Advertisement
Hong Kong society
OpinionLetters

Letters | Hong Kong’s AI push must include safeguards against AI hallucination

Readers discuss a lesson learned from a recent academic scandal, easing the rules of tax residency in Hong Kong, and dire predictions about the Chinese economy

Reading Time:3 minutes
Why you can trust SCMP
1
HKU has launched an investigation into a published doctoral thesis, after it was found to contain fictitious AI-generated references. Photo: Shutterstock
Letters
I refer to your report, “Journal defends work with fake AI citations after Hong Kong university launches probe” (November 10).
The recent controversy involving a paper written by a University of Hong Kong PhD candidate highlights a growing concern in this artificial intelligence era – AI hallucination. This phenomenon, where AI tools fabricate information that appears credible but is factually false, poses serious challenges to both academic integrity and public trust in innovation.
As AI becomes more embedded in daily life, the risk of such AI hallucinations extends beyond academia. The Hong Kong government has announced plans to establish a high-level AI Efficacy Enhancement Team to help integrate AI applications into public administration.
Advertisement

This initiative is commendable, as it recognises the technology’s potential to boost efficiency in government work. However, the HKU case is a timely reminder that AI must assist but not replace human judgment.

No matter how advanced AI becomes, human oversight remains irreplaceable. To ensure accountability, it is essential to ensure human gatekeeping mechanisms at key decision points. Whether in policymaking, data analysis or academic research, professionals must verify and interpret AI-generated outputs with critical discernment.

Advertisement

AI can be a powerful tool for productivity, but only when guided by responsible human governance. Convenience must never come at the cost of credibility.

Advertisement
Select Voice
Choose your listening speed
Get through articles 2x faster
1.25x
250 WPM
Slow
Average
Fast
1.25x