What is an intelligence explosion?

The idea of an "intelligence explosion" was first articulated by statistician I.J. Good in 1965. It is a scenario where an AI becomes capable of rapidly building an AI with a higher level of capability than its own, and then this new AI does the same, and so on, leading to explosive growth in AI capabilities.

This term is used much less often than it used to be. Related concepts that are more commonly discussed are:

  • Recursive self-improvement: the process of AI systems iteratively improving themselves, which could result in an intelligence explosion.

  • Takeoff speed: how quickly the process of AI surpassing humans will occur.

This video by Emergent Garden explains in more detail what an intelligence explosion could look like if it used more advanced versions of current language models.