748: The Five Levels of AGI — with Jon Krohn (@JonKrohnLearns)

Narrow AI

The concept of narrow AI is exemplified by systems that excel in specific tasks, often surpassing human capabilities in those areas. Jon Krohn explains that these systems are categorized into levels based on their proficiency compared to skilled adults. For instance, Grammarly and image generation models like DALL-E 2 are considered expert narrow AI, performing better than 90% of skilled adults in their respective tasks 1. At the virtuoso level, systems like IBM's Deep Blue and AlphaGo demonstrate capabilities that exceed even the best human players in chess and Go 1.

They're better than 90% of skilled adults at spelling and grammar, checking, or generating images.

--- Jon Krohn

These examples highlight the impressive achievements of narrow AI, yet they remain limited to specific domains without the generality of AGI.

   

Emerging AGI

Emerging AGI systems represent the cutting edge of AI technology, capable of performing a wide range of non-physical tasks. Jon Krohn notes that these systems, such as OpenAI's GPT-4 and Google's Gemini, are at the forefront of AI development, yet they only match or slightly surpass unskilled human performance 2. Despite their broad capabilities, these systems are still classified as emerging AGI, indicating that they have not yet reached the level of skilled human performance across diverse tasks.

Now, in terms of emerging AGI, so in terms of an emerging general algorithm, so we're still at this level one where things are just equal to or somewhat better than an unskilled human.

--- Jon Krohn

This distinction underscores the ongoing journey towards achieving true artificial general intelligence.