OpenAI released GPT-4, a large language model that it said performed at human level on various professional and academic benchmarks. The model accepted image as well as text inputs.
What happened
- OpenAI said GPT-4 scored around the top 10% of test takers on a simulated bar exam, compared with the bottom 10% for its predecessor.
- It was made available to paying ChatGPT Plus subscribers and to developers through a waitlist for its API.
- Microsoft confirmed that its new Bing chat had been running on GPT-4.
- OpenAI published limited technical detail, citing competition and safety, which drew criticism from researchers.
Why it mattered
GPT-4 raised the bar for AI capability and accelerated adoption in business, while also intensifying concerns about misinformation, security misuse, and the pace of AI development. Within weeks, an open letter organised by the Future of Life Institute called for a six-month pause on training systems more powerful than GPT-4.
Lessons for organisations
Organisations adopting large language models should assess risks such as data leakage, inaccurate output, and prompt injection before deployment, and keep a human in the loop for important decisions. Contracts with AI providers should cover how prompts and data are stored and used.
Source: TechCrunch
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.
