More AI Hallucinations in an Oregon Court

Last fall, I wrote about an instance where a lawyer was caught using fake citations that were generated from artificial intelligence. It happened again in Oregon District Court, where Magistrate Judge Mark Clarke sanctioned a lawyer fairly severely in Couvrette v. Wisnovsky. The Judge questioned whether the lawyer meaningfully participated in the case:

At issue before the Court is whether Mr. Murphy willfully violated [Local Rule] 83-3 by failing to meaningfully participate in the case. Mr. Murphy argues that the Court should not impose a sanction because he did not participate in the summary judgment briefing and he was unaware that the Local Rules required him to review his associated pro hac vice counsel’s filings prior to submission.

If you’re explaining, you’re losing.

What is a hallucination anyway? If you’ve worked with AI to generate written summaries before, you’ve probably seen this. The computer has a tendency to just make stuff up. Since people tend to do that as well, one might think that’s a true sign of convergence with human intelligence. But no, AI is incredibly error-prone because it’s all just inductive reasoning. It chases correlations and is inevitably getting fooled by randomness. David Hume would know better, but a lot of dumb money in the stock market has been chasing a deductive-reasoing-free approach to knowledge that makes users vulnerable to what in statistics are called type 1 errors.

I’m not saying these new computational tools are worthless. We just need to skip even the notion of trust and focus on verification. I’m sure this lawyer wishes he at least fact checked his citations.

Eric Shierman lives in Salem and is the author of We were winning when I was there.