Roundblock转发了
Randomized trial AI for legal work finds Reasoning models are a big deal for lawyers: Law students using o1-preview (the first available reasoner) had the quality of their work on most tasks increase (up to 28%) & time savings of 12-28%. There were a few hallucinations, but a RAG-based AI with access to legal material (Vincent) reduced those to human level. Combining both with be the future. Big changes to law appear to be coming: "Our findings demonstrate that reasoning models improve not only the clarity, organization, and professionalism of legal work but also the depth and rigor of legal analysis itself."
One thing that strikes me is how naturally AI and the legal profession fit together—not just because AI improves efficiency, but because lawyers are already trained in structured questioning and reasoning. The Socratic method—challenging assumptions, testing logic, iterating toward better conclusions—is at the core of legal education. AI, especially reasoning models, thrives in a similar environment: it refines responses when challenged, improves through iteration, and exposes weaknesses in logic. So maybe the real power of AI in law isn’t just time savings or improved clarity—it’s that the best lawyers will learn how to use AI as an intellectual sparring partner, extending their reasoning even further. The real question is: Will lawyers be able to get past their learned risk aversion enough to actually try?
It's fascinating to see such improvements. I wonder how reasoning models will impact legal ethics education. Could AI help address bias in legal decision-making?
Ethan Mollick, the integration of AI in legal work is transformational! Imagine the efficiency boost combined with human expertise. ?? #FutureOfLaw
Generative AI seems to fit deeply into the legal realm, especially in leveraging precedents to reason about current cases. The news about hallucinations, such as citing incorrect precedents, has nothing to do with the real benefits of this technology…it simply shows a lack of understanding of how it actually works.
Glenn Tomai Esmond Yong
What is a "human level" for hallucinations? Do humans tend to make up cases or laws that don't exist?
A lot of good dense stuff in the paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5162111