Social RoboticsOpenAI describes GPT-3, a 175-billion-parameter language model
A paper from more than 30 OpenAI researchers showed that a much larger model could tackle new tasks from a few examples.
OpenAI's GPT-3 paper described a model with 175 billion parameters, more than a hundred times larger than GPT-2, trained on nearly a trillion words of web text, books and Wikipedia. It could perform tasks such as translation and article writing from a handful of examples in the prompt, without retraining, though it still struggled with some common-sense reasoning and showed gender and racial bias in early tests. For social robotics, GPT-3 marked the arrival of fluent open-ended text generation, the capability that researchers would soon start wiring into robots for conversation.
Source: VentureBeat
This archive entry summarises reporting published elsewhere; follow the source link for the full story. The summary was drafted with AI from that reporting.