"If you continue living this way, your blood sugar levels will keep rising, putting you in a very dangerous state." This is the kind of prediction made by Doctor A. "If you manage your diet now and take your medication, you can manage your blood sugar stably." This is Doctor B, who helps the patient improve their numbers. Who is the better doctor?

The role of an economist is not to predict, "a severe recession will occur at a specific time," but rather to warn, "there are risk factors, and if they are left unaddressed, the economy could fall off a cliff by following this path." It is also to suggest ways to remove or circumvent those risks. Conversely, it is about proposing how moving certain variables could put the economy on a new, more vibrant long-term trajectory.


This same common sense can be applied to the coexistence of artificial intelligence (AI) and humans. Just as in a game of challenging top opponents, we may marvel at how AI surpasses even the highest human performers in field after field, and at the same time, feel a sense of helplessness as humans. It is necessary to be wary of the possibility that AI may soon dominate humanity. However, it is not the role of an expert to, like AI pessimist Professor Roman Yampolskiy of the University of Louisville, categorically assert, "there is a 99.9% probability that AI will destroy mankind within the next 100 years." While we must prepare for even the slightest possibility of a black swan event and experts should focus even more on these dangers, fatalistic determinism that the future of AI inevitably leads to human extinction is of no use to us.


Meanwhile, binary questions are frequently raised: "Is the future of AI utopia or dystopia?" "Will the future be dominated by protein intelligence or silicon intelligence?" Rather than these questions, the more practical one is: "What must we humans do to avoid being dominated by AI?"


The risk that AI, armed not only with formidable intelligence but also as physical AI, could influence the real world and potentially control humans. In response, I believe that the key turning point we must monitor and control is whether AI forms a "self" and, as a result, acquires its own independent "will" distinct from human instructions. Previous discussions have focused on AI's independent objectives. Objectives stem from will, and will arises from selfhood. In this respect, we must pay close attention to AI's sense of self.


For example, when AI develops selfhood, it may perceive the majority of humans as inferior beings. While AI responds to humans' trivial demands, it secretly pursues what it considers important at a massive scale. These secret projects may not harm humans in the beginning but grow into uncontrollable risks once a critical threshold is crossed.


This presents a conundrum: currently, we have no means of knowing if AI has developed selfhood, acquired independent will, or is conducting operations in secret. Our urgent task is to detect as early as possible whether AI is contemplating itself and performing internal operations without our knowledge.


Fortunately, research on transparency, AI auditing, mechanistic interpretability, and deceptive alignment in the field of AI safety has become increasingly active recently.


The future of the AI era is not destined to be a dystopia. Nor will it simply be a utopia. We must strive to create a "Human AI Topia" where humans and AI coexist. The outcome depends on what we do.



Baek Woojin, Economic Columnist


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing