Source URL: https://simonwillison.net/2025/Apr/20/ethan-mollick/#atom-everything
Source: Simon Willison’s Weblog
Title: Quoting Ethan Mollick
Feedly Summary: In some tasks, AI is unreliable. In others, it is superhuman. You could, of course, say the same thing about calculators, but it is also clear that AI is different. It is already demonstrating general capabilities and performing a wide range of intellectual tasks, including those that it is not specifically trained on. Does that mean that o3 and Gemini 2.5 are AGI? Given the definitional problems, I really don’t know, but I do think they can be credibly seen as a form of “Jagged AGI” – superhuman in enough areas to result in real changes to how we work and live, but also unreliable enough that human expertise is often needed to figure out where AI works and where it doesn’t.
— Ethan Mollick, On Jagged AGI
Tags: gemini, ethan-mollick, generative-ai, o3, ai, llms
AI Summary and Description: Yes
Summary: The text highlights the dual nature of AI’s reliability, presenting a nuanced understanding of its capabilities. Specifically, it introduces the concept of “Jagged AGI,” suggesting that while some AI models like o3 and Gemini 2.5 display superhuman capabilities, their unreliability necessitates human oversight. This perspective is vital for professionals in AI and security, emphasizing the need for understanding AI’s limitations as well as its strengths.
Detailed Description: The provided text discusses the evolving landscape of artificial intelligence and its ramifications. Here are the major points:
– **Dual Nature of AI**: The author notes that while AI can outperform humans in various tasks, it also exhibits inconsistencies and unreliability.
– **Comparison to Calculators**: The text draws an analogy between AI’s performance and that of calculators, suggesting that both types of tools have distinct capabilities but highlight a fundamental difference in how they process and produce outcomes.
– **General Capabilities**: The mention of AI, including models such as o3 and Gemini 2.5, points to their ability to tackle various intellectual tasks beyond their specific training sets.
– **Jagged AGI**: The term “Jagged AGI” introduces a new perspective on AI classifications, emphasizing that while certain models can achieve superhuman performance, they still require human intervention to navigate their limitations.
– **Need for Human Expertise**: Due to AI’s unreliability in specific areas, the text stresses the importance of human oversight, particularly in determining when AI can be trusted to perform effectively.
This analysis holds particular significance for professionals in security and compliance, indicating the need for strategies that incorporate human judgment alongside AI capabilities. As AI technology continues to evolve, understanding its limitations will be crucial for deploying it securely in various applications, particularly in sensitive domains where misinformation and errors could have severe consequences.