Machine Learning Study Group
 |
Welcome! We meet from 4:00-4:40 p.m. Central Time. Anyone can join. Feel free to attend any or all sessions, or ask to be removed from the invite list as we have no wish to send unneeded emails of which we all certainly get too many. |  |
Contacts: jdberleant@ualr.edu and mgmilanova@ualr.edu
Agenda & Minutes
105th meeting, Mar. 15, 2024
- DU will lead discussion on "Efficacy of ChatGPT 4.0 in Translations between English and Spanish for Healthcare Patient Requests for Medical Advice"
- Discussion, suggestions, and questions:
- Why translate with ChatGPT and not Google Translate (for example)?
- Will you measure improvement in outcomes by collecting before-and-after data?
- What is your hypothesis? (This was given in the Central Questions subsection.)
- For the actual proposal, the claims here will need to be established through citations to existing background work.
- In full proposal, explain with citations that Google Translate does poorly in this domain. This helps establish the need to try ChatGPT or other AIs.
- People were asking about BLEU scores so in the proposal explain the issues with it.
- It's good that you have found something novel because research should have a novelty component. Another one is keep in mind is significance to the field.
- Clarify how these translated messages and originals are being treated confidentially per HIPAA and so on.
- Some comments after the meeting suggested that the prompts in the document were pretty rudimentary and maybe you could get better results with modest improvements in the prompts.
For next week:
- Updates, announcements, questions, etc.?
- Readings: We are reading
- Sparks of Artificial General Intelligence: Early experiments with GPT-4 (https://arxiv.org/abs/2303.12712).
- We will start with "Throughout the paper we emphasize limitations"
- Here is a Youtube video about the article: https://www.youtube.com/watch?v=qbIk7-JPB2c (found by IE)
- We have read chapter 1 part 4: https://huggingface.co/learn/nlp-course/chapter1/4 up to "Note that the first attention layer in a decoder block pays attention to" and weGPT
No comments:
Post a Comment