Dear privacy pros,

I hope this introduction finds you in good health and good spirits. To our Muslim colleagues, I wish you a wonderful Ramadan and Selamat Hari Raya Aidilfitri in advance!

ChatGPT and other powerful artificial intelligence systems continue to dominate headlines. In the few months since the last time I wrote on this topic, OpenAI released ChatGPT-4. It has apparently made great strides in functionality and accuracy, and promises an improvement five times more powerful than earlier models. GPT-4, short for generative pretrained transformer 4, already proved it is able to pass a bar exam (something which I derided the earlier version of ChatGPT for), as well as a final exam at Stanford's medical school and a Wharton Business School Master of Business Administration test. I find some solace in the fact that my boys still have the last laugh, as apparently even GPT-4 flunked the Primary School Leaving Examinations.

I think it is interesting, and potentially highly revelatory, that the world’s foremost experts in AI are now pushing for a pause on further development of AI systems more powerful than GPT-4. An open letter from the Future of Life Institute calling for this pause makes for a rather sobering read, with plenty of allusions to apocalyptic scenarios where "powerful digital minds that no one — not even their creators — can understand, predict, or reliably control" may potentially "outnumber, outsmart, obsolete and replace us."

Other experts appear to be heeding the call. The Center for Artificial Intelligence and Digital Policy, a U.S. tech advocacy group, called on the U.S. Federal Trade Commission to stop OpenAI from commercially releasing further iterations on GPT-4. It is pertinent to note that potential privacy breaches and security weaknesses feature prominently in the concerns highlighted in the group's complaint.

The United Nations Educational, Scientific and Cultural Organization similarly released a recommendation calling on governments around the world to implement a global AI ethics framework. Perhaps the Singapore Personal Data Protection Commission’s Model AI Governance Framework and other regulatory frameworks around AI ethics and governance would provide a useful template in this regard.

From a privacy perspective, in formulating such a global AI ethics framework, it is also useful to think about the rights of original content creators in this ecosystem, which will undoubtedly be a critical generator of huge economical value in the not-so-distant future. Perhaps Adobe's recent move to compensate creators who contribute the content used to train its generative AI tool Firefly out of the future revenue derived from images, videos, audio etc. generated by the tool can serve as a model for other companies in this space to emulate. Adobe’s decision to give creators the ability to mark their content with a "Do Not Train" tag is equally important and deserves to be heralded as a groundbreaking decision that should be applied to other types of data, including personal information.

With these closing thoughts, I hope you enjoy going through this week's Asia-Pacific Digest.