GPT-4 is a deep learning model developed by OpenAI that is capable of accepting both image and text inputs and emitting text outputs. It has achieved human-level performance on various professional and academic benchmarks, although it is less capable than humans in many real-world scenarios.
GPT-4 is a large multimodal model that has been trained on a vast corpus of data in order to learn to generate coherent and contextually appropriate text in response to various inputs.
It has been designed to be more reliable, creative, and able to handle much more nuanced instructions than its predecessor, GPT-3.5. GPT-4’s capabilities have been tested on a variety of benchmarks, including simulating exams that were originally designed for humans.
GPT-4 has also been evaluated on traditional benchmarks designed for machine learning models, where it considerably outperforms existing large language models alongside most state-of-the-art models.
GPT-4’s text input capability is currently available via ChatGPT and the API, with the image input capability being prepared for wider availability through collaboration with a single partner.
OpenAI has also open-sourced OpenAI Evals, a framework for automated evaluation of AI model performance, to allow anyone to report shortcomings in their models to help guide further improvements.