Safe and Profitable SuperIntelligence

An iQ White Paper by Dr. Craig A. Kaplan

July 2024

ABSTRACT

SuperIntelligence is defined as advanced AI that outperforms humans on any cognitive task. It has a long history but may arrive faster than many researchers think. When it arrives, it will create trillions of dollars in profits and tremendous benefits for humanity. However, SuperIntelligence also poses an existential threat if not developed safely.

Safe designs require solving three related safety problems: 1) The Transparency Problem, 2) The Design Problem, and 3) The Alignment Problem.

Our approach addresses these problems. It uses machine learning techniques to train and customize autonomous AI agents, called Advanced Autonomous Artificial Intelligences (AAAIs). These AAAIs work transparently and safely together on a network using a common architecture to enable SuperIntelligence.

Our system designs are compatible with the approaches of the AI Alliance, META, IBM, Microsoft, and many other leading technology companies. The designs are the most profitable path to SuperIntelligence because they enable developing SuperIntelligence faster and more safely than existing approaches.

​

FULL PAPER