[Paper Review] One Small Step for Generative AI, One Giant Leap for AGI: A Complete Survey on ChatGPT in AIGC Era

Chaoning Zhang, Chenshuang Zhang+14 more|arXiv (Cornell University)|Apr 4, 2023

Artificial Intelligence in Healthcare and EducationMedicine89 citations

TL;DR

A comprehensive survey of ChatGPT in the AIGC era, covering OpenAI, GPT history, core technologies, applications, challenges, and future outlook toward AGI.

ABSTRACT

OpenAI has recently released GPT-4 (a.k.a. ChatGPT plus), which is demonstrated to be one small step for generative AI (GAI), but one giant leap for artificial general intelligence (AGI). Since its official release in November 2022, ChatGPT has quickly attracted numerous users with extensive media coverage. Such unprecedented attention has also motivated numerous researchers to investigate ChatGPT from various aspects. According to Google scholar, there are more than 500 articles with ChatGPT in their titles or mentioning it in their abstracts. Considering this, a review is urgently needed, and our work fills this gap. Overall, this work is the first to survey ChatGPT with a comprehensive review of its underlying technology, applications, and challenges. Moreover, we present an outlook on how ChatGPT might evolve to realize general-purpose AIGC (a.k.a. AI-generated content), which will be a significant milestone for the development of AGI.

Motivation & Objective

• Introduce OpenAI and the context of ChatGPT within the AGI/AIGC landscape.
• Summarize the technology and core techniques underpinning ChatGPT (Transformer, autoregressive modeling).
• Review the applications of ChatGPT across scientific writing, education, medicine, and other fields.
• Identify challenges, including technical limitations, misuse, ethics, and regulation.
• Discuss future directions toward general-purpose AIGC and AGI.

Proposed method

• Synthesize and organize existing literature on ChatGPT up to GPT-4.
• Explain the Transformer backbone and autoregressive modeling with formal definitions.
• Trace the GPT model lineage (GPT-1 to GPT-4) and key training/data details.
• Categorize applications by domain (scientific writing, education, medicine, others).
• Highlight challenges and regulatory/ethical considerations and provide an AGI-oriented outlook.

Experimental results

Research questions

• RQ1 What are the foundational technologies and training paradigms that enable ChatGPT (Transformer, autoregressive modeling)?
• RQ2 How has ChatGPT evolved from GPT-1 to GPT-4 and what are its capabilities and limitations?
• RQ3 What are the main applications of ChatGPT across domains such as scientific writing, education, and healthcare?
• RQ4 What challenges, misuse risks, ethical concerns, and regulatory issues accompany ChatGPT’s deployment?
• RQ5 What is the envisioned trajectory for ChatGPT toward general-purpose AIGC and AGI?

Key findings

• ChatGPT (GPT-4-based) supports multimodal inputs (text and images) and achieves high performance on professional tasks, with human-level abilities in several standards.
• GPT model progression (GPT-1 to GPT-4) shows dramatic increases in parameters, data, and capabilities, including zero-shot/few-shot performance improvements and RLHF for GPT-3.5.
• Transformer and autoregressive modeling underpin contemporary LLMs; ChatGPT leverages self-attention and decoder-only architectures for fluent, context-aware generation.
• Applications span scientific writing, data analysis, literature review, content generation, proofreading, and academic peer-review roles, with varying effectiveness and caveats.
• Challenges include technical limitations, potential misuse, ethical concerns, and regulatory considerations that shape deployment and governance.
• The survey outlines an outlook toward general-purpose AIGC and the broader goal of achieving AGI beyond ChatGPT’s current capabilities.