Wang Kai Collection / 2023 AIGC Industry Report Archive

A large collection of AIGC industry reports amassed by Wang Kai in 2023.This batch no longer represents current best practices(from a 2026 perspective, the tool stack and judgments are outdated), but as"a collection of the industry's immediate reactions when the ChatGPT moment occurred"it has value as a historical record of the era.

Timeframe: ChatGPT launch in 2022-11 → full year of 2023. Over 20 PDFs in total.

I. Foundational English-language Papers

Sparks of Artificial General Intelligence (Microsoft Research, 2023-03)

Authors: Sébastien Bubeck et al., 17 people, Microsoft Research
arXiv:2303.12712(2023-03-22)
Core thesis: GPT-4 displays "sparks" of AGI — the early, incomplete version demonstrated cross-domain general capabilities
Impact:The concept of "AGI sparks" originated from this paper, and every industry report in the subsequent 12 months cited it
Wang Kai's application:Wang Kai's AI industry judgments and observations on large-company competitive landscape core evidence for the "L2 → L3" transition in the 5-level AI framework

Azure OpenAI Service Documentation (Microsoft, 2023)

Full set of GPT-3 / Codex / Embeddings model integration documentation
Complete Quickstart / How-to / Tutorial / Reference materials
Wang Kai's application: Compare with later OpenRouter / Claude API to observe business model evolution

II. Chinese Official / Industry-level Research Reports

AIGC White Paper (2022) — CAICT + JD Exploration Research Institute

Publication date: 2022-09 (two months before ChatGPT launched!)
Nature:Forward-looking national-level think tank research — predicting that AIGC would soon become a major technology direction
Implication: Chinese institutions were in fact systematically tracking AIGC as early as mid-2022; ChatGPT only accelerated public awareness
Content: Technology development history / international and domestic landscape / legal risks / application scenarios

ChatGPT Research Framework (2023-02) — Industry Special Research

A comprehensive industry perspective on "how ChatGPT came to be / how it evolved / commercialization prospects"
ChatGPT's daily active user growth outpaced Instagram, with 13 million average daily unique visits in January
Giants entering the field: Baidu / Tencent / Google / Microsoft issued responses
An overview of the evolution from Transformer to GPT 1/2/3 to ChatGPT

AIGC Applications / Commercialization and Beneficiary Targets (2023-02) — Huaxi Securities Media & Internet

Securities analyst Zhao Lin (S1120520040003)
Core viewpoints:Commercialization conditions for both B-end and C-end are mature, with strong near-term implementation potential
A complete industry chain of "data supply → model customization → content distribution" has formed
Industry logic resembles fast-moving consumer goods / fragmented entertainment / content production methods are being transformed

Domestic and International Large Model Evaluations (2023) — Wenxin vs GPT

Test dimensions: Common-sense creation / inductive reasoning / mathematics and code / application scenarios
Conclusion: The GPT series leads in deductive reasoning / logical reasoning, while Wenxin performs slightly better in inductive summarization tasks
All three major models still have room for improvement in emotional reasoning
Wang Kai's application:Wang Kai's AI industry judgments and observations on large-company competitive landscape foundation for "Chinese vendor progress assessment" alongside OpenAI/Anthropic/Google evolution

AIGC / ChatGPT Insurance Industry Application White Paper (2023-05) — ZhongAn FinTech Research Institute

Theme: GenAI implementation scenarios in the insurance industry
Regulatory preparation / legal risks / industry institutional development
Wang Kai's application: Used as a reference for "vertical industry AI applications" (insurance = a highly regulated traditional industry)

State of Generative AI 2023 — Qiming Venture Partners × Weijin Research

Joint report by VC + media research
"Integration of research and entrepreneurship" judgment — US giants + startups driven by dual engines
GPT-4 artificial general intelligence "sparks"

AI + New Era of Marketing (2023-06) — Minglue Technology + School of Management, Fudan University + Miaozhen Marketing Academy

Perspective: AI implementation in the marketing industry
China's digital economy foundation → 1999/2020 GDP comparison → AI driving the next two decades
Expert consensus on "disrupting productivity" / "improving efficiency" / "innovating experience" / "freeing up manpower"

2023 AIGC Report — Aifenxi

Huang Yong / Wen Hongwei et al.
Industry analyst B2B perspective

The Era Transformation Triggered by ChatGPT (2023-06-26) — Colliers East China

Perspective of a real estate research institution
AIGC industry chain + Shanghai "4+X" base layout
Prosperity of computing power / chips / high-bandwidth networks

III. Introductory / Textbook-level ChatGPT Introductions

ChatGPT From Beginner to Master (2023, 116 pages)

All content in this book was completed by ChatGPT — serving as a demo in itself
Basic knowledge / implementation methods / advanced applications / latest progress

The Past and Present of ChatGPT (2023-04-11, Li Piji)

Author: Li Piji, Nanjing University of Aeronautics and Astronautics (former Baidu / Tencent AI Lab engineer)
In PPT format, popular science on natural language processing
Background: PhD from The Chinese University of Hong Kong / Master's from Shandong University

Introduction to Instruction Tuned LLMs

Distinguishing LLM vs Instruction Tuned LLM
Principles of instruction fine-tuning based on RLHF
When GPT-3 is insufficient / GPT-4 already includes Instruction Tuning optimizations

Why ChatGPT Is a Productivity Revolution — Fu Sheng (Cheetah Mobile / OrionStar)

Early Chinese entrepreneur-level speech
Influence: Fu Sheng was one of the earliest domestic technology leaders to make high-profile statements
Time anchor: The industry's "shock period" after the 2023 Spring Festival

IV. In-depth Special Reports

Commercial Content Value Discovery (2023)

Strategy / industry discourse / brand stories / product technology / market / business models
Angle: Viewing the AI transformation of commercial content from a traditional news perspective

AIGC 2023 (2 different reports)

Marked general GAN / Confidential
Thematic expansion (technology / business / pragmatic)

V. Unreadable PDFs (7 files)

Text extraction failed for the following PDFs (possibly scanned images / severe OCR defects / encoding issues); handled per "archive original files, wait for user to decide whether to OCR":

File	Size	Status
file_005	20MB	Completely empty
file_006	2.1MB	Completely empty
file_015	40MB	Completely empty
file_017	5.8MB	Completely empty
file_018	1.5MB	Completely empty
file_021	5.0MB	Severe garbled text
file_022	3.7MB	Severe garbled text (similar to 021)

See ocr-review notes.

Historical Value Assessment

Viewing 2023 reports from a 2026-05 perspective:

Most technical details are outdated (more comprehensively covered by the Foundation Models for NLP textbook Foundation Models for NLP (Springer textbook close reading))
But high value as a "historical record of the era" — observing how the industry reacted to the ChatGPT shock
Seeing which predictions came true and which were wrong (retro-back testing the quality of industry judgments)

Correct predictions

AIGC commercialization conditions are mature → already confirmed ($87 billion in GenAI financing, 191 new unicorns; see details in AI Agent Market Investment Trends 2025 (Data Report))
Writing + content production implemented first → confirmed ("Writing accounts for the majority of work tasks"; see details in Wang Kai Collection / 2024-2026 AI Frontier Papers and Reports Close Reading How People Use ChatGPT)
Computing power + chips are key bottlenecks → confirmed (NVDA's 10-year logic of accelerated computing, Aschenbrenner's trillion-dollar cluster prediction)

Incorrect predictions

"B-end near-term implementation" was significantly overestimated — real data (SOTA data 2025-09): work users account for only 30%, the majority is personal use
"Domestic large models will catch up in 1–2 years" was significantly overly optimistic — in reality DeepSeek was the real breakthrough for Chinese large models, and that came at the end of 2024
"AIGC will replace search engines" was partially misjudged — the reality is coexistence with search engines, not replacement