Social RoboticsOpenAI releases GPT-4, a model that accepts images as well as text
The successor to GPT-3 became available to paying ChatGPT users and could process far longer passages of text.
OpenAI launched GPT-4 on 14 March 2023, describing it as its most capable model while admitting it still had flaws. It accepts image and text input and produces text, and the company said it could handle up to about 25,000 words at once. Access initially required a paid ChatGPT Plus subscription, and image input was not yet open to the public. A stronger general-purpose model with visual understanding made richer conversation and scene description more practical for robot developers building on commercial APIs.
Source: Al Jazeera
This archive entry summarises reporting published elsewhere; follow the source link for the full story. The summary was drafted with AI from that reporting.