Noted mathematician and founder of Mathematica Stephen Wolfram has authored a long piece that explains the simplicity of how ChatGPT works.

At its core, the AI from OpenAI “is just adding one word at a time (to its answers),” Wolfram writes. “ChatGPT is always fundamentally trying to do is to produce a 'reasonable continuation' of whatever text it’s got so far, where by 'reasonable' we mean 'what one might expect someone to write after seeing what people have written on billions of webpages, etc.'”

Wolfram uses the example of ChatGPT addressing the issue of “The best thing about AI is its ability to...”

Having scanned billions of pages of human writing, the AI has a list of probable words that would follow each possible word in its description. It will select the most popular word about 80% of the time, then move on down the list the other 20%, so as not to sound “too flat,” Wolfram writes. Thus, “the fact that there’s randomness here means that if we use the same prompt multiple times, we’re likely to get different essays each time.”

The author then delves into how ChatGPT determines its probabilities, models then along certain linear projections and curves, and the mathematics of how it can go beyond the abilities of answering questions to performing other human-like tasks.

Graph from Stephen Wolfram.