GPT-4 Prompt Tries To Generate Fundamentally New Knowledge

by asking for an explanation for any poorly understood phenomena that humans have a hard time explaining, GPT is forced to create novel hypotheses

Brandon Gorrell

This weekend, @gfodor tweeted a GPT-4 prompt that generates seemingly novel explanations for poorly understood phenomena.

Here’s the prompt:

What’s an example of a phenomenon where humanity as a whole lacks a good explanation for, but, taking into account the full set of human generated knowledge, an explanation is actually possible to generate? Please write the explanation. It must not be a hypothesis that has been previously proposed. A good explanation will be hard to vary.

Twitter users have been sharing screenshots of GPT-4’s responses to the prompt, including explanations of the Fermi Paradox, the Roanoke Colony disappearance, and the persistent presence of war and violence throughout human history.

The short, straightforward prompt is a simple but creative command that ‘forces’ GPT-4 to generate explanations that have possibly never been proposed, suggesting that GPT-4 is capable of creating ‘new knowledge’. That the explanation must be “hard to vary” means that if “it was disproven, could you tweak it into a new explanation that isn’t falsified yet. If so, it’s a bad explanation,” according to @gfodor.

Presumably, you can get GPT-4 to propose new explanations for specific phenomena of your choosing by slightly tweaking the language in the prompt. I used the following to ask GPT-4 specifically about the Big Bang:

[PHENOMENON OF YOUR CHOICE] is a phenomenon where humanity as a whole lacks a good explanation for, but, taking into account the full set of human generated knowledge, an explanation is actually possible to generate. Please write the explanation for [PHENOMENON OF YOUR CHOICE]. It must not be a hypothesis that has been previously proposed. A good explanation will be hard to vary.

GPT-4 proposed that the big bang can be explained by “the fractal multiverse.”