|
Getting your Trinity Audio player ready...
|
Liming Zhu and Qinghua Lu, leaders in the study of responsible AI at CSIRO and Co-authors of Responsible AI: Best Practices for Creating Trustworthy AI Systems delve into the realm of responsible AI through their extensive work and research.
Artificial Intelligence (AI), currently a major focal point, is revolutionising almost all facets of life, presenting entirely novel methods and approaches. The latest trend, Generative AI, has taken the helm, crafting content from cover letters to campaign strategies and conjuring remarkable visuals from scratch.
Global regulators, leaders, researchers and the tech industry grapple with the substantial risks posed by AI. Ethical concerns loom large due to human biases, which, when embedded in AI training, can exacerbate discrimination. Mismanaged data without diverse representation can lead to real harm, evidenced by instances like biased facial recognition and unfair loan assessments. These underscore the need for thorough checks before deploying AI systems to prevent such harmful consequences.
The looming threat of AI-driven misinformation, including deepfakes and deceptive content, concerning for everyone, raising fears of identity impersonation online. The pivotal question remains: How do we harness AI’s potential for positive impact while effectively mitigating its capacity for harm?
Responsible AI involves the conscientious development and application of AI systems to benefit individuals, communities, and society while mitigating potential negative impacts, Liming Zhu and Qinghua Lu advocate.
These principles emphasise eight key areas for ethical AI practices. Firstly, AI should prioritise human, societal, and environmental well-being throughout its lifecycle, exemplified by its use in healthcare or environmental protection. Secondly, AI systems should uphold human-centred values, respecting rights and diversity. However, reconciling different user needs poses challenges. Ensuring fairness is crucial to prevent discrimination, highlighted by critiques of technologies like Amazon’s Facial Recognition.
Moreover, maintaining privacy protection, reliability, and safety is imperative. Instances like Clearview AI’s privacy breaches underscore the importance of safeguarding personal data and conducting pilot studies to prevent unforeseen harms, as witnessed with the chatbot Tay generating offensive content due to vulnerabilities.
Transparency and explainability in AI use are vital, requiring clear disclosure of AI limitations. Contestability enables people to challenge AI outcomes or usage, while accountability demands identification and responsibility from those involved in AI development and deployment. Upholding these principles can encourage ethical and responsible AI behaviour across industries, ensuring human oversight of AI systems.
Identifying problematic AI behaviour can be challenging, especially when AI algorithms drive high-stakes decisions impacting specific individuals. An alarming instance in the U.S. resulted in a longer prison sentence determined by an algorithm, showcasing the dangers of such applications. Qinghua highlighted the issue with “black box” AI systems, where users and affected parties lack insight into and means to challenge decisions made by these algorithms.
Liming emphasised the inherent complexity and autonomy of AI, making it difficult to ensure complete compliance with responsible AI principles before deployment. Therefore, user monitoring of AI becomes crucial. Users must be vigilant and report any violations or discrepancies to the service provider or authorities.
Holding AI service and product providers accountable is essential in shaping a future where AI operates ethically and responsibly. This call for vigilance and action from users is instrumental in creating a safer and more accountable AI landscape.
Australia is committed to the fair and responsible use of technology, especially artificial intelligence. During discussions held on the sidelines of the APEC Economic Leaders Meeting in San Francisco, the Australian Prime Minister unveiled the government’s commitment to responsibly harnessing generative artificial intelligence (AI) within the public sector.
The DTA-facilitated collaboration showcases the Australian Government’s proactive investment in preparing citizens for job landscape changes. Starting a six-month trial from January to June 2024, Australia leads globally in deploying advanced AI services. This initiative enables APS staff to innovate using generative AI, aiming to overhaul government services and meet evolving Australian needs.
