In 2009, I launched FantasySCOTUS. On the time, the thought was novel: a league the place folks may predict the result of Supreme Court docket circumstances. I quickly realized I had the makings of a nascent prediction market. We started to mixture the votes and supply knowledgeable predictions on all the pending circumstances earlier than the Court docket. The most effective gamers would predict as much as 85% of the circumstances precisely. By 2014, I partnered with colleagues to develop a machine studying algorithm that may predict Supreme Court docket circumstances. We referred to as it {Marshall+}. At its peak circa 2016, the algorithm was practically as correct as our greatest users–approximately 70% accurate.
Admittedly, over time, curiosity in FantasySCOTUS has light. I chalk it up to some components. First, the Supreme Court docket has grow to be extra predictable. With the brand new appointees, fewer votes are up for grabs typically. I feel the novelty of predicting outcomes has considerably worn off. As an alternative, the main target shouldn’t be on affirm/reverse, however how the Court docket guidelines. Second, coding every case for our algorithm proved to be too time-consuming. The juice was now not well worth the squeeze. Third, AI has fully obliterated all of our work. It’s now doable to ask an AI agent to foretell a case, and that work shall be much more correct than what our algorithm may need generated with hours of preparation. I keep FantasySCOTUS at this level principally as a novelty. There are a devoted group of gamers who nonetheless make predictions, and I’m grateful for his or her dedication. However I doubt this league will ever return to its heyday.
Within the early days, I used to be usually requested if the Supreme Court docket Justices would examine FantasySCOTUS. I had no approach of understanding, however I might joke with reporters that Justice Kennedy may examine the league to see which method to vote.
Now, it appears, issues have come full circle.
On Monday, the Supreme Court docket heard oral argument in Jules v. Andre Balazs Properties. The query offered issues arbitration. I will not even faux to grasp the info. Fairly, I used to be intrigued by a single query requested.
Adam Unikowsky represented the Petitioners. Adam has made headlines of late for his work on AI. His most bold project employed Claude to automate a Supreme Court docket oral argument. You must hearken to it. It sounds actually actual. And, it appears, the Justices are conscious of Adam’s efforts.
In the course of the seriatim spherical, Justice Alito requested Adam:
JUSTICE ALITO: Properly, simply out of curiosity, do you assume we must always ask Claude to determine this case?
The transcript notes there was “laughter.” I heard the distinct and extended laughs of Justices Kagan and Barrett. There could have been others.
Adam swiftly replied:
MR. UNIKOWSKY: No. I –I adhere to the sensible judgment of –of this Court docket.
I feel the impact of this joke is that legal professionals will now be afraid to even counsel they use AI. But it surely appears clear that many legal professionals are utilizing AI to generate questions that could be posed at oral argument–though they might not admit it. I give Adam credit score for being clear together with his strategies. Although, Adam could have gotten burned.
For no matter it’s value, Professor Scott Dodson requested Claude to generate a opinion within the case within the style of Justice Kagan.
