March 14, 2023Science & TechnologyAmericas

GPT-4 and the Multimodal Era

OpenAI released GPT-4, a multimodal AI that could process images and text, pass professional exams, and reason at near-human levels across domains.

On March 14, 2023, OpenAI released GPT-4, a large multimodal model that could accept both text and image inputs. GPT-4 passed the bar exam in the 90th percentile, scored in the top ranges of the SAT, GRE, and AP exams, and demonstrated sophisticated reasoning across domains from law to medicine to coding. The model showed significant improvements in reliability, nuance, and the ability to follow complex instructions. Its vision capabilities allowed it to analyze charts, interpret handwriting, and describe images in detail. GPT-4 catalyzed a wave of AI integration across industries — from customer service to software development to scientific research — and convinced many that artificial general intelligence, once a fringe idea, might be achievable within years rather than decades.

More in Science & Technology

History, delivered weekly.

A curated dispatch of forgotten moments, pivotal turning points, and the stories behind the dates. No spam, just history.