Answer Complex Questions From an Arbitrarily Large Set of Documents With Vector Search and GPT-3
David Shapiro ~ AI via YouTube
Overview
Syllabus
- The need for answering questions from arbitrary data sources
- The benefits of a multi-document answering system
- Building a chatbot to answer questions and summarize documents
- Breaking up text into chunks for indexing
- Building an index
- Converting unicode to ascii to prevent gpt3 errors
- Saving data as a JSON file
- Building the index
- Building a Knowledge Base
- Building the index
- Searching for memories
- Using GPT-3 to answer questions about a text
- The majority's decision to allow states to ban abortion
- Answering a question with a superintelligence
- Generating answers to questions with GPT-3
- Joining answers into one big block
- Creating a detailed summary of chunks
- Trying to fix a broken search
- Fixing the bug in the gpt3 completion function
- GPT3's difficulty with complex questions
- The gpt3 log
- The Supreme Court's decision on abortion
- The Supreme Court's decision on abortion
- The Supreme Court overturns a lower ruling banning abortion
- The final result of the superintelligence question
- The Supreme Court overturns Roe v. Wade
- The Supreme Court overturns Roe v. Wade
Taught by
David Shapiro ~ AI