February 26, 2024Philosophical Transactions of the Royal Society A Mathematical Physical and Engineering SciencesOpen Access

Large language models as tax attorneys: a case study in legal capabilities emergence

View Full Paper

Ask AI

Bookmark

Authors

JNJohn J. NayHarvard University DKDavid KaramardianStanford University SLSarah B. LawskyIllinois College

Discussion

Member takes

Implication

Evaluation demonstrates emerging legal reasoning capabilities in large language models, indicating high accuracy under few-shot prompting without matching expert tax lawyers.

Key Points

Large language models demonstrate emerging legal reasoning in tax law, achieving high accuracy with relevant authority without matching expert lawyers.
Evaluation across thousands of tax examples shows that few-shot prompting and retrieval of legal authority significantly boost GPT-4 performance.
Automated testing pipelines highlight that emerging artificial intelligence capabilities could transform legal services and future AI governance.

Cite This Study

Nay et al. (2024) studied this question.

synapsesocial.com/papers/68e777acb6db6435876ecae8 https://doi.org/10.1098/rsta.2023.0159

View Full Paper

Ask AI

Bookmark

Implication

Evaluation demonstrates emerging legal reasoning capabilities in large language models, indicating high accuracy under few-shot prompting without matching expert tax lawyers.

Large language models as tax attorneys: a case study in legal capabilities emergence

Discussion

Member takes

Implication

Key Points

Cite This Study

Large language models as tax attorneys: a case study in legal capabilities emergence

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider