OpenAI launched its first commercial product, an API giving developers access to models from the GPT-3 family. GPT-3, described in a research paper the previous month, had 175 billion parameters, more than a hundred times as many as GPT-2.
What happened
- The API offered a general-purpose ‘text in, text out’ interface that could be applied to almost any English language task.
- Access began as a private beta, with applications reviewed before going into production.
- OpenAI said it would terminate access for harmful uses such as harassment, spam, and deliberate deception.
- In September 2020 Microsoft announced an exclusive licence to the underlying GPT-3 technology.
Why it mattered
Demonstrations built on the API showed GPT-3 writing essays, code, and conversation, marking the start of large language models being offered as a commercial service. It also set the pattern of offering powerful models through controlled online access rather than releasing them openly, an approach many AI developers later followed.
Lessons for organisations
Before adopting third-party AI services, check how prompts and outputs are stored and used, and set rules on what data staff may share with them. Treat AI outputs as drafts that need human review before they are relied on.
Source: OpenAI
Part of our Top stories archive of headline-making events in information security, privacy, and AI. If you would like help applying the lessons to your organisation, contact us.
