# Sparks of Artificial General Recommender (AGR): Early Experiments with ChatGPT

**Authors:** Guo Lin, Yongfeng Zhang  
**arXiv:** [2305.04518](https://arxiv.org/abs/2305.04518)  
**Date:** 2023-05-08  
**Summary page:** https://www.summarizepaper.com/en/paper/2305.04518v1/

## Key points

- Study explores feasibility of developing an Artificial General Recommender (AGR) using Large Language Models (LLMs)
- Ten fundamental principles proposed for an AGR, along with testing protocols
- Evaluation of ChatGPT as a potential AGR through recommendation-oriented dialogues
- Findings demonstrate ChatGPT's potential as an AGR but identify limitations and areas for improvement
- Emphasis on further research to address limitations and enhance ChatGPT's performance as an AGR
- Study contributes to understanding conversational recommender systems using advanced language models

## Summary

This study by Guo Lin and Yongfeng Zhang explores the feasibility of developing an Artificial General Recommender (AGR) using Large Language Models (LLMs). An AGR is designed to engage in natural dialogues and generate recommendations across different domains. The researchers propose ten fundamental principles that an AGR should adhere to, along with corresponding testing protocols. They then evaluate whether ChatGPT, a sophisticated LLM, can meet these principles by engaging in recommendation-oriented dialogues with the model and observing its behavior. The findings of the study demonstrate that ChatGPT has the potential to serve as an AGR. However, several limitations and areas for improvement are identified. The researchers emphasize the need for further research to address these limitations and enhance the performance of ChatGPT as an AGR. Overall, this study contributes to our understanding of developing conversational recommender systems using advanced language models. It provides valuable insights into the capabilities and challenges associated with utilizing LLMs like ChatGPT for generating recommendations through natural dialogue interactions.