# Sparks of Artificial General Intelligence: Early experiments with GPT-4

**Authors:** Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, Harsha Nori, Hamid Palangi, Marco Tulio Ribeiro, Yi Zhang  
**arXiv:** [2303.12712](https://arxiv.org/abs/2303.12712)  
**Date:** 2023-03-22  
**Summary page:** https://www.summarizepaper.com/en/paper/2303.12712v5/

## Key points

- Artificial intelligence (AI) has been advancing rapidly, with large language models (LLMs) at the forefront of progress.
- OpenAI's latest model, GPT-4, is a significant leap forward in AI capabilities.
- An early version of GPT-4 exhibits more general intelligence than previous AI models and can solve novel and difficult tasks spanning various fields without needing special prompting.
- GPT-4's performance is strikingly close to human-level performance and often surpasses prior models such as ChatGPT.
- GPT-4 could reasonably be viewed as an early but still incomplete version of an artificial general intelligence (AGI) system.
- The paper emphasizes discovering limitations and discusses challenges ahead for advancing towards deeper and more comprehensive versions of AGI.
- The societal influences resulting from recent technological leaps in AI research are reflected upon, along with future research directions that may require pursuing a new paradigm beyond next-word prediction to achieve true AGI.

## Summary

Artificial intelligence (AI) has been advancing rapidly, and large language models (LLMs) have been at the forefront of this progress. The latest model developed by OpenAI, GPT-4, is a significant leap forward in AI capabilities. In this paper titled "Sparks of Artificial General Intelligence: Early experiments with GPT-4," the authors report on their investigation of an early version of GPT-4 while it was still in active development by OpenAI. The authors contend that this early version of GPT-4 is part of a new cohort of LLMs that exhibit more general intelligence than previous AI models, including ChatGPT and Google's PaLM. They demonstrate that beyond its mastery of language, GPT-4 can solve novel and difficult tasks spanning mathematics, coding, vision, medicine, law, psychology and more without needing any special prompting. Moreover, in all these tasks, GPT-4's performance is strikingly close to human-level performance and often vastly surpasses prior models such as ChatGPT. The authors believe that given the breadth and depth of GPT-4's capabilities, it could reasonably be viewed as an early but still incomplete version of an artificial general intelligence (AGI) system. However, they also put special emphasis on discovering its limitations and discuss the challenges ahead for advancing towards deeper and more comprehensive versions of AGI. The paper concludes with reflections on societal influences resulting from recent technological leaps in AI research. It also highlights future research directions that may require pursuing a new paradigm beyond next-word prediction to achieve true AGI. Overall, this paper provides valuable insights into the current state-of-the-art in AI research with specific focus on LLMs like GPT-4. It sheds light on the potential implications for society as we move closer to achieving AGI systems while also acknowledging the challenges we face along the way.