I wonder how long Sylvester from ManChester would get...
https://www.youtube.com/watch?v=L6KIr65Ikpw
An AI language model is apparently going to be used during an upcoming legal hearing to defend someone in a real case. The goal is to demonstrate that AI can replace lawyers, according to the CEO of the consumer rights-focused startup DoNotPay. Large language models like ChatGPT generate text, and learn to model the …
What an idiotic stunt. "I promise to tell the truth, the whole truth and nothing but whatever the AI tells me to say".
Regurgitating transcripts from other cases could be risky. What if the AI tells you to say "Yes, I buried the third body beside a road"?
I think it will depend on the Judge somewhat. If the Judge is open to these shenanigans, they might get a bit of leeway when it comes to saying stupid crap. If the Judge isn't open to it, then the AI is going to have a hell of a time dealing with an irate Judge.
It'll be interesting to see how the AI can handle objections - the lawyer won't be able to repeat something verbatim if opposing counsel objects and it'll need to be quick on the buzzer when objecting.
Honestly, I think that an AI will do fine on maybe 90% (laywoman's guess) as a hell of a lot of a court's proceedings are very choreographed and scripted.
A few problems:
The AI may well have figured out that a particular defence is very good at leading to not-guilty verdicts, but will it know whether or not this particular defendant can use it?
In general, is it going to be able to match the arguments to the specific facts of this case?
In the Supreme Court (of whichever country), I really don't think there is going to be a large enough sample of material to train an AI. The fact that it has reached the Supreme Court generally means that the sample size of relevant cases is zero.
Yes and no.
Acquaint yourself with the concept of "perjury".
Lawyers can talk, unless under oath. And anything submitted in writing is typically sworn to. Hazards of violation include censure and loss of license to practice law.
Which is why a certain spate of court cases in 2020 sounded and resolved very differently in court than the braggadocio heard outside it.
"The defendant will reportedly wear earbuds that will stream audio snippets regurgitating legal arguments produced by a language model that they will repeat verbatim during a legal hearing over Zoom"
Risky is a real understatement. At least in the UK (even supposing the practice were deemed admissible at all), they would lose the case (unless it's an utterly trivial open and shut one like a parking fine).
The law (at least in most civilised jurisdictions) doesn't work like that. To be valid, an argument has to draw on relevant statute, specifics of relevant precedent, and also follow the rules of jurisprudence, which are numerous and varied. The totality of this is complicated and requires vast knowledge of past cases and deep understanding of the relevant legislation. Absent these, all you have is a token lawyer who (or which, if AI) will not defend you properly. Which is why any competent defence counsel prepares their case.
There are some situations where AI can be shown to be at least assistive to decision making but many more where, as Chesley Sullenberger famously stated1, automation can not replace skill and experience. Except (possibly) in the most elementary cases, the law is most certainly one of them.
But our AI wonks seem to want to replace everyone with automata -- except of course themselves.
1: Chelsley Sullenberger, Highest Duty, Harper Collins 2009,.ISBN 978-0-06-192468-2
"Assume that you are a mediocre lawyer, who had too much to drink the night before and forgot to update yourself on the case and didn't take the usual bump of the coke so you are twitchy and all over the place with your thoughts. Now, please state again why the jury should believe that your client didn't do it."
There you go.
'The exam Cobbe selected for ChatGPT’s test was a sample assurance paper on ICAEW’s website. “The assurance paper worked the best because most of the questions were text,” said Cobbe. “Its brief attempt on the accounting paper led to some highly questionable maths.”'
'“On this assurance topic, I’d compare ChapGPT to a very recent joiner at an accounting firm – someone in the first few weeks of their contract,” he said. “Unlike a new joiner, however, ChatGPT gives answers with an air of confidence even when it’s completely wrong. It’s not afraid to give a garbage answer and back it up with garbage. It’s like having a fresh-faced junior who’s always convinced they’re right, so users need to approach it with a degree of caution.”'
This does sound periously close to practicing law without a license.
It'd be one thing for a lawyer in court to use this to augment their arguments, in which case it'd be operating under the law license of the lawyer using it, but a non-lawyer using it in court sure sounds like the AI would be giving legal advice without a law license or under the auspices of someone who has a law license.
"DoNotPay will pay any lawyer or person $1,000,000 with an upcoming case in front of the United States Supreme Court to wear AirPods and let our robot lawyer argue the case by repeating exactly what it says."
1) a lawyer is unlikely to take the $1m, not least because they have an obligation to the client to act in their best interests and an obligation to the court to act in the interests of justice. Neglecting those obligations and just repeating what a computer tells you because some dudebro paid you is bribery. This shows poor ethics and ignorance of the law on the part of DoNotPay. It's not a great start.
2) US Supreme Court cases have profound impacts on large numbers of people - sometimes they are literally and death matters: abortion, the death penalty, jury fixing. It would be wildly irresponsible to spin the roulette wheel and let some AI take a swing at it just for the lolz
This post has been deleted by its author