Dust Profile picture
Build Smarter Teams with Generative AI Internal data vectorized, not rasterized - Smart read and write tools - Your own powerful workflows
Nov 14, 2022 5 tweets 1 min read
Teaching LLMs to teach themselves new tasks by teaching them to generate few shot examples from high-level tasks descriptions ♾️

Given a new task description the model generates few-shot examples and prompt itself with them!

dust.tt/spolu/a/b39f8e…

🧵👇Some learnings Interestingly fails with text-davinci-002 but works really well with code-davinci-002.

Looks like instruction following training interferes with this “meta meta learning” or “self-teaching” process.
Oct 31, 2022 7 tweets 3 min read
WebGPT reproduced from advanced prompting only.

Dust-based web-search assistant demo answers questions by searching the web, summarizing content and compiling a final answer with references:

dust.tt/spolu/a/41770f… This app performs the following steps:

1/ search google, take first 3 results
2/ call to a @Replit to get content (headless browsing)
3/ 1st model call to clean-up and summarize each site
4/ 2nd model call (few-shot prompted) to generate final answer with references