Large Language Models as Philosophical Inquiry
Lead Research Organisation:
UNIVERSITY OF READING
Abstract
Large Language Model (LLM) technology has been at the heart of recent advances in artificial intelligence, including chatbots such as Chat-GPT and Claude, and many other applications besides. But should the success of this technology change the way we think about language itself? This project conducts a philosophical investigation into whether LLMs reveal fundamental insights about language. While some argue that LLMs can function as descriptive and predictive accounts of language comparable to those developed by linguists, others have claimed that they provide no insight into human language whatsoever. This project will develop a new middle ground: that LLMs can serve as scientific models of public languages. This “modelling approach” requires us to move away from a view of linguistic study as solely concerned with linguistic cognition and instead recognise a notion of language understood as the set of public conventions that unite a linguistic community.
The project has three aims, corresponding to three work packages:
Develop the “modelling approach”: a new philosophical approach to LLMs that treats them as scientific models of public languages.
Conduct two model analysis studies that analyse how LLMs process specific linguistic phenomena in order to further linguistic and philosophical inquiry.
Host an interdisciplinary conference and publish a special issue of Ratio, bringing together speakers from across academia and industry to explore the relationship between LLM technology and the nature of language.
The research will be interdisciplinary in nature, combining philosophical analysis informed by recent debates in philosophy of language and philosophy of science with empirical investigations of language model technology that makes use of recent innovations in explainable AI. As such, the project will establish links between philosophy, computational linguistics, and AI research that have until now been underexplored. The development and application of the “modelling approach” will not only provide new insight into the nature of language, but introduce new methodologies to philosophical investigation.
The project has considerable impact and engagement potential within AI industries. As companies seek to put LLM technology to use in various commercial applications, they will benefit from a greater understanding of how such technologies provide insight into the nature of language. Engaging with stakeholders from industry is an important part of the project, ensuring that the modelling approach is informed by the practical business of language model construction, and that the insights gained from the approach can then be fed back into industry.
The research is extremely timely given the rapid advancements in LLM technology and the ongoing debates about its significance for understanding language. LLM technologies are fast becoming embedded into many aspects of life, yet our understanding of the insight that the technology holds is still relatively limited. This project would seek to address that shortcoming.
The project has three aims, corresponding to three work packages:
Develop the “modelling approach”: a new philosophical approach to LLMs that treats them as scientific models of public languages.
Conduct two model analysis studies that analyse how LLMs process specific linguistic phenomena in order to further linguistic and philosophical inquiry.
Host an interdisciplinary conference and publish a special issue of Ratio, bringing together speakers from across academia and industry to explore the relationship between LLM technology and the nature of language.
The research will be interdisciplinary in nature, combining philosophical analysis informed by recent debates in philosophy of language and philosophy of science with empirical investigations of language model technology that makes use of recent innovations in explainable AI. As such, the project will establish links between philosophy, computational linguistics, and AI research that have until now been underexplored. The development and application of the “modelling approach” will not only provide new insight into the nature of language, but introduce new methodologies to philosophical investigation.
The project has considerable impact and engagement potential within AI industries. As companies seek to put LLM technology to use in various commercial applications, they will benefit from a greater understanding of how such technologies provide insight into the nature of language. Engaging with stakeholders from industry is an important part of the project, ensuring that the modelling approach is informed by the practical business of language model construction, and that the insights gained from the approach can then be fed back into industry.
The research is extremely timely given the rapid advancements in LLM technology and the ongoing debates about its significance for understanding language. LLM technologies are fast becoming embedded into many aspects of life, yet our understanding of the insight that the technology holds is still relatively limited. This project would seek to address that shortcoming.
Organisations
People |
ORCID iD |
| Jumbly Grindrod (Principal Investigator) |