Mirco Sambrotta: If God looked into AI systems, would he be able to see there whom they are speaking of?

Organized by the Department of Logic

Further information

Invitation pdf

Mirco Sambrotta (Slovak Academy of Sciences):
If God looked into AI systems, would he be able to see there whom they are speaking of?

Abstract
Can Large Language Models (LLMs), such as ChatGPT, be held accountable for their language production? Affirmative answers hinge on recognizing them as genuine language users, capable of mastering the use of words and sentences through adherence to inferential rules. However, implementing or algorithmically programming the ability to follow such rules is impractical; it can only be acquired through training that transcends mere formalism. Therefore, LLMs can only be considered authentic language users and be held responsible for their outputs to the extent that they are trained to adhere to inferential rules. At the same time, though, LLMs can be trained to follow inferential rules only to the extent that they are held accountable for their results. This interdependence underscores the nuanced nature of their accountability in language production.