It’ll render ‘extremely compelling legal verdicts.’ Sure….
The post Elon Musk Feeds AI ‘All Court Cases,’ Promises It Will Replace Judges Because He’s An Idiot appeared first on Above the Law.
Elon Musk is not a lawyer. Nor is he really a genius inventor. But he is a guy with confidence inversely related to his competence and expertise, so it makes a lot of sense that he’s bragging that he’s inventing a legal tech product that will replace the whole system! An AI offering that will replace judges and “render extremely compelling legal verdicts.”
Grok, Musk’s version of generative AI, is just like OpenAI’s or Anthropic’s, except it spits out answers with a dash of conspiracy theory and all the humor and wit of a concussed rooster pecking on a Ouija board. Science fiction is full of artificial intelligences modeled after their creator and — on that count — the famously unfunny Musk seems to have succeeded.
Someone who still posts on X recently suggested that Grok might soon be able to summarize large pieces of legislation — you know, the thing that every commercial grade GenAI product is already doing — “so politicians can’t hide stuff from us.” You might ask, “Wouldn’t relying on summaries actually exacerbate the power of politicians and lobbyists to bury small, unrelated issues in a several hundred page bill?” And you ask that because you’re not the sort of person dumping your life savings into fake money schemes.
In response to this message about Grok’s summarizing functions, Musk wrote:
No, it will not.
But this does encapsulate the Elon Musk experience: throw some cases into an algorithm already struggling with basic fact inquires and announce that it will replace litigation! This approach tracks the overall development model for Grok, which Musk champions for training on its access to everyone’s public Tweets.
For most technologists, Garbage In, Garbage Out is a precautionary axiom. For Musk it’s a design philosophy.
To be blunt, “all court cases” means a lot of bad, cursory, and confusingly drafted opinions that aren’t particularly useful to anyone outside the parties. And sometimes not even clear to them.
Not to mention that a lot of times judges are just plain wrong and no one bothers to clean up the record. Many an erroneous summary judgment opinion sits safely on the books because the underlying case settles before trial. By way of a notable example, Kathryn Mizelle’s opinion that “sanitation” doesn’t mean “something that’s sanitary” because sanitation departments pick up garbage — is still perfectly undisturbed because afterward the Biden administration dropped the whole sanitary mask order on its own in light of the pandemic coming to an end. Nonetheless, Mizelle’s opinion fits squarely into the “all court cases” category even though it’s less Marbury v. Madison than that answer from Billy Madison.
Legal tech providers with far greater expertise in this field and a much deeper reservoir of secondary sources have put in a lot of effort to make legal AI work. When Thomson Reuters showed me some early AI work, they felt that “hallucinations” — AI being outright wrong — could be controlled with the benefit of their broader library, but that the challenge in building something that can work for legal is solving for results that are real but wrong. Misreading dicta, imputing parenthetical quotes in string cites to the case at hand, granting too much weight (or far too little weight) to a particular concurrence… these are all higher level challenges that actually serious legal tech providers are spending massive amounts of money to solve.
Meanwhile, Musk is just tossing cases into the hopper and letting Grok sort it out and deliver “extremely compelling legal verdicts” in favor of the plaintiff while taunting the defendant as “the fourth generation of imbecile” or some drivel. It’s the sort of half-assery he’s already shown toward his new job as head of federal government efficiency (maybe we cut every employee with an odd social security number!). Or actually, “co-head” because the efficiency group can’t even streamline its own leadership.
Though it’s also impossible to read this announcement outside of the broader context of Musk’s dealings with the law. He locked himself into a corner in buying Twitter because he blew off standard legal advice, waiving protections that he would — hopelessly — try to resuscitate after the fact. He cultivates fanboys who bombard judges ruling against him with angry letters. He’s working hard to move all of his legal exposure to appear before N.D. Texas Judge Reed O’Connor who just happens to be invested in Musk’s business. Of course he dreams of a chatbot that spits out opinions!
It would save him a lot of trips to Northern Texas.
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news. Joe also serves as a Managing Director at RPN Executive Search.
