Understanding Artificial Intelligence, Artificial General Intelligence, and Artificial Superintelligence?
Ben Goertzel, CEO of SingularityNET and the Artificial Superintelligence (ASI) Alliance, is widely recognised as a pioneer in the field of artificial intelligence (AI). He is also credited with introducing the concept of Artificial General Intelligence (AGI), and has been at the forefront of discussions around AI’s future. Recently, Goertzel shared valuable insights on key AI terms—AI, AGI, and Artificial Superintelligence (ASI)—which are often used interchangeably but have distinctly different meanings. According to Goertzel, understanding these terms is essential for navigating the evolving AI landscape.
As Goertzel himself noted, "I am seeing constant confusion these days about the basic meanings of terms like AGI, ASI, etc."
Despite the fluidity and complexity of these terms, he aims to clarify them for those within the AI research world. Here’s a breakdown of what these terms mean, according to Goertzel’s perspective:
Artificial Intelligence (AI):
At its core, AI refers to any software or hardware that performs tasks people consider "smart." These tasks typically involve problem-solving, pattern recognition, and decision-making. Goertzel explains that AI is simply about enabling machines to do something that humans do with intelligence—whether it's recognizing speech, interpreting images, or making predictions based on data.
“AI is any software or hardware doing something that people consider smart when people do it,” Goertzel said.
Today’s AI systems are typically narrow in scope, meaning they excel at specific tasks but are limited to those applications. Examples of narrow AI include virtual assistants, chatbots, and recommendation systems.
Artificial General Intelligence (AGI): Beyond Task-Specific Intelligence
AGI, on the other hand, refers to an AI system that can generalise knowledge and apply it to a wide range of tasks, just as humans do. AGI goes beyond performing specific tasks and can
“generalize imaginatively and fluently beyond its training data and programming,” as Goertzel explains.
It has the ability to adapt and apply learning from one domain to solve problems in entirely different areas, mimicking human flexibility and creative thinking.
While there are theoretical frameworks for AGI, such as Marcus Hutter’s book Universal AI, practical tests to measure AGI's capabilities are still evolving. Goertzel points out that current AI systems, including large language models (LLMs), “do very poorly” on AGI tests like Francois Chollet’s AGI-ARC, which measure general intelligence across humans and AI.
Human-Level AGI (HLAGI): Cognitive Capabilities Similar to Humans
Human-Level AGI (HLAGI) is a critical step on the path toward full AGI. It refers to AI that possesses cognitive abilities similar to a normal adult human, such as the ability to reason, solve complex problems, and generalise knowledge creatively. Goertzel stresses that HLAGI is capable of generalising beyond its training data and imagining new solutions, much like humans do. Achieving HLAGI would represent a significant breakthrough in the development of AI, providing the foundation for even more advanced systems.
Artificial Superintelligence (ASI): A New Frontier of Intelligence
The next level, ASI, or Artificial Superintelligence, represents a leap far beyond human-level intelligence. ASI would not only replicate human cognitive abilities but surpass them across every conceivable domain, including creativity, strategy, and problem-solving.
"ASI is expected to come with the ability to self-analyze and self-modify to a level far beyond HLAGIs," Goertzel notes.
This means ASI would not only be smarter than humans but could also enhance and improve itself autonomously, creating even more advanced AIs at an accelerating pace.
While ASI promises to revolutionise industries and tackle complex global challenges, Goertzel acknowledges that even in a world with ASI, humans may still hold some unique value. Just as plants in the Amazon or our gut bacteria provide essential contributions despite having less intelligence, humans may continue to offer valuable perspectives even after ASI has surpassed us in cognitive abilities.
What Comes After ASI? Ultraintelligence?
Goertzel speculates on what might come after ASI—Ultraintelligence. This could represent an intelligence that masters complex, paradoxical concepts and potentially manipulates aspects of the multiverse. While still theoretical, this notion highlights the boundless possibilities for AI to expand beyond our current understanding. Goertzel humorously leaves this as “TBD”—a sign that the future of AI is still wide open and full of potential.
AI’s Evolving Journey
Ben Goertzel’s reflections on AI, AGI, and ASI provide a roadmap for understanding the stages of AI development. While these concepts are still fluid and evolving, they help frame the potential trajectory of artificial intelligence. Goertzel’s insights reflect a vision of AI that is both exciting and challenging, one where human and machine intelligence might coexist in ways we are just beginning to understand.
As AI continues to evolve, these concepts will play a vital role in shaping the future of technology and its integration into society. The journey from AI to AGI and ASI may seem distant, but the pace of advancements suggests we are moving toward a future where intelligent machines play a central role in our world, with humanity still offering unique value to the emerging intelligence around us.
