The world is being quietly rearranged by people who write very long documents.


The title they went with The Bicameral Model: Bidirectional Hidden-State Coupling Between Parallel Language Models Noisy translates that to

AI models can now share internal thoughts to use tools with high accuracy


AI models can now communicate directly through their internal processing, not just by writing text. This means one AI can silently guide another to use tools like calculators or code, solving complex problems with much higher accuracy.
Until now, if an AI needed to use a tool, it had to write out its request in plain language, like talking to a human. This paper shows AI models can be built to share internal signals, making them much more efficient and reliable at using external tools. This changes how AI systems can be designed to tackle problems that require precise calculations or code execution, making them far more capable for specific tasks.
Watch for new AI systems that integrate this direct internal communication to achieve high accuracy on tasks that currently require complex prompting or human oversight, especially in areas like scientific research or engineering design.

If you insist
Read the original →