3 ms·
Can someone help me understand how (or why) Large Language Models like ChatGPT and Bing/Sydney follow directives at all - or even answer questions for that matt
by curiousgiorge 4y ago
Can someone help me understand how (or why) Large Language Models like ChatGPT and Bing/Sydney follow directives at all - or even answer questions for that matter. The recent ChatGPT explainer by Wolfram (https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-why-does-it-work/ https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-...) said that it tries to provide a '“reasonable continuation” of whatever text it’s got so far'. How does the LLM "remember" past interactions in the chat session when the neural net is not being trained? This is a bit of hand wavy voodoo that hasn't been explained.
- layer8 4y agoSupposedly, directives are injected as prompts at the beginning of each session (invisibly to you the user). It’s exactly the same as if they weren’t automatically injected and you typed them in instead. The model is trained such that producing output consistent with such prior directive prompts is more likely. But it’s not ironclad, as the various “jailbreaks” demonstrate.
- valine 4y agoIt’s predicting the next word given the text so far. The entire chat history is fed as an input for predicting the next word.
- stevenhuang 4y agoIt's fine tuned to respond in a way that Humans recognize as a Chat bot. I've bookmarked the best explanation I found on how ChatGPT was trained from GPT3, once I get home I'll share it here.
- stevenhuang 4y agoHere it is: https://reddit.com/comments/10q0l92/comment/j6obnoq https://reddit.com/comments/10q0l92/comment/j6obnoq