Update FAQ

FAQ.md

@@ -1,4 +1,4 @@
-### Why does the h2oGPT say it was trained by OpenAI or Open Assistant?
+### Why does h2oGPT say it was trained by OpenAI or Open Assistant?
 
 ![](https://user-images.githubusercontent.com/6147661/233486736-812d7b95-8c2f-438e-be76-ec4845c28a33.png)
 
@@ -10,6 +10,7 @@ answer and similar answers. In other words, they "contaminated" the training dat
 Most of the knowledge of the model is from pre-training on the billions of tokens, the fine-tuning only turns that language
 model into a chatbot by returning short answers for short questions, or in other words, pre-training creates language
 understanding and some knowledge, while fine-tuning injects style. Certain simple personality traits can be modified by fine-tuning however, and we are working on giving h2oGPT proper personality: https://github.com/h2oai/h2ogpt/issues/73
+Update: We continued fine-tuning with the [h2oai/h2oai/openassistant_oasst1_h2ogpt](https://huggingface.co/datasets/h2oai/openassistant_oasst1_h2ogpt) dataset, which includes some personalization, and the effects are noticeable.
 
 ### Is h2oGPT multi-lingual?