Skip to content
TheChatArchiveThe Chat Archive
Human-written demonstration (OpenAssistant volunteers)· OpenAssistant (human-written)Hugging Face OpenAssistant Conversations (oasst2)

Explain how neural networks work in Layman's terms

In OpenAssistant data, “assistant” turns were written by human volunteers as demonstrations. They are not model outputs.
  1. User

    Explain how neural networks work in Layman's terms

  2. Assistant

    Neural networks are trained to produce satisfactory outputs based on the inputs they receive. The way in which they respond to any input is decided by an enoumeous collection of interconnected numbers who collectively controll what the network is going to output. By randomly changing the numbers inside the black box every time it produces an unsatisfactory output, the network eventually starts to create better results. Nobody, not even the experts, knows what each of the numbers does or how they work together. But the important part is that black box doesn't need to be understood, it just needs to work.

    Even though the intelligence, structure and development of a human brain and a neural network are very different, the total mystery surrounding the way they compute stuff is very similar.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.