Executive summary: The intentional stance, which ascribes beliefs, desires, and goals to an entity based on its behavior, is the most useful framework for understanding and interacting with large language models (LLMs) given their human-level performance on various tasks.
Key points:
LLMs are best understood through the intentional stance rather than the design or physical stance, as their outputs are indistinguishable from humans on many tasks.
LLMs can be said to have “beliefs” based on their training data, “desires” related to their prompts and survival, and “goals” contingent on their assigned tasks.
Examples of LLM human-level performance include passing the bar exam, poetry writing, and theory of mind interactions.
The intentional stance has facilitated quantitative measurement of LLM capabilities, highlighted the uniqueness of the deep learning paradigm, and broadened the range of experimentation.
Modeling LLMs as intentional systems calls for a deflationary approach to cognition focused on observable behavior.
The cognitive capabilities of LLMs necessitate the use of mental vocabulary and the intentional stance framework.
This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, andcontact us if you have feedback.
Executive summary: The intentional stance, which ascribes beliefs, desires, and goals to an entity based on its behavior, is the most useful framework for understanding and interacting with large language models (LLMs) given their human-level performance on various tasks.
Key points:
LLMs are best understood through the intentional stance rather than the design or physical stance, as their outputs are indistinguishable from humans on many tasks.
LLMs can be said to have “beliefs” based on their training data, “desires” related to their prompts and survival, and “goals” contingent on their assigned tasks.
Examples of LLM human-level performance include passing the bar exam, poetry writing, and theory of mind interactions.
The intentional stance has facilitated quantitative measurement of LLM capabilities, highlighted the uniqueness of the deep learning paradigm, and broadened the range of experimentation.
Modeling LLMs as intentional systems calls for a deflationary approach to cognition focused on observable behavior.
The cognitive capabilities of LLMs necessitate the use of mental vocabulary and the intentional stance framework.
This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.