ChatGPT's launch and the sudden public encounter with generative AI

A research preview becomes a phenomenon

OpenAI published its blog post introducing ChatGPT on 30 November 2022, describing a model trained to interact in a conversational way, released as a free preview to gather feedback. Unlike prior demonstrations of large language models, which had largely remained within research and developer communities, ChatGPT was placed directly in front of the public through a simple chat interface, with no technical setup required.

The reaction was immediate and broad. Within days the service was reportedly handling over a million users, and it went on to become one of the fastest-growing consumer applications in history. Its fluency at drafting essays, summarising documents, writing code and holding sustained conversation gave millions of people their first hands-on encounter with a large language model, often with little context for its limitations.

“Millions of people got their first hands-on encounter with a large language model, often with little context for its limitations.”

Fluency without reliable grounding

ChatGPT's most consequential characteristic was one that took the wider public longer to appreciate than technologists: its confident, well-formed answers were not reliably tied to verified fact. The model could, and regularly did, generate plausible-sounding but false information, a behaviour that quickly acquired the informal name 'hallucination'. Educational institutions, newsrooms and employers found themselves reacting within weeks to a tool whose output was fluent enough to pass as human-authored and unreliable enough to require verification every time.

The launch also intensified an existing debate over academic integrity, as students began using the tool to draft assignments, and over authorship and disclosure more broadly — questions that outpaced the policy guidance most institutions had in place.

A profession suddenly in demand

For the computing profession, ChatGPT's launch converted an abstract debate about AI safety and capability into an urgent, public one. Employers, regulators and educators looked to technologists for guidance on where the technology was reliable, where it was not, and how it should be governed, often faster than professional bodies or standards organisations could formally respond.

• Explain model limitations, including hallucination risk, in plain terms to non-technical stakeholders.
• Establish organisational policy on acceptable use of generative AI in professional work product.
• Verify factual claims from generative AI outputs before they reach clients, students or the public.
• Support digital literacy efforts that help the public understand what these tools do and do not know.

What the Society advises

The Society regards the weeks following ChatGPT's release as a demonstration of why technical literacy is now a matter of broad public interest, not a specialist concern. Members are advised to apply the Society's standards of accuracy and honesty to any professional use of generative AI tools, to disclose their use where material, and to pursue CPD that keeps pace with a technology now embedded in daily public life.