Abstract:Perhaps the most ambitious scientific quest in human history is the creation of general artificial intelligence, which roughly means AI that is as smart or smarter than humans. The dominant approach in the machine learning community is to attempt to discover each of the pieces required for intelligence, with the implicit assumption that some future group will complete the Herculean task of figuring out how to combine all of those pieces into a complex thinking machine. I call this the "manual AI approach". This paper describes another exciting path that ultimately may be more successful at producing general AI. It is based on the clear trend in machine learning that hand-designed solutions eventually are replaced by more effective, learned solutions. The idea is to create an AI-generating algorithm (AI-GA), which automatically learns how to produce general AI. Three Pillars are essential for the approach: (1) meta-learning architectures, (2) meta-learning the learning algorithms themselves, and (3) generating effective learning environments. I argue that either approach could produce general AI first, and both are scientifically worthwhile irrespective of which is the fastest path. Because both are promising, yet the ML community is currently committed to the manual approach, I argue that our community should increase its research investment in the AI-GA approach. To encourage such research, I describe promising work in each of the Three Pillars. I also discuss AI-GA-specific safety and ethical considerations. Because it it may be the fastest path to general AI and because it is inherently scientifically interesting to understand the conditions in which a simple algorithm can produce general AI (as happened on Earth where Darwinian evolution produced human intelligence), I argue that the pursuit of AI-GAs should be considered a new grand challenge of computer science research.

Generative AI: An AI paradigm shift in the making?

Generative Artificial Intelligence: A Systematic Review and Applications

Generative Ai: potential and pitfalls

Generative AI: An In-depth Exploration of Methods, Uses, and Challenges

The Generative AI Paradox: "What It Can Create, It May Not Understand"

The Paradox of Artificial Creativity: Challenges and Opportunities of Generative AI Artistry

On the Challenges and Opportunities in Generative AI

Generative Artificial Intelligence: Trends and Prospects

Future of software development with generative AI

Scientists' Perspectives on the Potential for Generative AI in their Fields

Generative Artificial Intelligence and GPT using Deep Learning: A Comprehensive Vision, Applications Trends and Challenges

Generative AI: Overview, Economic Impact, and Applications in Asset Management

Generative AI Systems: A Systems-based Perspective on Generative AI

AI-GAs: AI-generating algorithms, an alternate paradigm for producing general artificial intelligence

Enhancing Reality: Exploring the Potential of Generative Artificial Intelligence

Generative AI: is it a paradigm shift for higher education?

Generative AI: Exploring the Applications of Generative Models in Creative Industries

AI Horizons: Unveiling the Future of Generative Intelligence

Can There be Art Without an Artist?

Generative AI: A systematic review using topic modelling techniques

The Road Ahead: Emerging Trends, Unresolved Issues, and Concluding Remarks in Generative AI—A Comprehensive Review