Domenic Denicola • 2/19/2023

ChatGPT Is Not a Blurry JPEG of the Web

This article critiques Ted Chiang's analogy that LLMs are a 'blurry JPEG of the web,' arguing it's misleading. It proposes a better analogy: LLMs are simulators that instantiate temporary simulacra (like ChatGPT's 'assistant') based on prompts. This framework explains model behavior, jailbreaks, and capabilities without attributing false agency.

0 comments

#artificial intelligence #large language models #simulation