Fix embedding model mismatch between Question_answering_using_embeddi…

[sheldonrampton]

Summary

The article on "Embedding_Wikipedia_articles_for_search" is intended to show how the embeddings were generated for the article on "Question_answering_using_embeddings." However, the "Embedding_Wikipedia" code specifies a different EMBEDDING_MODEL than the "Question_answering" model. As a result, the "Question_answering" code fails to understand the embeddings and returns returns "I could not find an answer" when it tries to answer the questions. This pull request fixes that EMBEDDING_MODEL mismatch.

[sheldonrampton]

After submitting this pull request, I realized that the embeddings file which is referenced in
https://cookbook.openai.com/examples/question_answering_using_embeddings
references this embeddings path:
https://cdn.openai.com/API/examples/data/winter_olympics_2022.csv
which was generated using text-embedding-ada-002 as its EMBEDDING_MODEL. If you accept my pull request, therefore, you will need to regenerate the embeddings file using text-embedding-3-small as the embedding model.

Alternately, you could fix the embedding model mismatch by using text-embedding-ada-002 for both the "Question answering" and the "Embedding Wikipedia" articles.