John Bliss

I started a blog providing up-to-date analysis of research on legal AI--did GPT-4 ace law exams? Is legal AI hallucinating as much as reported? This is a fast-moving field, and too often our discussions are based on outdated or inaccurate info. Subscribe: ai-lawyering.blog

@IsaiahPoritz A limitation: they studied GPT 3.5, which scores below the 1st percentile on UBE, not GPT 4, which scores in the 68th percentile and is integrated in legal AI apps (Lexis, Co-Counsel, start-ups) that reduce hallucinations. The authors used GPT-4 to identify 3.5's hallucinations.

We all know the stories of lawyers using ChatGPT to draft hallucination-ridden briefs. But just how bad is it? A new Stanford study has the numbers: top AI models hallucinated 75% of the time when answering questions about a court’s core ruling. news.bloomberglaw.com/business-and-p…

@ssballakrishnen @dsandomierski @DeucesTecum @HLS_CLP @bryonfong @ProfQuintanilla @ProfMarkovic @KevinTFrazier @Christophkw @dsfarbman @NoahRFeldman @beidelson @ProfAdediranT @CohenProf @danepps @CassSunstein @cwschmidt1 @DavidLuban @JonathanHChoi @Dschwarcz

@ssballakrishnen @dsandomierski @DeucesTecum @HLS_CLP @bryonfong @ProfQuintanilla @ProfMarkovic @KevinTFrazier @Christophkw @dsfarbman @NoahRFeldman @beidelson @ProfAdediranT @CohenProf @danepps @CassSunstein @cwschmidt1 @DavidLuban

Law teachers planning spring courses: How should we adapt to advances in generative AI, e.g. GPT-4 acing law exams with the best-performing prompts (Choi & Schwarcz) and students now gaining access to Lexis+ AI. I just posted an article on this. (1/4)

I draw from surveys of students and faculty, a 1L pedagogical experiment, and the new educational literature on teaching with generative AI. See papers.ssrn.com/sol3/papers.cf…, forthcoming with Jurimetrics. (4/4)

· What considerations should we weigh as the technology continues to develop? E.g., emerging professional competencies, legal ethics, potential improvements in legal learning, equity concerns, academic integrity? (3/4)

· How should we approach assessments? E.g., stop giving take-home exams? AI-proofing exam questions? Detailed AI citation requirements? Implementing AI bans? · Should we integrate this tech in our teaching? What would this look like? (I suggest exercises and assignments) (2/4)

An @LSI_Journal article by @ProfJBliss, assistant prof @SturmCOL, examines how globalization has effected the civic role of lawyers. His work focuses on lawyers' experience with civic professionalism in the U.S and China and how these differ dramatically ow.ly/D3pk50DTjaK

Check out the latest blog from The Good Project where guest contributor @ProfJBliss discusses teaching good work in law and how to help students will sustain "public-interest values" regardless of their career path. buff.ly/2Z1aYuM

Thanks for the welcome Sam. So excited to join you at @SturmCOL!