Paper
- Ringel, Daniel, Creating Synthetic Experts with Generative AI (July 15, 2023). Available at SSRN: https://ssrn.com/abstract=4542949.
- Appendix: Details, Notes, Parameters, IRB information
How to Create your own Synthetic Expert using Generative AI
This website provides supporting materials, code, data, and tutorials for the paper Creating Synthetic Experts with Generative AI by Daniel M. Ringel (2023).
News
[November 15, 2023] Watch my Talk at the efl Annual Conference in Frankfurt: I will introduce Syntehtic Experts in the context of Opportunities and Challenges of Generative AI. Program and Free Tickets
[October 19, 2023] Recording of Wharton AI Talk: My talk on Synthetic Experts is now available on YouTube
[September 28, 2023] Top 10 at SSRN: Synthetic Experts rank #1 in Top 10 downloads of recent papers in Marketing Science eJournal Ranking
[September 8, 2023] Presentation of Synthetic Experts at the Wharton Business & Generative AI Workshop in San Fransico. Check out the PDF presentation slides!
[September 6, 2023] Anonymize texts with Synthetic Twins: Python notebook now available on GitHub
[August 26, 2023] Create your own Synthetic Expert with new Python notebooks available on GitHub. Label text with generative AI and fine-tune LLMs for your purpose.
[August 23, 2023] Work with several demo Datasets to explore the use and creation of Synthetic Experts.
[August 19, 2023] Easily apply the MMX Synthetic Expert of this research to YOUR data. New Python notebook available on GitHub.
[August 16, 2023] This research is currently being revised and extendet. I will make materials available as they are completed.
Code
- Python Notebooks: GitHub Repo
- Functions: UseSynExp.py
Supporting Documents
Tutorials
- Get ready to fine-tune LLMs [Setup_Python_GPU.ipynb]
- Query the OpenAI API [Query_OpenAI_API.ipynb]
- Parse OpenAI API responses [Parsing_API_Responses.ipynb]
- Fine-tune a pre-trained LLM from Huggingface [Fine-tuning_LLMs.ipynb]
- Hyperparametertuning with Optuna [Hyperparametertuning.ipynb]
- Deploy your Synthetic Expert and use it for Inference [ipynb]
Repository
Code will be maintained on GitHub
Fine-Tuned Model
Marketing Mix Classifier on Huggingface's Model Hub
Questions, Comments?
Get in touch at dmr@unc.edu