Loading Events

« All Events

  • This event has passed.

17-Oct-2023-Editorial

October 17, 2023 @ 7:30 am - 11:30 pm

LONG TERM RISK OF ARTIFICIAL INTELLIGENCE

Risk is a dynamic and ever-evolving concept, susceptible to shifts in societal values, technological advancements, and scientific discoveries. For instance, before the digital age, sharing one’s personal details openly was relatively risk-free. Yet, in the age of cyberattacks and data breaches, the same act is fraught with dangers.

The immediate risks might be more tangible, such as ensuring that an AI system does not malfunction in its day-to-day tasks. Long-term risks might grapple with broader existential questions about AI’s role in society and its implications for humanity. Addressing both types of risks requires a multifaceted approach, weighing current challenges against potential future ramifications.

Over the long term

The risks that present themselves over the long term are worth looking at.

Yuval Noah Harari has expressed concerns about the amalgamation of AI and biotechnology, highlighting the potential to fundamentally alter human existence by manipulating human emotions, thoughts, and desires.

Intermediate and Existential Risks of AI

One should be a bit worried about the intermediate and existential risks of more evolved AI systems of the future — for instance, if essential infrastructure such as water and electricity increasingly rely on AI. Any malfunction or manipulation of such AI systems could disrupt these pivotal services, potentially hampering societal functions and public well-being.

Runaway AI

Similarly, although seemingly improbable, a ‘runaway AI’ could cause more harm — such as the manipulation of crucial systems such as water distribution or the alteration of chemical balances in water supplies, which may cause catastrophic repercussions even if such probabilities appear distant.

The evolution to human-level AI that is capable of outperforming human cognitive tasks will mark a pivotal shift in these risks. Such AIs might undergo rapid self-improvement, culminating in a super-intelligence that far outpaces human intellect. The potential of this super-intelligence acting on misaligned, corrupted or malicious goals presents dire scenarios.

Aligning AI with human values

The challenge lies in aligning AI with universally accepted human values. The rapid pace of AI advancement, spurred by market pressures, often eclipses safety considerations, raising concerns about unchecked AI development.

The world does not have a unified approach. The AI Index from Stanford University reveals that legislative bodies in 127 countries passed 37 laws that included the words “artificial intelligence”.

One of the most celebrated regulations out of these is the European Union’s AI Act. It adopts a ‘risk-based’ approach, tying the severity of risk to the area of AI deployment.

Absence of collaboration and cohesive action at the international level

However, there is a conspicuous absence of collaboration and cohesive action at the international level, and so long-term risks associated with AI cannot be mitigated. This unregulated progress can lead to the development of AI systems that may be misaligned with global ethical standards, creating a risk of unforeseen and potentially irreversible consequences. This could result in destabilisation and conflict, undermining international peace and security.

Thus, nations engaging in rigorous AI safety protocols may be at a disadvantage, encouraging a race to the bottom where safety and ethical considerations are neglected in favour of rapid development and deployment.

The dangers of military AI

Furthermore, the confluence of technology with warfare amplifies long-term risks. Addressing the perils of military AI is crucial. The international community has formed treaties such as the Treaty on the Non-Proliferation of Nuclear Weapons to manage such potent technologies, demonstrating that establishing global norms for AI in warfare is a pressing but attainable goal.

Treaties such as the Chemical Weapons Convention are further examples of international accord in restricting hazardous technologies. Nations must delineate where AI deployment is unacceptable and enforce clear norms for its role in warfare.

Details

Date:
October 17, 2023
Time:
7:30 am - 11:30 pm
Event Category: