"If we don't change the direction soon, we'll end up where we're going." Irwin Corey
AI technologies are forecast to add USD 15 trillion to the global economy by 2030. At this moment, around the world, at least 50 countries (including the EU as a whole) have developed, or are in the process of developing, a national AI strategy. Of these, 36 have (or plan to have) either separate strategies in place for public sector AI, or a dedicated focus embedded within a broader strategy.
This means it is quite safe to state that artificial intelligence is far from being a novelty for most individuals around the globe. It also shows that the new technology is already impacting, to a significant extent actually, the way we act as humans in general, and as users of digital products in particular.
AI-powered tools are integrated into various businesses from different areas, ranging from retail and education, to the medical system, cybersecurity and even the defense sector. With so many advancements and considering the effort and investments that the governments put into adopting and further developing AI, it is only normal that we ask ourselves: should this be regulated? Is there a standard we should have in mind when building AI products? Should there be laws, government accountability and corporate transparency?
What is regulation and why would AI need to be regulated?
Regulation of businesses also existed in the ancient Egyptian, Indian, Greek, and Roman civilizations. Modern regulation, however, can be traced back to the 19th and 20th century, when much of the ruling in the US was handled and enforced by regulatory agencies which produced their own administrative laws and procedures under the authority of statutes.
New technologies and groundbreaking discoveries or innovations have always called for new regulations or amendments to the existing ones. And, as with any other change, opinions are divided when it comes to the benefits of regulation. Some believe it hinders progress and is an impediment to corporate and small-business profits and a waste of resources. Others see it as a great way to protect individuals or businesses and to address the potential unintended consequences of disruption. The truth is that regulation should be able to maintain a balance between doing these while also fostering innovation.
With regard to AI, there are a lot of voices calling for such a regulation that would prevent governments and any other businesses from using the technology to their advantage. For instance, Bill Gates, who believes AI will “allow us to produce a lot more goods and services with less labor,” expects labor force dislocations and has suggested a robot tax. Elon Musk, who is well-known for his warnings about AI and its threat to the existence of human civilization, calls for proactive regulation “before it is too late.”
But why would we need to regulate AI? Why not let the technology follow its course and turn into this unprecedented, amazing tool?
1. AI can be both good and evil
Artificial intelligence has indeed proven it can bring unexpected benefits to numerous sectors. From less boring and time consuming jobs, decreased costs, improved quality and performance, better decision-making processes, to more accurate and faster medical diagnostics as well as early detection of cancer and other chronic diseases, AI turns out to be a great ally in improving our lives.
But it is not all rainbows and sunshine. AI can also be biased, manipulative and inconsistent. There were numerous cases in which AI was used for targeted political adverts - remember Facebook and Cambridge Analytica who served up manipulative fake news stories to people based on their psychological profiles.
2. Machines make mistakes too
And this is not actually the real problem. The problem is that, most of the times, we tend to believe that an algorithm is always right, because it's all mathematics and logic. Well, this is not even remotely true. In her book "Hello World", Hannah Fry presents a series on interesting events in which algorithms were misleading, but humans decided to trust them blindly. For instance, on March 22nd 2009, a man almost drove off a cliff because his GPS told him it had found a shortcut to his house. Although the road was narrow and full of dirt, the man stated he had no reason not to trust his TomTom sat-nav. In other shocking examples, she presents how some algorithms used in the judicial system have allowed judges to wrongfully convict several citizens.
3. Ethics is a big conundrum
Ethics in artificial intelligence has been long debated. Sadly, we still have no final decision or course of action on this.
A few years ago, five technology giants have joined forces to create an ethics standard for AI. Researchers from Alphabet, Amazon, Facebook, IBM and Microsoft have met to discuss the most tangible issues, such as the impact of AI on employment, transport and even wars. Furthermore, Google employees pledged to not use AI for military applications.
Yet, people are still wandering whether or not AI will be used to power warfare or how we should address the subject of driverless cars.
What is currently being done?
Experts, researchers, governments and other AI enthusiasts are becoming aware of the fact that the future development of machine learning, NLP, deep learning etc, could easily get out of control.
We are already starting to see different groups that want to impose certain regulation on how this technology is used. The main purpose of this is to make sure AI will benefit people, not cause them harm - voluntarily or not.
In 2015, a thousand experts, including physicist Stephen Hawking, the co-founder of Apple, Steve Wozniak, Elon Musk, co-creator of PayPal, linguist Noam Chomsky, and Demis Hassabis, chief executive of Google’s AI company, signed a petition that was drawing attention to the dangers of artificial intelligence and demanded it be regulated.
Furthermore, President Trump’s executive order, “Maintaining American Leadership in Artificial Intelligence,” calls for a coordinated approach to regulation of AI. The strategy is based on five principles:
· Driving technological breakthroughs
· Developing appropriate technical standards
· Training a workforce for the future
· Fostering public trust and confidence
· Ensuring international trade of AI-enabled products
Although all these initiatives are not enough to regulate the field of AI, they represent a promising start. It is clear artificial intelligence needs regulation and that almost everyone seems to be on board with this idea. The question that still remains now is - can AI be regulated?
The Stanford project, "A study into one hundred years of Artificial Intelligence" is a report published by a group of experts from Stanford University led by Eric Horvitz, a Microsoft researcher. The authors claim it will be impossible to regulate AI.
“The consensus of the study agrees that attempts to regulate AI in general would be an error, since there is no clear definition of what AI is (it’s not just one thing), and the risks and considerations to bear in mind are very different for the different domains”. The Stanford project, "A study into one hundred years of Artificial Intelligence"
Talking about arms control, Henry Kissinger, former US secretary of state, believes it may be a lot harder to control the development of AI weapons than nuclear ones.
Indeed, AI will be difficult to regulate as a whole, because it is too diverse and has its own way of developing and pushing through the known boundaries. However, this does not mean that we should stop trying. AI is the technology of the future so we need to understand and manage it if we want to have a harmonious human-machine relation.
At this point, no one holds the correct answer related to regulation of artificial intelligence, but we'll continue our part in making everyone aware of the economic, social and humanitarian implications of AI.
In the end, we leave you with an interesting report about the AI readiness level of governments around the world.