# ChatGPT-based Investment Portfolio Selection

**Authors:** Oleksandr Romanko, Akhilesh Narayan, Roy H. Kwon  
**arXiv:** [2308.06260](https://arxiv.org/abs/2308.06260)  
**Date:** 2023-08-11  
**Summary page:** https://www.summarizepaper.com/en/paper/2308.06260v1/

## Key points

- Authors explore the potential uses of generative AI models, specifically ChatGPT, for investment portfolio selection
- Trusting investment advice from GPT models can be challenging due to model "hallucinations"
- Authors propose an alternative approach using ChatGPT to obtain a universe of stocks from the S&P500 market index
- Various portfolio optimization strategies leveraging AI-generated trading universe are compared
- ChatGPT is effective in stock selection but may not perform well in assigning optimal weights to stocks within a portfolio
- Combining stock selection by ChatGPT with established portfolio optimization models leads to better results
- Hybrid approach combining AI-generated stock selection and quantitative optimization techniques can lead to more robust and favorable investment outcomes
- Careful validation and verification of outputs from generative AI models like ChatGPT is important to mitigate risks associated with model hallucinations.

## Summary

In the paper titled "ChatGPT-based Investment Portfolio Selection," authors Oleksandr Romanko, Akhilesh Narayan, and Roy H. Kwon explore the potential uses of generative AI models, specifically ChatGPT, for investment portfolio selection. They acknowledge that trusting investment advice from Generative Pre-Trained Transformer (GPT) models can be challenging due to model "hallucinations," which require careful verification and validation of the output. To address this challenge, the authors propose an alternative approach. They utilize ChatGPT to obtain a universe of stocks from the S&P500 market index that have the potential to be attractive for investing. The authors then compare various portfolio optimization strategies that leverage this AI-generated trading universe. They evaluate these strategies against quantitative portfolio optimization models and popular investment funds. The findings of their study indicate that ChatGPT is effective in stock selection but may not perform as well in assigning optimal weights to stocks within a portfolio. However, when combining stock selection by ChatGPT with established portfolio optimization models, better results are achieved. By blending the strengths of AI-generated stock selection with advanced quantitative optimization techniques, the authors observe the potential for more robust and favorable investment outcomes. This research suggests a hybrid approach for more effective and reliable investment decision-making in the future. It highlights how combining AI-generated stock selection with established portfolio optimization models can lead to improved investment outcomes. The authors emphasize the importance of carefully validating and verifying outputs from generative AI models like ChatGPT to mitigate any risks associated with model hallucinations. Overall, this study contributes valuable insights into leveraging generative AI models for investment portfolio selection while emphasizing the need for a cautious and blended approach in decision-making processes.