Source: Pexels
We often come across headlines of how AI is revolutionizing industries. Certainly, its impact is everywhere to be seen and can not be disregarded as such — be it in the form of enhancing healthcare outcomes, elevating customer experiences, optimizing supply chains (something I covered in detail in one of my previous articles) to opening up new business stream altogether.
AI is indeed a transformative force.
However, with its benefits come the risks and challenges too. Ethical concerns such as privacy, bias, accountability, and cybersecurity have raised severe concerns since the AI industrialization has picked up traction.
In this article, we will learn how to navigate the fine balance building AI regulation while simultaneously fostering innovation.
The Big Responsibility That Comes With AI Regulation
In order to ensure that the benefits of AI continue to serve society and humanity at large, it is important to regulate AI — at the right time with the right guardrails.
Now, this is easier said than done. It puts a big responsibility on policymakers, businesses, and technologists.
Wondering what is so “big” about this? Here is the thing:
Firstly, ethics is an abstract and esoteric concept. There is no one person, group, team, organization or authority that can comprehensively cover the ethical landscape and set the precedence for everyone.
Secondly, the rapid rate at which the technology is developed often makes it challenging for policymakers to fully understand new technologies and establish effective safeguards.
Lastly, regulatory frameworks must be developed ensuring a careful balance between fostering innovation while safeguarding individuals and society.
The Need for AI Regulation
With the increasingly pervasive nature of AI, its impact on society is growing too which further raises concerns, such as:
Data Privacy
Before we even discuss issues that are exclusive to AI systems, let’s quickly touch upon the foundational issue that has been concerning society for a long time.
You guessed it right: it’s data issues, primarily data privacy — or should I say, the breach of data privacy.
Data regulation like that of the General Data Protection Regulation (GDPR) has set the precedence in terms of handling users’ data responsibly, thereby raising the bar for consent and transparency.
Algorithmic Bias
Now, comes the part where AI algorithms have the potential to reflect (in worse-case scenarios even amplify) the bias stemming from the data it is trained on.
Historical data may be biased leading to the biased outcomes from the model. Its impact has been seen in areas like recruitment, credit-lending, healthcare and more. It is critically important to ensure we build fair and non-discriminatory AI systems, which is unfortunately an ongoing challenge for both developers and regulators.
One rule of thumb that can help build such fair systems involves embedding ethics by design, which means that if may not be fair by default, but we can make those design choices to build responsible AI. It involves asking the right questions at the ideation phase itself, which could be as simple as: Who is missing from this training dataset? Who have we not considered while training the model but would be served with the predictions when deployed?
Accountability
AI systems, especially deep learning ones, are often “black boxes,” which makes it challenging to understand the reasoning behind their decision-making processes. This lack of transparency leads to obscurity around accountability when AI systems cause harm or make erroneous decisions. Building transparency in AI systems is key to ensuring that AI systems are aptly understood and can be audited, when need be.
Safety Risks
With good comes the other side too. Yes, AI systems have been exposed to adversarial attacks, where bad actors can manipulate them into producing harmful and unintended outputs. Furthermore, deploying AI in certain areas, such as autonomous vehicles, surveillance, military applications, and other sensitive areas, raises concerns about its potential misuse.
In nutshell, regulations bring the much-needed safety net to deploy AI in a way that is aligned with societal values.
The Balancing Act of Innovation with Protection
Now that we clearly understand the need for AI regulation, let’s quickly cover the challenges behind putting them into practice, starting with fear of inhibiting innovation.
AI has become synonymous to “greatest innovation of recent times” (or even the greatest innovation of all time), which means those who are able to leverage it effectively will be able to progress faster than others.
AI has also been assumed to be a symbol of wielding power, for it is seen as a driver of economic growth and technological advancement. Now, with this context, it is clear that regulations are perceived to be restricting innovation creating barriers to entry for new ideas and startups.
It is important to note that regulations imply additional measures for compliance that has associated costs, which is a major factor that leads to discouraging investment in AI research and development.
So, what is the right thing to do? It requires policymakers to maintain a balance between setting necessary safeguards while allowing room for experimentation and growth.
One such effort in this direction is adopting a risk-based regulatory framework, similar to the EU AI Act. Its underlying principle ensures a risk-based grading that puts necessary guardrails and emphasis on high-risk AI systems.
It considers the fact that there is no “one-size-fits-all” approach. And that, one blanket rule does more harm than good. Therefore, regulators are tasked with the challenge of considering the specific contexts in which AI is deployed while designing regulatory framework.
Now that we have discussed contextualizing, there is another dimension to consider — geographical boundaries. AI is a global technology, which means it transcends boundaries. So, multiple regulatory approaches across countries cannot just create confusion, but also disturb international collaboration. Global cooperation is key to develop common AI principles, frameworks, standards and guidelines for AI regulation.
I hope that, by now, it has become clear how challenging it is to build regulations. One underlying principle that can help address all these challenges is to adopt an adaptive approach to AI regulation. It is difficult to get it all right at once, plus it runs the risk of quickly becoming outdated as the technology evolves. Therefore, continuous monitoring and iteratively update the regulations is key to building robust safeguards around one of the most innovative technologies of all time.
Vidhi Chugh is an AI strategist and a digital transformation leader working at the intersection of product, sciences, and engineering to build scalable machine learning systems. She is an award-winning innovation leader, an author, and an international speaker. She is on a mission to democratize machine learning and break the jargon for everyone to be a part of this transformation.