OpenAI has launched GPT-4, the latest version of its successful online chatbot ChatGPT. GPT-4 has the ability to recognize and explain images, allowing it to generate recipes based on the contents of a fridge.
The model also has improved text interpretation and generation capabilities, with the ability to process thousands more words at a time than its predecessor, GPT-3.
ChatGPT Plus subscribers currently have exclusive access to the new model.
According to Dr. Peter Van der Putten, director of AI at Pegasystems and a professor of AI at Leiden University, GPT-4 is capable of generating more elaborate and specific prose and demonstrates a deeper understanding of a problem than ChatGPT.
The new model has the ability to respond in specific styles, such as the Shakespearean pirate style seen in a demo.
OpenAI acknowledges that GPT-4 may still produce falsehoods, but safeguards have been improved.
GPT-4’s ability to generate text in various styles could lead to its use in creative fields, such as writing stories, poetry, and scripts in the voice of established creators.
It may also be used to make users sound more interesting on dating apps.
However, the improved specificity and reasoning abilities of the model may also increase the trust that people have in it.