Race to the Bottom on 'Safety' Affecting 'Responsible AI Development'

AI Ethics should not be treated as an aftermath, rather organizations must prioritize the incorporation of AI ethics at the very outset of the design process of AI systems to ensure that these systems work as intended and are developed in a responsible manner. It also helps organizations to protect themselves from

1. The Risk of Reputational Damage
2. Regulatory Fine, or
3. Legal Action.

With the increase in adoption of AI systems and its potential to impact so many aspects of our daily lives, it also highlights the potential risks around automated decision-making systems and the need for better governance and ethical standards when deploying such systems. The legal community should have a good understanding of the responsible development and deployment of AI systems in order to inform, translate and advise on the legal implications of the AI systems. Responsible development of AI systems ensures that the AI systems are safe, secure, and socially beneficial.

Cost of Responsible AI Development:

Responsible AI development can be seen as a collective action problem. If AI companies prefer to develop AI systems with risk levels that are closer to what is socially optimal. As individual AI companies may prioritize their own profits or competitive advantage over responsible AI development. This may compromise the collective welfare of the society.

Collaboration and adoption of ethical and responsible practices, along with industry standards and regulations can help address this problem and ensure that the development of AI systems is aligned with ethical principles and values.

What is Responsible AI development?

Responsible AI development aims to work on safety, security, and structural risks associated with AI systems.

Ensuring Safety: Aims to mitigate risks and ensure AI systems behave as intended and act in ways that are acceptable by people.

Ensuring Security: AI systems might be vulnerable to a wide range of security threats and can be exploited by malicious actors for harmful purposes. Attacks on AI systems also include manipulating the data used to train the system, exploiting the vulnerabilities in the AI algorithms, and attacks that target the hardware and software platforms on which the AI systems runs. Responsible AI development aims to protect and detect attacks on AI systems, such as adversarial attacks and data poisoning attacks. Ensuring that the AI systems can be trusted to perform their intended functions safely and reliably, without posing a risk to users or society at large.

Evaluating and Mitigating Structural Risks: Evaluating the structural impact of AI is an important aspect of AI governance, and also requires a multidisciplinary approach that not only considers the technical aspects of AI but also its broader societal and economic implications.?

The Structural Impacts of AI:

1.?????Potential for job displacement and changes in the labour market: As systems become more advanced and able to performs tasks previously done by humans. This may lead to job losses in certain industries and sectors. This can have significant social and economic implications and requires policymakers and stakeholders to consider how to address these challenges to ensure a just transition to the new economy.

2.?????Potential for use in military applications (Autonomous Weapons): AI systems used in military applications raises ethical and legal concerns. Policymakers and stakeholders need to consider how to regulate and control their use to mitigate the risk of military conflict and ensure compliance with international law.

3.?????Threat to Political and Social Institutions: Through the use of automated propaganda, surveillance, and disinformation campaigns, AI systems can pose a threat to political and social institutions. Assessing the structural impact of AI requires understanding these risks and developing strategies to implement them, such as ensuring transparency and accountability in AI systems, and promoting ethical and responsible development and deployment of AI systems.

Releasing unsafe products can have serious consequences for the companies beyond just the immediate financial costs of recalls or lawsuits. It can result in a loss of reputation which can be difficult and costly to rebuild. This can also lead to increased regulatory scrutiny and intervention, which could limit the company’s ability to innovate and compete.

Competitive Pressures among AI Companies Leading to “Collective Action Problems”:

In an AI development race, responsible development could be prey to a “Race to the Bottom” dynamic.?

Competition in the AI industry can have an adverse impact on the development of responsible AI, leading to a situation where each company has more incentives to prioritize the speed of AI system development over responsible and ethical consideration. In order to be ahead in the race of AI to launch first the new AI product or feature to the market, can result in cutting corners or overlooking potential risks of AI. This can create a race to the bottom where companies prioritize speed and profit over the potential risks and ethical considerations.

If AI companies prefer to develop AI systems with minimum risk levels, it can result in socially optimal outcome. If each company pursue their own self-interest, this can lead to an outcome where the risks of AI systems are higher than what is socially desirable. Because each company may choose to ignore the potential risks of AI systems in order to gain a competitive advantage or reduce development costs. Imagine a situation where companies focus on increasing their development speed while lowering their investment in safety, security, and impact evaluation.

We might ask why racing to the bottom on product safety is not ubiquitous or in other words why the trend of prioritizing speed-to-market over product safety is not as prevalent in other industries that value quick product development, such as pharmaceutical industry.

The most plausible explanation to this difference is that the Pharmaceutical Industry is heavily regulated by external agencies, such as Food and Drug Administration (FDA), which needs extensive testing and evaluation before they can be approved for use.

Additionally, potential liability for harms caused by defective drugs can be enormous leading to pharmaceutical companies prioritize safety and effectiveness in their products. Market forces also have huge roles to play as the consumers often demand safe and effective drugs and failures to meet these demands can result in significant financial and reputational harm to the company.

In recent years, researchers and activists have made significant efforts to highlight the biases present in the AI systems used in critical societal domains such as criminal justice and widely used technological platforms such as recommender systems. Similar concerns have been raised regarding the efficacy, bias, and other characteristics of medical AI systems, as well as emerging technologies such as self-driving vehicles. It is crucial to analyze and communicate these risks among wide range of stakeholders in order to generate interest in cooperation towards responsible AI development.

Joint research can concretize the benefits of AI, including collaborative efforts aimed at using AI for social good, which creates a shared upside. Collaborative research can also facilitate more socially responsible decisions regarding the publication and deployment of AI systems by enabling stakeholders to conduct collaborative risk analyses, leading to a shared upside.

Thankyou for reading this article. Hope you find this informative.

Rupa Singh

Founder and CEO (AI-Beehive)

Author of 'AI Ethics with Buddhist Perspective'