that safety, security and privacy can be guaranteed. Currently there is no guarantee that anyone is safe or that privacy laws are not being broken. Part of the problem, however, is that government itself may be the problem. The NSA whistleblower Edward Snowden showed that the government is essentially spying on people by hacking into their devices and reading their messages. Thus, there is a need not only for oversight of these industries but for the government to regulate itself.
In spite of the NSA revelations, it is clear that the government should be involved in regulating industries that use AI because there is currently no regulation or oversight whatsoever, and people have seen numerous negative effects—from flash crashes in the markets prompted by AI-driven algorithms to literal crashes of cars driven by AI-fueled steering to privacy breeches by AI-operated devices that pick up voices and learn how to interpret them based on sound waves. None of these AI-driven approaches in any of these industries have been perfected or shown to be entirely safe, yet they are sold to people as though they were perfectly sound and without side effects. The reality has shown otherwise.
Basic Principles
Basic principles that should be followed include the use of oversight committees for each industry in which AI is used in the creation of some software, some product, or so method that will enable or enhance human life to be lived more efficiently. These committees should be tasked with applying a utilitarian ethics principle to the problem of whether the AI is suitable for consumption. Beyond whether the AI actually works as advertised, these committees should look at the greater good or at whether the products actually would help society to achieve a greater good or whether they would actually just put society at greater risk.
Utilitarian ethics is concerned with maximizing welfare and using that position to justify whatever course of action is adopted. For various industries it would have to be shown that their AI products actually can maximize welfare rather than reduce it. Cars that use AI for driving autonomously would have to make the case to these committees that their product is actually beneficial to society and that there is zero risk of the product malfunctioning or failing to respond appropriately on the road. Device makers that listen to people and respond to requests based on voice activation would have to show that such devices are worth the risks of an over-dependency upon them being cultivated. The problem of too much technology and too much reliance on AI is that it can be addictive and habit-forming, just like too much dependency on opioids to relive pain. In the end, they can cause problems for everyone who takes them and lead to more suffering than was intended as the drugs were meant to relieve pain—not cause it.
If products that use AI are actually going to cause more trouble than good, then the committees should not authorize them to come to market. The industries that use AI should be regulated in the same way the pharmaceutical industry is regulated, from a utilitarian ethical perspective. The government should have this power, as it is the best positioned to provide this service. The only drawback is that it is unclear whether the government is actually capable of policing itself. If industries are incapable of regulating themselves, it stands to reason that government will face this same problem. At the end of the day, there has to be an end to oversight.
All the same, companies should have to apply for permission to test products and to have those tests examined by the committees just like companies in the health industry have to have their products tested. AI can be just as powerful and deadly and disastrous as drug products; therefore, regulation and oversight should be just the same. Without such oversight, the risks that attend to the advent of AI in society are not going to be addressed, because the industries that use AI to sell products are in the business of using the next big thing to make money.
Conclusion
This policy paper has shown how important it is for industries that use AI to promote their products be regulated by government. It has highlighted some of the problems of this policy, particularly the problem of government to regulate itself, but it has also shown that there is no alternative because regulation has to start and stop somewhere. The paper has highlighted the reasons for this needed—particularly that AI products are not 100% safe or effective and that numerous defects have been found across all industries. The need to protect people, to protect their privacy, to protect their personal information, and to protect their lives all shows that government has to take an active role in overseeing the use of AI across all industries.



References
Chokshi, N. (2018). Is Alexa Listening? Amazon Echo Sent Out Recording of Couple’s
Conversation. Retrieved from https://www.nytimes.com/2018/05/25/business/amazon-alexa-conversation-shared-echo.html
Duhigg, C. (2018). Did Uber steal Google’s IP? Retrieved from
https://www.newyorker.com/magazine/2018/10/22/did-uber-steal-googles-intellectual-property
Forsberg, M. (2003). Why is speech recognition difficult. Chalmers University of
Technology. Chalmers University of Technology.
Kosoff, M. (2018). Uber is losing the self-driving car war. Is it too late to catch up?
Retrieved from https://www.vanityfair.com/news/2018/12/uber-races-to-get-its-self-driving-cars-back-on-the-road
Matousek, M. (2018). Elon Musk says he agrees that there are too many robots on the
Model 3 production line. Retrieved from https://www.businessinsider.com/elon-musk-says-model-3-production-using-to-many-robots-2018-4
Nance, K., & Ryan, D. J. (2011, January). Legal aspects of digital forensics: a research
agenda. In 2011 44th Hawaii International Conference on System Sciences (pp. 1-6). IEEE.
Navarro, F. (2018). You’re not paranoid, your phone really is listening to you. Retrieved
from https://www.komando.com/happening-now/464613/youre-not-paranoid-your-phone-really-is-listening-to-everything-you-say
Pesce, M. (2018). Voice assistants are always listening. Retrieved from
https://www.theregister.co.uk/2018/02/26/voice_assistants_are_always_listening_so_why_wont_they_call_police_if_they_hear_a_crime/
Shanahan, M. (2015). The technological singularity. MIT Press.
Yip, J. (2018). Algorithmic Trading using Sentiment Analysis on News Articles.
Retrieved from https://towardsdatascience.com/https-towardsdatascience-com-algorithmic-trading-using-sentiment-analysis-on-news-articles-83db77966704