AI as a System
Asst. Prof. Jittat Fakcharoenphol, Vice-Chair of the Computer Engineering Department at Kasetsart University and the translator of “The Ethical Algorithm,” (by Michael Kearns and Aaron Roth, published by Salt) commented that it was particularly challenging to lay down ethical guidelines for AI applications. The task, he emphasized, requires clear operational procedures.
For example, the governance of an ethical algorithm must have clear definitions and components. If an algorithm is expected to be unbiased and non-discriminatory, drafters of the governance guidelines must define the term “bias” very clearly. To date, research has shown there are so many definitions of bias and the word can be interpreted in various ways. In Prof. Jittat’s opinion, the governance guidelines should identify biases clearly and set the right frameworks to reduce bias-related risks in data inputs for AI training.
“AI is a system,” he said. “Relevant parties should understand this fact for them to properly learn about, use and govern AI. Because it is a system, AI-related problems should be addressed together, not separately.”
Legitimacy of Data Usage
Montri Stapornkul, Head of Personal Data Protection at True Corporation Public Company Limited (True), described AI as a modern infrastructure with a crucial role to play at both enterprise and national levels. The utilization of AI can be divided into two parts: technology usage and data usage. Montri believes governance should be in place for data usage to protect the rights of data owners.
“True issued a charter for ethical AI to ensure that our data usage is transparent, legitimate, and in line with stated purposes. Our ethical AI application is reflected through our respect for data owners’ rights,” Montri said.
According to him, AI systems have three types of risks: 1) Risks related to Data Privacy Policy; 2) Risks related to Data Usage Conditions; and 3) Risks related to Public Understanding of AI.
To provide assurances that consumers will enjoy maximum benefits from AI applications, True has drawn up “True’s Ethical AI Charter,” which comprises four pillars: 1) Good Intent; 2) Fairness and Bias Mitigation; 3) Data Privacy; and 4) Transparency.
“Data is the foundation for AI applications. Although Thailand has not yet had a dedicated AI law, existing legislations had already laid down clear frameworks on data usage,” Montri said, referring to the Personal Data Protection Act B.E. 2562 and the National Broadcasting and Telecommunications Commission’s Announcement on the Protection of Telecom Service Users’ Rights to Privacy and Communications Freedom. “Metaphorically, AI development is like raising a child. When we prepare ethical guidelines for AI, it will progress well and become a good member of our society.”
His opinion was echoed by Raewat Tankittikorn, Head of Channel Excellence at True. Raewat said whether AI would be ethical or not depended on the kinds of data used for its training with the purpose of AI usage as another key factor.
