Synthetic Creativity - by Cavin - Deep Markets

About Synthetic Creativity - by Cavin - Deep Markets

Published on Deep Markets on Oct 3, 2022, 'Synthetic Creativity' by Cavin examines the proliferation of machine-generated content. It places machine learning in a sequence with writing, the printing press, the internet and digital tools as ways to reduce the cost of creating and distributing information.

The excerpt focuses on synthetic text, describing OpenAI's GPT-3 and its few-shot performance without fine-tuning, and uses such as brainstorming, short-form sales and marketing copy, summarization, translation and transcription. It states that current models are not yet capable of sufficiently cohesive long-form stories.

Key features

• Traces idea propagation from writing to printing press to internet
• Describes GPT-3 as 175-billion-parameter autoregressive language model
• Explains few-shot performance without task-specific fine-tuning
• Discusses synthetic text for stories, copywriting and language tasks
• Lists summarization, translation and transcription as horizontal uses
• Notes vertical assistants for specific workflows and industry styles

Synthetic Creativity - by Cavin - Deep Markets at a glance