Without revealing the specifics of the case, can you provide a legally-equivalent example as to what this cutting edge AI actually managed to do?My team used a new cutting edge due-diligence AI to review approx 17 million pages of information in a complex litigation. In less time than it took to write this post, it found about 10 pages of info that are incredibly relevant to the litigation. I also had a team of about 10 lawyers who also reviewed the same dataset over a 6 week time period. That’s about 6 days x 6 weeks x 10 hours x 10 attorneys x $600 per hour for a first year associate in NYC. Or about $2 million of fees charged to a big Fortune 100 company. The team missed 3 of these 10 pages and found another 8 or so pages of info that were semi-relevant to the matter. So $2 million of human work was about the same as a few seconds of electricity. And this is just the beginning. YMMV.
I'm curious because it's hard to tell from your post if what the AI did was truly impressive or just quite mundane. Was it a matter of your team feeding a list of relevant words/parameters to the AI and then having it comb through the data and tell you which pages matched those parameters? So it's really just a better type of "Find" function? That's still quite wonderful, but totally conceivable and not quite impressive (proportional to the hype it generates).
Or did it return things based on an open-ended question? I don't know ... say, "The U.S. government contends that the removal of our app from digital platforms does not violate the First Amendment. We intend to argue that PAFACA triggers First Amendment scrutiny, and furthermore, that it fails to satisfy intermediate scrutiny even if strict scrutiny does not apply. Identify the most relevant evidence."
I guess I'm trying to understand the nature of what it accomplished. Did it find things that were unexpected, or was it just simply much faster at doing a fairly mundane task?
Statistics: Posted by Caduceus — Sun Feb 23, 2025 12:20 am