Strengths of Systems like ChatGPT (Still Requires Constant and Total Verification of EVERYTHING)
- Providing General Legal Information
- Explaining legal concepts
- Providing elements of a cause of action or defense
- Critiquing existing statutes or case law precedent
- Accurately summarizing prominent cases
- Comparing the laws of different jurisdictions
- Exploring legal history
- Answering Legal Questions
- Analyzing fact patterns, identifying and explaining relevant case law precedent, and applying the law to the facts (relevant case law is often accurately described, but case names and citations are sometimes made up)
- Analyzing fact patterns, identifying and explaining relevant statutory text, and applying the law to the facts (relevant statutory text is often accurately described, but statutory citation is sometimes wrong or made up)
- Assisting with Document Drafting
- Generating initial drafts of contracts, contract provisions, etc.
- Generating initial drafts of demand letters
- Generating initial drafts of motions, arguments for briefs, etc.
Weaknesses of Systems like ChatGPT
- Accuracy in General
- Answers by predicting the next word in a sentence based on its training data, so you can't trust anything it says
- Accuracy of Primary Materials Citations
- Often accurately describes relevant statutory text but can't provide correct citations to the statutes discussed
- Often accurately describes case law precedent in a jurisdiction and correctly applies it to facts but can't provide correct citations/case names to the statutes discussed
- Accuracy of Text of Primary Materials
- Generally unable to accurately provide the text of a statute or regulation
- Generally unable to accurately provide the text of a case
- Currentness
- Cut-off date of training data is September 2021
- It can't search the internet, so it is unaware of any new statutes, statutory amendments, new cases, or negative treatment of statutes or cases
