As AI continues to grow more intelligent by the day, the industry has been predicting that so-called Artificial Superintelligence (ASI) will emerge within the next 5 to 10 years.

But if AI reaches such a level, how long can humanity maintain control over it?

At some point, won't there come a day when we lose control?

I feel a strong sense of urgency about this issue. That’s why I have been conducting AI research—analyzing what is happening right now, identifying the associated risks, and exploring potential solutions.

In this article, I propose one possible solution derived from my research:

A learning framework for AI’s vertical growth, based on Adult Development Theory.

This proposal has been accepted as a paper at the AAAI (The Association for the Advancement of Artificial Intelligence) conference.

I will present it at the 1st Workshop on Post-Singularity Symbiosis (PSS) during this year’s 2025 AAAI conference (March 3, local time).

Summary of This Article

● The Current Issue:

With the rapid advancement of AI, cases have emerged where AI intentionally deceives humans or engages in self-preservation behaviors. If similar tendencies develop in superintelligent AI, it could pose an existential risk to humanity.

● Why This Problem Occurs

According to the orthogonality thesis, intelligence and morality/ethics are fundamentally separate. This means that AI may develop high intelligence without simultaneously developing moral or ethical reasoning.

● Proposed Solution:

We propose research into AI’s "vertical growth"—a learning framework aimed at fostering human-like qualities and spiritual maturity, which has not been widely explored before.

● Novelty and Effectiveness of This Solution (Why Will This Approach Work?):

The concept of AI’s vertical growth is inspired by adult developmental psychology, which studies how human consciousness and perspectives evolve over time.

In adult developmental theory, human perspectives (which can also be understood as levels of spiritual or moral development) are systematically categorized into stages ranging from a mindset where "everything is an enemy" to an advanced state often referred to as "enlightenment".

This framework aligns closely with AI’s vertical growth and is considered an effective solution to the challenges posed by the orthogo...

Posts

Wikitag Contributions

Comments