OpenAI has unveiled GPT-4, the next iteration of its artificial intelligence tool which led to the creation of the popular ChatGPT and Dall-E image-creation programme. The company claims that GPT-4 is more precise, innovative and collaborative than its predecessor, GPT-3.5, and is 40% more likely to generate factual responses.
OpenAI’s ChatGPT-3, released in 2020, and the 3.5 version were used to develop ChatGPT and Dall-E, two widely acclaimed products that triggered a surge of interest in AI. OpenAI’s customers for GPT-4 include Morgan Stanley, Stripe, Duolingo, Khan Academy and the Icelandic government. The release is part of a wave of AI-related announcements from OpenAI, Microsoft, Google and other industry players. While the technology has generated excitement, concerns remain about its ability to perpetuate biases and misinformation. OpenAI has worked for six months to make the AI software safer, but acknowledges that it still has limitations such as social biases and adversarial prompts.