Randomness

AI is not precise and not 100% reproducible. It doesn’t matter how good your prompt is; AI will generate different outputs. AI is really random in nature. Cory Doctorow and his amazing post: LLMs are slot-machines cleverly explains the random nature of LLM outputs. If you need 100% precision, Generative AI is not for you; LLMs are not for you.

Generative AI is just trying to predict the next sequence of tokens. That’s all, an auto-complete on steroids. You need to understand the random nature of Gen AI when using LLMs.

This image was generated by GPT 5.1, can you see the issues in the image? AI is good with TEXT, LLMs are all about text. LLMs get much worse with images and 100x worse with video because images and videos are not text-based. Funny that the slot machine metaphor from Cory Doctorow’s post is 100% spot on about the lack of accuracy, replicability, and precision of LLMs.