AI has become the latest buzzword in the world and it seems like everyone is racing to understand its capabilities and potential impact.
It was popularly assumed that the future of AI would be the automation of mundane repetitive tasks requiring low-level decision-making. However, AI has advanced quickly, driven by more powerful computing systems and the accumulation of vast data sets. One key area, machine learning, stands out for its capability to process and analyse enormous volumes of data while improving over time. This has revolutionised numerous sectors, including education, health care, banking, retail, manufacturing, and even judiciary.
AI also has integrated itself into our daily lives more than we realise. This also raises questions about their control and alignment with human values. From smartphone assistants that manage our tasks to biometric devices with facial recognition and motion sensors, AI is embedded in many things we use daily. Devices like robot vacuum cleaners mapping out our homes, smart refrigerators, air conditioners, thermostats, lighting systems, and even social media platforms all rely on AI to function seamlessly.
Let us discuss a few scenarios that cause these ethical concerns.
AI deep fake dead voices
One recent development that sparked my interest in this area is the use of AI to synthesise the voices of deceased individuals in movies in India. While it was awe-inspiring and nostalgic to hear those voices again, it raises serious ethical questions about how far we are pushing the ethical boundaries in the use of AI. There are no established rules or precedents in this space.
This issue extends beyond the deceased, as many people (who are alive) are facing similar challenges, with their faces/voice or likeness being used without consent, leaving them vulnerable and struggling to protect their rights.
Autonomous car
An autonomous car uses sensors to perceive its environment and drive with minimal human input, processing vast data through its onboard system. In any traffic scenario, the car must make moral decisions, such as how to react in an emergency. While AI can be programmed to choose the best course of action, it may still be influenced by the biases present in the data provided by its developers
Loan approvals
Banks are highly regulated and are legally on the hook if the algorithms they use to evaluate loan applications end up inappropriately discriminating against classes of consumers. This scrutiny arises from the potential for algorithms to inadvertently replicate or exacerbate existing biases found in historical data, leading to discriminatory outcomes in lending practices.
ChatGPT or GPT:
Training large-scale AI models requires powerful data centers and servers, which consume vast amounts of electricity. A study from the University of Massachusetts, Amherst estimated that training a single AI model (specifically, a large neural network) can emit as much carbon dioxide as five cars do over their entire lifetimes, including manufacturing. This highlights the environmental burden AI technologies can impose if their development relies heavily on energy-intensive practices.
Ethics of Artificial Intelligence
The UN had proposed a few recommendations on ethical concerns about AI to tackle the faster growth of AI among the communities.
Some of the key aspects of these recommendations included the following.
Protection of Human Rights:
- Ensuring AI respects human dignity, privacy, and non-discrimination.
- AI must be designed in a way to uphold human rights and freedom and remove the existing societal bias and barriers that might affect certain individuals or groups.
- Data collection for AI models must be conducted with respect for individuals’ privacy and autonomy, ensuring that informed consent is obtained and that the data is used responsibly.
Transparency and Accountability:
- Requiring AI systems to be understandable and those developing or deploying them to be held accountable.
- This includes clear documentation of how the AI models work and the audits that are in place to protect against any misuse.
- Users also must have control over how their data is collected, stored, and processed, with clear transparency about how their information will be used to prevent violations of their rights and dignity
Promoting Inclusion and Fairness:
- Fostering access to AI benefits for all, particularly vulnerable and marginalised groups
- The needs of all human beings must be taken into consideration so that everyone can benefit, and all individuals can be offered the best possible conditions to express themselves and develop.
- The potential for AI to replace human jobs could deepen existing inequalities, particularly for marginalised or underrepresented groups, including women and cultural minorities.
Environmental Sustainability:
- Encouraging the development of AI systems that consider their environmental impact.
- This includes using renewable energy and minimising carbon footprint associated with the training of these models and algorithms.
Although many may be unaware of these recommendations and how member nations implement them, Australia has taken proactive steps by aligning with these guidelines and establishing a comprehensive set of governance policies. For more information, you can visit the official site detailing Australia’s AI ethics principles here.
Conclusion:
As AI continues to advance and integrate into various aspects of life, the intersection of AI, cybersecurity, and governance becomes increasingly critical. Proactive measures must be implemented to safeguard against potential risks, and mitigating threats to security and privacy. Although we have established some governance policies for AI development, significant work remains to enhance accountability and eliminate existing biases in the models we create.
