What is GPT-4?

- [Instructor] We've defined a language model as something that takes in text input, or a prompt, and outputs text data. Now, this was the case for GPT 3 and GPT-3.5, but now it turns out that GPT-4 is multimodal. This means it can take in input from more than one source, so that's either text, or text and image data, and it'll output text. As of this recording, the option to take in input from text and image data has not been yet released. Let's look at a couple of examples of the capability of the text-only model. OpenAI is the research company that created the GPT-4 model. The OpenAI researchers wanted to demonstrate how capable the GPT-4 model was compared to humans. They tested GPT-4 on some high school exams, like the SAT and Advanced Placement, along with some postgraduate exams, like the GRE. Now, these exams that are used regularly in the US. They also tested it on some professional exams, like the Bar exam,…