Rohan Anil on Opaque Chain-of-Thought as Bytestream
Rohan Anil replies that chain-of-thought in opaque models is a bytestream of the internal reasoning.
TLDR
In a reply to @sjgadler and @CFGeek, Rohan Anil states that chain-of-thought outputs from opaque models amount to a bytestream of the underlying opaque reasoning. The post comes from @arohan, identified as a research engineer and former Google DeepMind engineer. No further posts or confirmations appear in the packet. The comment addresses distinctions in how different AI reasoning processes can be described.
Combined views
97
1 Source, first seen 31d ago