Jenner & Block LLP partner Adam Unikowsky, widely regarded as a standout in the Supreme Court bar's next generation, detailed the AI experiment in his legal newsletter Monday night and discussed it further on Tuesday, saying in an interview that the computerized counsel proved "very clever."
It did so by answering the same questions that Unikowsky fielded from justices in October when arguing a civil rights case called Williams v. Reed
. Unikowsky, who earned undergraduate and graduate degrees at the Massachusetts Institute of Technology before attending Harvard Law School, prevailed in the case, but he suggested Tuesday that the chatbot's version of his side's argument also would've won — and done so with even snazzier syntax and citations."You read the transcript, there's a few answers where I get caught up and make mistakes, and the AI doesn't really make mistakes," Unikowsky told Law360, adding that the application impressed him by consistently connecting its answers to the case's central themes.
"I don't want to put myself down too much — I do have bills to pay as a practicing lawyer," Unikowsky joked Tuesday. "But I think many of its answers were as good or better than mine."
Those answers can be found in a transcript and audio that Unikowsky included — alongside the Supreme Court's genuine transcript and audio — in his Monday night newsletter, which contained provocative assertions belied by its anodyne headline: "Automating oral argument." In the post, Unikowsky indicated that his test of AI's oral advocacy abilities drew inspiration from a 2023 episode in which a company promised $1 million to any attorney who would deliver a Supreme Court argument by wearing earbuds and simply echoing the responses of an AI "robot lawyer."
The seven-figure offer seemed gimmicky at the time, but early 2023 was "the Paleolithic Era of AI," Unikowsky wrote. In the ensuing 30 months, the technology has progressed by leaps and bounds. Unikowsky utilized some of the latest digital instruments — such as voice AI platform ElevenLabs and AI model Claude Opus 4 — to create a near-match of his own voice and sophisticated answers to the actual questions the justices posed during arguments in Williams v. Reed.
The result was a startlingly realistic oral argument. Perhaps equally startling were some of Unikowsky's conclusions in his newsletter, including his assertion that "a robot lawyer would be an above-average Supreme Court advocate," and "AI will soon surpass even the best human oral advocates."
Those takeaways are notable partly because law firms have focused more on AI's efficiency with writing and research than on its ability to speak eloquently and incisively in high-pressure settings like the Supreme Court. In his Substack newsletter, Unikowsky averred that the emphasis has been backward, writing that "oral argument should be the first, not the last, frontier of AI-assisted legal practice."
Although professional norms are one obstacle to AI arguments, there aren't many court rules or ethics guidelines that explicitly prohibit automated oral advocacy, according to Unikowsky. One of the biggest hurdles may simply be that "no one has really asked" to deliver arguments via AI, he told Law360 on Tuesday.
But that's not to say that no one is using AI to deliver arguments, Unikowsky added. In the post-pandemic age, when various courts each day hold oral arguments via Zoom or telephone, lawyers can easily read real-time responses to judicial queries by glancing at AI apps such as ChatGPT.
"I guarantee you that ill-prepared and stressed-out lawyers are already using AI during remote hearings," he wrote in his newsletter.
In the post, Unikowsky anticipated various objections to his bullish AI outlook and sought to rebut them. As one example, he singled out the supposed "authenticity" of analog arguments, insisting that legal precedent shouldn't hinge on "a lawyer and a judge having a special moment of luminous kinship at oral argument."
"Wiping out authentic human connections is a feature, not a bug, of AI," Unikowsky wrote.
A reader comment on Unikowsky's post derisively said, "If we're going to replace the lawyers with AI, why not do the same thing with the justices themselves." Asked Tuesday about that comment, Unikowsky signaled that he's actually open to judicial work facilitated by high-tech tools.
"I'm definitely sympathetic to the use of AI in deciding cases. I don't think that we're going to just turn over the keys of the Supreme Court to AI. That's a bit of a tough ask," he said. "But, for example, I support the use of AI in arbitration," in the "incredibly slow" process surrounding veterans' benefits appeals and perhaps for arguments from litigants representing themselves.
It might seem unimaginable to let AI handle judging and oral advocacy. But it wasn't long ago that it seemed unimaginable for AI to handle myriad legal tasks. Now, those tasks are rapidly being handed to digital assistants, even amid a steady stream of cautionary tales about lawyers facing sanctions for filing briefs with AI-hallucinated case citations. Moreover, some members of the federal bench — such as Eleventh Circuit Judge Kevin C. Newsom — have emerged as enthusiastic AI proponents.
"Many judges are not implacably hostile to the use of AI," Unikowsky said Tuesday.
It's possible that AI fans and foes will both find fodder for their views in Unikowsky's newsletter. That's because a comparison of his actual argument and the AI argument reveals pros and cons, depending on one's perspective.
For instance, Unikowsky's real responses — like those of the justices and practically every human being — occasionally feature small pauses or hesitations, followed by repeated words. As an example, Unikowsky at one point in the argument said, "So I think it would — it would depend." By contrast, the AI's syntax is practically flawless — something the newsletter alluded to when it stated, "You can tell it's AI because it sounds a little too perfect."
As for the voices, they're difficult to distinguish, but Unikowsky wrote that he could've matched them even more precisely. He intentionally refrained from doing so, however, to err on the safe side: "I used my own voice to generate that voice, but the AI voice doesn't sound exactly like me. It's possible to create AI-generated voices that sound exactly like you, but I would like to stave off the DeepFakeOcaplyse for a few more days."
--Editing by Alanna Weissman.
For a reprint of this article, please contact reprints@law360.com.
