ChatGPT Turns 3 (restofworld.org)

ChatGPT turned three on November 30, 2025 — a brief era that saw rapid mainstreaming of large language models. Since its 2022 launch, OpenAI’s chatbot has attracted roughly 800 million weekly users, supports more than 20 languages, and is especially popular outside the U.S. in markets such as India, Brazil, Indonesia and the Philippines. Its adoption has reshaped workflows (e.g., judges in Colombia using ChatGPT/Copilot in 85% of cases), education, healthcare (robot companions in South Korea), and political messaging — including documented election manipulation like 170 fake X accounts in Ghana — underscoring both productivity gains and new social risks. Technically, ChatGPT exposed strengths and limits of current LLMs. It excels in high-resource languages but produces fabrications, nonsense, and safety failures in underrepresented tongues (Bengali, Swahili, Urdu, Thai), driven by sparse training data and moderation gaps. Those shortcomings have spurred a wave of localized GPT-like models and tools — from ITanong and Latam-GPT to Nigeria’s Awarri — and alternative entrants in China (Qwen, DeepSeek) that claim competitive performance at lower cost. Practically, this means more API-driven integration, growing reliance on local fine-tuning and data curation, and heightened focus on multilingual robustness, provenance, and governance as the community moves from novelty to infrastructure.