GPT-4 is an OpenAI large language model released in March 2023. Unlike its predecessor, it accepts both text and image inputs, making it a multimodal model rather than a text-only system. Its capabilities span language, coding, examination questions, and practical administrative writing. section 1.16 section 1.7
GPT-4 can produce useful answers and drafts, but strong performance on selected tests does not make it a dependable professional replacement. It performs relatively well on easy LeetCode problems while struggling with medium and hard problems, and generating code differs from operating successfully as an engineer. section 1.9 section 1.7
Its demonstrated uses include:
- Answering academic and professional examination questions, including the Uniform Bar Examination, LSAT, SAT, GRE, and advanced placement tests; it achieved human-level performance on most listed exams and placed in the top 10% on the bar examination. section 1.9
- Drafting healthcare administration, including an insurer appeal letter, patient email responses, common-question replies, paperwork, and an accessible script about alcohol use disorder. section 6.2
- Assisting with programming problems, while leaving a meaningful distinction between solving an isolated challenge and handling an entire software project. section 1.7
A healthcare administration example
A rheumatologist used GPT-4 to draft an appeal after an insurer denied coverage for a drug. The insurer granted the request, reportedly saving the patient $1,500 per month, while the practice also used the model for email replies and paperwork. Such results illustrate practical leverage, not guaranteed accuracy or independent judgment. section 6.2
