The legal system has never been so easy.
Thanks to e-Trial from Cinco, it's fun!
All you have to do is enter your plea, choose an e-Jury, and submit your evidence. Then, sit back and watch as each e-Trial automatically generates a legally binding verdict. Plus, the trial is never wrong! Cinco e-Trial is perfect for handling cases such as fraud, divorce, assault, rape, murder, tax evasion, petty larceny, destruction of property, and more.
> the company's AI-creation runs on a smartphone, listens to court arguments and formulates responses for the defendant. The AI lawyer tells the defendant what to say in real-time, through headphones.
Based on AI’s propensity for being confidently incorrect, I’m not sure I’d want to be repeating AI output in a court of law and representing it as truth.
But this xan be be a. ery good complement to an actual layer.
The layer would spend less time preparing (so cheaper) and will basically direct the IA and act as safeguard when the AI gets so confidently incorrect.
Then I imagine the defendant would risk enjoying a stretch in jail for contempt of court, for creating an unauthorized recording or live "streaming" of the court?
I'm not even convinced by the idea of having a person repeat what they hear through headphones - something like this could easily slow proceedings down, and that's without even going into whether or not the AI can stick to a consistent and accurate set of facts.
>The ultimate goal, according to Browder, is to democratize legal representation by making it free for those who can't afford it, in some cases eliminating the need for pricey attorneys.
this is the core issue I have with companies and tech like this: they try to solve social issues with technology and business. the issue of accessible defense is an issue of institutional failure that rests on the shoulders of the state. tech i think may bandage the problem, but just like how Tesla won't solve green travel this won't fix the underlying issue
Working within the existing system, though, is typically the fastest way to address the grievance. AI is probably the most scalable approach to competent counsel, far easier (given ChatGPT) than training new lawyers and convincing them to become public defenders.
AI will give crappy representation to those that can't afford it who will then lose from those that can. This isn't the fastest way to address this particular grievance, it is the fastest way to ensure it won't get addressed at all.
The fact that you can't get a half decent lawyer to represent you unless you are willing to go bankrupt isn't a feature, it's a bug and this just further increases that divide.
True for wealthy countries, but in really poor countries access to things like medical care or legal services isn't a policy issue, there is a legitimate scarcity. Barring foreign aid, the only two alternatives are no access to these services, versus access to an automated version that's worse than a human but better than nothing. So we're not dealing in a world of perfect solutions, just less bad solutions, which is fine. Note that I'm speaking generally, not about the particular service this article is about.
I disagree. Technology is the perfect solution for problems like these. So many societal issues today arise because systems developed centuries ago cannot keep up with a population that has grown exponentially since then. Tech's core feature is being able to scale. There will be more and more systemic reliance on automation and AI as time goes on, since that's the only sensible way out.
But the problem with law isn't the number of people, it's with the number of laws/binding contracts. If the US government can't even count the number of possible ways you can commit a crime then how is any of it supposed to work?
It's a way to show that the legal system is too complex. In the 80s the Justice Department tried to count the total number of criminal laws and they failed.[0] If the government can't even keep track of the number of them, then how is it reasonable to expect regular people to follow all of them?
This is a feature, not a bug for the classes which benefit from the ever expanding legal corpus. Coincidentally, politicians and lobbyists are often lawyers themselves.
That's the opposite of true. And I think you know that. It's absolutely insane that HN of all places is arguing against the concept of locality-of-reference.
Given the Option A (letting DoJ prosecutors only have to carry around Title 18) versus Option B (letting lay people looking up a topic by subject matter see what crimes and penalties pertain to that subject matter in the title for that subject matter), which one is friendlier to lawyers? Which one is friendlier to lay people?
Yeah, a DoJ official in 1982 probably found it inconvenient that 17 U.S.C. § 506 and 18 U.S.C. § 2319 aren't in the same deadtree volume for them to lug around. But if I'm a lay person, I'm glad that the criminal penalties for copyright infringement aren't invisible from within the chapter on copyright infringement in the title on copyright.
If I were to perform a code review for a junior engineer, and they had structured their source files so that all the constructors and destructors for all of their classes were in one independent file, rather than co-located with the classes to which they pertain, they would not pass code review. The argument that it's difficult to even know how many constructors and destructors they have, and therefore memory management is too complex, would be distinctly unpersuasive. It baffles and frustrates me that HN seems to come to the exact opposite conclusion with regard to the law.
That is obviously untrue. Why do you feel the need to make obviously false claims to support your position?
You need to obey the UCMJ?
You need to obey campaign finance laws?
You need to obey the Stop Trading On Congressional Knowledge Act?
You need to obey state law as incorporated by the Assimilative Crimes Act in jurisdictions you've never been to?
You need to obey HIPAA? The Privacy Act? The Freedom of Information Act?
A huge number of laws simply do not apply to you if you aren't engaged in a particular activity. E.g., you are not a covered entity for HIPAA unless you work for a medical provider or insurer. And a huge proportion of those criminal penalties only apply to government employees, not to average civilians at all.
You need to be aware of that before making decisions in life though. You also need to be aware of them to know how others around you can legally act.
Also, there's the problem that you don't know what is written in the laws that you don't know. Maybe it does pertain to you, but without knowing the law you can't know that.
The notion that monolithic tech actually scales is not a given. The strategy typically used to disrupt an industry is to provide a convenience for free that actually costs the company something. These costs (and liabilities) are familiar to the incumbent.
That institutional knowledge disappears upon a successful disruption. The startup will now be faced with learning these things over again, “breaking things” in the process.
Once enough growth has happened, monetizing via ads or increased pricing occurs. After this, you may have to squint your eyes a bit to tell if the situation before was all that much worse- or merely consisted of a different set of trade-offs.
I think that's utopian, all i see in this story is a future where some people can afford expert defendants (who'll be assisted by AI but much more effective than AI alone), while poor people are stuck with representing themselves with an app - and the government deems the problem solved, further eroding laws that guarantee legal support.
I mean, if the AI app is better than public defenders, it’s still progress. Rich people can already afford experts, so the inequality would be lessened.
Tech often misses obvious downsides, especially human, in pursuit of unrealistic idealism.
A good lawyer is not just knowledgeable of law, but of people. He needs to get 12 people in front of him to agree with him. And he may even be actively misleading them when he's defending a someone whom he knows to be guilty. This comes down to reading people, reading reactions, and understanding what each of those 12 people needs to hear to have them act like he wants. Not coincidentally, the most common background of politicians: lawyers.
I'd argue that the most likely scenario here is not closing the divide but widening it. Instead of public defenders, poor people will just get the bot, while those of means will get lawyers. It's akin to tech support today. In the not so distant past, it was easy to get a human on the line. Now in many cases unless you're a "special" client, or get your issue on the front page of some big social media site, you get the bot. And it sucks.
If poor people can only get the bot, that means the Constitution was changed. I don’t see how that could be the tech’s fault. It’s more likely that the AI would free up public defenders from things like having to answer their clients’ questions, as they could just ask the AI, allowing them to be more effective.
> It’s more likely that the AI would free up public defendants from things like having to answer their clients’ questions, as they could just ask the AI, allowing them to be more effective
Public defenders would be more effective if they didn't have to talk to their clients?
I read it more as, clients have a mix of questions. Clients don’t go through the system everyday and likely have some general worries and procedural questions as a result. Clients also have a case and need to coordinate with their lawyer. AI could enable more of the latter to happen within the limited time the lawyer has for each client.
Sorry, I can’t ever write clearly on the phone, this site is very mobile-unfriendly. I meant that defendants would be able to talk to the AI about their case to learn as much as they want. A public defender can’t commit that much time to each of their cases.
The problem with this is "how do you assure the AI isn't bold face lying to the defendant".
For a natural text generation/chat system like ChatGPT, at the moment this is effectively impossible and frankly I don't see it being sufficiently fixed for something as complex as teaching law to a non-law-student any time in the near future.
So you make this as understood as possible and the lawyer can fill in the gaps and clarify.
It's not like a "jailhouse lawyer" will do any better. Though it might cost something you didn't expect, such as a new witness against you or a shower pal.
I think these discussions always seem focused on the ideal state of the "AI is perfect and gives correct answers" and not the current state where the "AI constructs sentences that are structured correctly, but are filled with inaccuracies and illogical nonsense."
If we want to steelman AI why shouldn’t we also steelman the court system? Assuming a perfect AI is about the same as assuming a well functioning court system. Otherwise we should be assuming a flawed, potentially racist, confidently incorrect AI vs a flawed, excessively slow, definitely racist court system.
I don't think it's too far fetched to imagine the courts telling poor people they have to settle for an AI to represent them unless they can fork over the money to hire one themselves.
If an AI is just as capable (if not more capable) than a real lawyer, we'll be seeing rich people and corporations using them instead of humans. Until we see that, I'll assume that if poor people are forced (by law or finances) to use them they'll be getting the short end of the stick.
An overwhelming majority of Federal cases don't involve a jury and instead consist of:
1. A scaling army of prosecutors and "professional witnesses" which are compensated/promoted based on conviction rate.
2. A single "freely" provided and extremely overworked defense attorney with three times the caseload of the prosecutor and where every bit of help they give out takes from the time of the other defendants they "defend."
3. As a result of 1&2 you are pressured to sign a plea and threatened with triple and quadruple punishments for the same activities. All while being stuck in a cell (why else would you need a free attorney)? And we are not talking the difference of 3 months and a year. Instead we are talking about the difference of 5 years and 20 years.
4. The result is an overwhelming number of factually innocent people in prison for crimes they never committed. From people who sat down and said, what happens if it goes wrong vs what happens when I do 5 years of adult daycare?
So yes, any kind of AI is likely to improve the situation as the situation is currently completely unacceptable to anyone who has seen a family member go through it.
> A good lawyer is not just knowledgeable of law, but of people. He needs to get 12 people in front of him to agree with him.
While I agree with you that AI powered lawyer can have many downsides and drawbacks here you are building on a falsehood.
There is this illusions that computers are "book smart" but not "street smart". It is very deeply ingrained into our culture. Just think about Data from Star Trek. Very analytical, but struggles to understand emotions. Therefore also struggles to manipulate people's emotions, aka being emotionally persuasive.
This is a sci-fi trope, but is false as it comes to these generative models. (or can be false depending on training data) If you train the system with mock juries it can learn how to twist them around its finger. Text written by such a model can be as emotionally persuasive as one written by the best lawyer.
In fact in reality, as the tech is today, it would be easier to reliably implement the "reading reactions, and understanding what each of those 12 people needs to hear to have them act like he wants" part than the "follows flawless logic, legal technicalities to a T" part.
It is of course still a very perilous path to make an AI-powered lawyer. It is just important to not base our reasoning on sci-fi tropes
> it would be easier to reliably implement the "reading reactions, and understanding what each of those 12 people needs to hear to have them act like he wants" part
Come to think about it. This part can be valuable, even as a startup. You get mock juries, you act out legal cases in front of them. You collect every single thing the actor lawyers say, and how they say it, and you also collect full transcript of their deliberations.
You model the data, and then you sell the services to lawyers. They can practice in front of your model jury initialised with what can be known about the real jurors and can see traces of their deliberations. Not just one run of the model, but many runs if they want to.
P.G. Moneybags here - you know when I have a support issue at either of those companies regarding the billion dollar advertising campaign I am running I never have any problems getting support.
The solution seems clear - poor people just need to run billion dollar advertising campaigns to improve their support.
While I agree that technology is the solution I disagree AI has anything to do with this.
This "AI Lawyer" is about a traffic ticket. If we want to make things more fair while also enforcing the laws on the book it seems like the more straightforward tech solution is to improve traffic cameras/monitors such that tickets are given out more appropriately. Or, if it still goes to court, why not just have a system where someone can represent themselves with a form or checklist which points out the reasons they were given a ticket by mistake?
Perhaps this is a good business model in the current system but its definitely not the best for society as a whole. Why do we need lawyer of any type to handle something as banal as speeding tickets?
> like the more straightforward tech solution is to improve traffic cameras/monitors such that tickets are given out more appropriately
This is not straightforward from a bureaucratic sense. There is no strong incentive (that I can identify - humanitarian ideals do not count) to improve the functionality here to make traffic tickets and their resolutions more sensible.
However, an AI that can resolve traffic tickets and citations might incentive a city to start gathering data more effectively - to win cases and get revenue - and that is a great incentive to improve the general tech.
> Why do we need lawyer of any type to handle something as banal as speeding tickets?
I also want to mention that this is a great point. The fact that my job provides legal insurance, the end result being I can get out of my traffic tickets by hiring a lawyer for free to show up and ask for a dismissal, is pretty dumb.
> Normative statements. You should be asking what are the incentives.
This is itself a normative statement.
I’m of the opinion that incentives aren’t the only thing worth considering — there’s something to be said for enshrining ethics into law (if only because it can adjust incentives).
Another thing worth considering is our mediocre track record of identifying incentives ahead of time.
FWIW I think adding more cameras and monitoring equipment would have a strong incentive since it could still increase the total number of tickets given out.
Either way the government is not purely incentivized to increase funds. Standard fines for traffic violations are changed up or down all the time. Sometimes cities decide their current traffic enforcement is too burdensome and they cut back on tickets. There could be political actors which have incentives to improve it. Whether or not it will ever happen is not clear but I dont think its impossible by any means.
This is just great press hacking by the company, jumping on the coverage ChatGPT has recently received. In essence, the company is just offering a checklist of reasons why the ticket was given by mistake but making it sound advanced and technologically impressive to a wide audience. Which seems like a perfectly fine societal use to me.
I always hire a lawyer for traffic court. Show up with a lawyer and wear a suit. Don't want them to think I'm one of the poors to stomp down. If you go to low-level court in middle America you'll understand what I mean -- 99% of the people there are drugged out and wearing rags generally look like their shit is not together at all and these are the people the lazy/burnt-out prosecutor is geared towards crushing effortlessly.
They basically always offer some bullshit non-moving violation to make me go away. Saves me thousands on insurance premiums and has also saved me from a criminal record for my more serious tickets (like allegedly driving gasp 36 mph by a school, which apparently comes with jail time in Ohio).
Weird that you're so experienced in defending driving like a jackass that you have a ton to write about your strategy and feelings on that topic.
Most people don't get ticketed often enough for speeding to feel one way or another.
And most people don't feel righteous about speeding in a school zone.
Convicted/not convicted doesn't map perfectly onto guilty/not-guilty. You yourself understand this perfectly or you wouldn't be hiring a lawyer for a moving violation.
Yeah, that's just a dumb claim. A bigger population means more people who can become lawyers. The problem isn't population, and never was, at least for this problem: it's always been the case that justice is arbitrary and money buys better lawyers.
> So many societal issues today arise because systems developed centuries ago cannot keep up with a population that has grown exponentially since then.
If you had 1 lawyer per 100 people in 1800 and you had 1 lawyer per 100 people today, I don't see a scaling problem with population.
Indeed in the UK in 2012 there were 120,000 solicitors for 64 million
Today it's 160,000 for 68 million
That's a 30% increase in lawyers with a 6% increase in population.
Why do we need more lawyers per person today than 10 years ago?
> If you had 1 lawyer per 100 people in 1800 and you had 1 lawyer per 100 people today, I don't see a scaling problem with population. [...] Why do we need more lawyers per person today than 10 years ago?
How many more laws are there? Being a lawyer is like 90% reading and collating documents to filter out relevant information. We have a lot more law and a lot more documentation to sift through these days.
We have remedies when a lawyer is incompetent. What sort of remedies do we have when the AI is incompetent? Was there some EULA? Did someone click it? Should that even be allowed? It is my position that the companies that produce AI lawyers (and "expert systems" of any type) should bear full liability for everything and anything that their product does wrong.
The defendant is pro se and can choose whether to include info from the AI. If you fuck up your own pro se case because your uncle or a dumb robot told you something stupid that's on you.
Nitpick: From a glimpse at the article, I think the defendant will represent himself, the "AI lawyer" running on the defendants smartphone will just give him hints. Not sure how it can be acceptable to have a third party listen into court proceedings in real time through a smartphone, but ok.
> The AI lawyer tells the defendant what to say in real-time, through headphones.
Yeah - and I wonder what will happen if it starts telling the defendant to say things which aren't grounded and rooted in fact, or which contradict things they previously said or submitted in writing.
Not sure a generative model is the best way to keep people clear of obvious cases of perjury/lying under oath etc.
Yes. If its one thing we've seen ChatGPT and the like struggling with, it's internal consistency and keeping the whole text coherent and tied to some fundamental overarching thread or strategy.
To play devil's advocate though, would a layman without AI help be any better at navigating this?
But courts are very likely (in my view) to take a very very dim view of the introduction of some kind of barrack-room "AI lawyer". It will slow things down by introducing potentially spurious arguments which wouldn't have been made by either a represented party, or unrepresented party.
Given the courts are a public service, this isn't necessarily zero sum - unrepresented parties do take longer in court, but I very strongly suspect in the life of a whole case, parties with an aspiring barrack-room AI lawyer will take up more court time. Especially if their argument loses coherency or starts to contradict itself.
This is not an AI problem, I have seen more than a few time Lawyers contradict in different parts of a case. One example off the top of my head is when Lawyers attempt aurge that social media platforms are both publishers and platforms so they can not be held liable at all under any circumstance
> One example off the top of my head is when Lawyers attempt aurge that social media platforms are both publishers and platforms so they can not be held liable at all under any circumstance
... that would be very surprising, since there is no legal definition nor distinction of publisher or platform.
Even TechDirt manages to get this right, but the myth that tech companies exist on the knife's edge of being converted under the law into publishers is stubborn.
Well, INAL, but during the legal interactions I had, none of which was in the US or got to actual trial, whenever the clients didn't follow the legal guidance of the lawyer, the lawyer was not accountable. Heck, even if you follow your lawyer's advice, the lawyer isn't accountable for the outcome. Only for the correct conduct of his practice. So I'd say, in this AI-supported court case (for speeding ticket after all), if the AI screws up I think the defendant is on its own. Or we see a follow up civil suite against the "AI" provider, which would be both, hilarious and highly interesting.
Clients not following their advice is always going to happen, and it's not the lawyer's problem as you say. And indeed, the lawyer can't (and historically in many jurisdictions shouldn't be creating compensation structures which imply they can) guarantee the outcome.
I agree the defendant is likely to be on their own, and I'm wondering whether the individual concerned is aware of this. The company seems to be willing to backstop their ticket if they lose.
I'm not sure that will help much if they are sent to jail for a few days for wasting court time for relatedly lying under oath or talking themselves into a generative AI's web of inaccurate statements. Such a follow-up suit would be both hilarious and interesting to follow indeed.
- safe transport for women at night: Find My Friends / Uber+Lyft
- access to government data: opendata initiatives, a plethora of cadastral map services
I'm going to say the obvious thing everyone is thinking. Everyone who has ever said "this is the core issue: you're trying to solve a social issue with technology" has never solved a social issue _or_ a technological issue and has no idea how to do it.
0: obviously I must clarify for this audience that "solved" in this context means that it has reduced the prevalence of some problem below some low base rate
Unfortunately the patriarchy has been clamping down on this to make it more difficult for women, who often have lower incomes, to get firearms that would better equalize their self defense capabilities. They impose FOID, licensing, and various other barriers to carry that are less affordable to women.
Tesla was a "forcing function" that caused other car manufacturers to build EVs. AI lawyer technology may be the catalyst needed to trigger reform of the legal system. It's clear that up until now there has been no reform forthcoming.
I think it's more likely that AI systems will act as a crutch, enabling even more overcomplicated procedures to grow and persist, until it's impossible for any mere human to understand.
With technology or a product that people will voluntarily pay for, you can make a specific, measurable, concrete difference in a bounded timeframe. If you fail, you can be sure to do so without causing immense suffering across a whole society in the way that government actions sometimes do when they go wrong. If you could "fix the underlying issue" in a permanent way without unintended consequences, that would be better, but no one knows how to do that reliably using politics, and the failure modes of airbnb, uber, and tesla are gentler than the failure modes of politics.
This is the core issue I have with comments like this: they refuse to try solving social issues with technology. I can understand why not business, but every time a solution involving tech pops up, there's immediately people dismissing it with a comment saying you can't solve societal issues with tech, with no proposed alternative solutions either. Ok, you don't want to solve it with tech, how then?
This is the core issue I have with comments like this: they refuse to try solving social issues with business. Every time a solution involving business pops up, there's immediately people dismissing it with a comment saying you can't solve societal issues with business, with no proposed alternative solutions either. Ok, you don't want to solve it with business, how then?
A an aside business has been shown to be the best path to solving societal problems, from medicine, to poverty business is that path. Leave government to solve a problem and as the old saying goes you would get the best iron lung money could buy but you would never get a polio vaccine.
I agree with you, but I also can understand why people may think that profiting from solving a social problem could create a slew of other issues, such as incentivizing the proliferation of that same social problem so that solving it can turn a profit. It's just business!
On the other hand I don't understand this hatred towards tech, aside from "it's new, and I don't understand it, therefore it's bad".
I don't have an issue with tech let say, but how he tech serves more as a stop gap for visible flaws while the underlying core remains flawed. this tech won't solve the problem of overzealous police officers, biased judges, and other problems stemming from the institution itself
How is this a social problem? Lawyers are expensive, because education takes time, motivation and competence. So it's an economical problem, something the state can't easily solve. Technology can make an impact here and sidestep or even ignore those problems.
There are multiple facets to this question depending on the type of law you are talking about.
Criminal, Civil, Contract, Torts, Family, etc.
In many areas of law I would agree with your statement, however with criminal law I have always felt that many governments have been in the habit of criminalizing poverty in many ways. Personally I think it should be mandated that at a minimum for every $1 given to a prosecutor's office to prosecute criminal cases, $0.50 should be given to public defenders (ideally $1 for $1 but I would take $0.50) currently is more like $0.10.
Education is always expensive, in every country, because it takes time. US-Universities are just exceptional expensive for whatever reasons. But time is something that this technology can remove for this cases, so I would say it still remains an economical problem.
Though, the ability to use the technology might come down to a social problem. Because it still needs a certain level of education and competence to use the technology. Just significant less than visiting a university for a whole decade or so.
> So it's an economical problem, something the state can't easily solve.
The state can simplify the law and eliminate or stop prosecuting laws that don't directly harm people, thus reducing the demand for lawyers. It's not just an economical problem.
It could also make things much worse, due to Jevon's paradox. The way rich societies throw away tons of food because it isn't scarce, or how modern software is slower than ever because hardware is too cheap. If half-baked AI's lower *some* of the costs of law practice, we might end up wasting more resources than ever before fighting against the proliferating volume of it.
See, the sword cuts both ways. Prosecutors' office would run more efficiently, would pursue much larger caseloads (lower priority ones) with fewer employees and less time. Civil lawsuits could become much cheaper to file; trolling could expand from niches (like patent law) to much broader, more socially-impactful forms of warfare. The financial risk/reward equation is what deters this, and lowering the costs means eliminating the deterrence. And imagine regulatory compliance costs, when regulators feel emboldened to design regimes so elaborate they themselves don't understand them, because they have AI assistants capable of parsing them.
I'm imagining a dystopian future where everyone wears a legal advice headset that is constantly whispering to them about things they are doing, or need to change to avoid a lawsuit. Kind of like a continual game of twister where if you misstep, all the watching legal AIs jump on you, immediately serving you electronically, with court dates pre-scheduled and all the evidence they need to win the case already submitted.
Its a state created problem and you expect them to solve it? Private entities are the way these problems are always solved efficiently. Cant wait for lawyers to be obsolete personally.
Public defenders are state entities. The state failing to fund its justice system to guarantee speedy trial, competent representation, etc. is a state problem, not a problem of technology.
You can represent yourself in court in many cases, the problem is that you will likely fail as you won't be very good at case law, abiding by all sorts of conventions and therefore lose.
If AI can help you make properly referenced arguments it might well solve a societal problem to a degree.
> this is the core issue I have with companies and tech like this: they try to solve social issues with technology and business. the issue of accessible defense is an issue of institutional failure that rests on the shoulders of the state.
Who's to say DoNotPay won't eventually sell this service to the state, in service of the public interest in accord with the Sixth Amendment? In that light, it may be a tool worth developing
I think it’s exactly the opposite. People broadly overestimate the effect of social and political solutions while overlooking the impact of technological and economic changes. For example, why did gender roles suddenly change in the western world in the mid 20th century? Was it because people wrote books and passed laws? Or was it because the economy shifted from physically demanding agricultural and industrial work to a services and knowledge work economy? I’d argue that the typical course is for social change to follow technological and economic change.
Similarly, the the problem with representation of defendants in the legal system is technological and economic, not social. It’s expensive to have a credentialed person spend lots of time working on behalf of someone else. That’s true for everything from auto repair to law. The social and legal system is simply responding to that economic reality. The public is willing to spend so much money to protect the rights of the accused and there is only so much you can accomplish by haranguing them into wanting to spend more money. By contrast, if you dramatically reduce the cost of providing the accused with a defense, you dramatically change the landscape of the social and political issues.
While a pure ChatGPT-style bot would be prone to hallucinations and overconfidence, isn't law actually a fertile field for working out grounded chat LLMS?
The lawsuits I've been involved in typically hinged on pretty arcane stuff which I highly doubt is going to work well without an entity that can quite literally think on its feet. Just showing up is 90%, but what you do when you show up is the other 10 and if you botch that you may still lose a case.
'Sorry, that wasn't in my training set' won't cut it if the penalty is >> the price of your legal fees, an extended stay in jail or in some countries the death penalty. Obviously for criminal cases you'd be fairly mad to try this kind of construct but that's exactly the ones where the legal fees are absolutely crippling.
I've yet to have a case that I brought that went to trial end up under 50K in costs and that was with what I thought was fairly simple stuff with relatively unsympathetic defendants.
I could see an AI lawyer being a great assistant for human lawyers. I dislike the idea of an AI lawyer being appointed to me if I cannot afford my own representation, especially as the county NN isn't going to have as much resources as Kirkland and Ellis's NN that's been trained on all the cases they downloaded from Lexus Nexus.
The problem is that half the working class wants those loopholes to stay in place because they have this delusional notion that one day, they'll be the ones making use of them as well.
"At no time at any location within the courthouse may any electronic device be used to make a photograph or recording in any fashion, whether audio, video or otherwise."
This is covered in the article, and they apparently think that it could be legal in two out of 300 courtrooms. You've linked to just one specific court's webpage.
> Does media coverage on criminal cases destroy the fabric of criminal proceedings?
From the perspectives of both sides (defense counsel and prosecution), I believe this is undesired as it primes potential jury members' opinions outside of court evidence.
I am fairly sure I saw live, in-courtroom coverage of both the Rittenhouse (as a prominent criminal case) as well as the Depp/Heart (a prominent civil) case.
Sorry, but can anyone explain how having one or more AI lawyers is any different from allowing an AI judge in a proceeding?
First of all, AIs are not beholden to the bar association. They aren't at risk of losing their license to practice. A poor AI will not convince a human judge as well as a human lawyer, leaving the defendant with an improperly executed defense. A great AI might, perhaps, be able to win a case, but then the question would be whether the human judge was able to understand the full implications of the AI's arguments. Only an AI judge would be able to rule impartially against an AI lawyer.
Which I'd assume we would all find to be a horrific path for our already completely fucked up justice system to follow. After all, if AI can be lawyer and judge, it would only make sense to have the laws written by AI legislators, approved by the AI executive branch, and enforced by the AI police and military.
I remember discussing ideas for law automation ca 2016 that weren't even quite as radical as this. Kind of shows that we tend to overestimate change in the short term and underestimate it in the long term. I don't know how legit this specific product is, but the likes of GPT have definitely put this into the realm of the imaginable.
The upside of this is that it could prevent self-represented defendants from trying to apply legal arguments or concepts in the wrong context. Pro se defendants do this all the time and it really pisses off judges. ChatGPT has a reputation for being "confidently wrong", but in my experience it's less likely to make that specific category of error.
The downside is that I hire a traffic ticket attorney because they've been handling a cases similar to mine, in the same court circuit, almost every day for years. Legal work isn't about knowing the law. It's about knowing how the system works in a particular place and time. ChatGPT was trained on many thousands of various legal cases. But it can't ever really understand how a given court works unless many people have written about that place and posted it to the internet.
I feel like it would be much better for people to just type their case into the ChatGPT prompt and ask "What argument should I use?", and then take that into the courtroom. Ask it what counter arguments to expect, what the standards of evidence are, etc. You can even prompt it with "Pretend you're a traffic court judge. How would you respond to this argument?" Essentially use it as a research tool for people who are brave or foolish enough to represent themselves. Not sure how much benefit there is in having that in the room with you.
We need AI judges, not lawyers. Then we can have prompts, which make you non-guilty on each occasion, may be also provided by AI for convenience. We would save so much money on taxes.
Some work a lawyer does is menial word churning. Some work a lawyer does requires much more mental effort.
If I could feed thousands of documents to an algorithm and then ask him stuff like "in which pages the tax returns of the defendant are discussed?" or "how many times is the defendant's wife talked about in a negative connotation?" it would make work much easier.
"I was apprehended under a level three predictive crime directive. My AI lawyer failed to cite the relevant statutes. Now I've been remanded to a medium security behavioral modification center. After incorrectly answering the AI managed multiple choice psychological evaluation questionnaire, I was denied parole. When released I'll have to start all over again with a social credit score of zero. My mother always said I would run afoul of the algorithm. If only I had listened..."
What if we take it from the other end and make an AI prosecutor?
Input parameters would be a description of the scene and output would be a list of crimes anyone involved has committed along with confidence rating of how well it would succeed in court.
Once this tech gets good, and everyone has access to a high-powered legal team for pennies, I am a little concerned that anyone will be able to DDoS the legal system.
It's already possible for giant corporations to bury smaller actors in paperwork. They can and do crush people with this.
I'm not sure if we (at least, in the US) have the institutional capacity to do so.
The legal system is interesting because it operates within its own framework.
Industries can be disrupted from the outside. People switch when the next best thing comes along: horses to cars, gas to electric, wired to wireless.
The legal system can only be (legally) disrupted from the inside, from within itself. If AI tech gums up the legal system itself, how will we use the legal system to fix the legal system? Regime change / revolution is probably not ideal.
The legal system is already under DDOS if you want a speedy trial. There are people in jail for years without being found guilty for anything simply because the courts don’t have enough capacity to manage their current criminal trial load.
173 comments
[ 3.0 ms ] story [ 210 ms ] threadBased on AI’s propensity for being confidently incorrect, I’m not sure I’d want to be repeating AI output in a court of law and representing it as truth.
The layer would spend less time preparing (so cheaper) and will basically direct the IA and act as safeguard when the AI gets so confidently incorrect.
It's going to suck so hard.
I'm not even convinced by the idea of having a person repeat what they hear through headphones - something like this could easily slow proceedings down, and that's without even going into whether or not the AI can stick to a consistent and accurate set of facts.
this is the core issue I have with companies and tech like this: they try to solve social issues with technology and business. the issue of accessible defense is an issue of institutional failure that rests on the shoulders of the state. tech i think may bandage the problem, but just like how Tesla won't solve green travel this won't fix the underlying issue
The fact that you can't get a half decent lawyer to represent you unless you are willing to go bankrupt isn't a feature, it's a bug and this just further increases that divide.
This strikes me as a bizarre thing to say and I'm curious what meaning you ascribe to this statement.
What would be the value of counting such a thing? Why would you want your tax dollars spent on such a task?
[0] https://blogs.loc.gov/law/2013/03/frequent-reference-questio...
Given the Option A (letting DoJ prosecutors only have to carry around Title 18) versus Option B (letting lay people looking up a topic by subject matter see what crimes and penalties pertain to that subject matter in the title for that subject matter), which one is friendlier to lawyers? Which one is friendlier to lay people?
Yeah, a DoJ official in 1982 probably found it inconvenient that 17 U.S.C. § 506 and 18 U.S.C. § 2319 aren't in the same deadtree volume for them to lug around. But if I'm a lay person, I'm glad that the criminal penalties for copyright infringement aren't invisible from within the chapter on copyright infringement in the title on copyright.
If I were to perform a code review for a junior engineer, and they had structured their source files so that all the constructors and destructors for all of their classes were in one independent file, rather than co-located with the classes to which they pertain, they would not pass code review. The argument that it's difficult to even know how many constructors and destructors they have, and therefore memory management is too complex, would be distinctly unpersuasive. It baffles and frustrates me that HN seems to come to the exact opposite conclusion with regard to the law.
The only reason this system works is because of privacy. Without it the system is just broken. But even with privacy it makes the system about luck.
That is obviously untrue. Why do you feel the need to make obviously false claims to support your position?
You need to obey the UCMJ?
You need to obey campaign finance laws?
You need to obey the Stop Trading On Congressional Knowledge Act?
You need to obey state law as incorporated by the Assimilative Crimes Act in jurisdictions you've never been to?
You need to obey HIPAA? The Privacy Act? The Freedom of Information Act?
A huge number of laws simply do not apply to you if you aren't engaged in a particular activity. E.g., you are not a covered entity for HIPAA unless you work for a medical provider or insurer. And a huge proportion of those criminal penalties only apply to government employees, not to average civilians at all.
Also, there's the problem that you don't know what is written in the laws that you don't know. Maybe it does pertain to you, but without knowing the law you can't know that.
That institutional knowledge disappears upon a successful disruption. The startup will now be faced with learning these things over again, “breaking things” in the process.
Once enough growth has happened, monetizing via ads or increased pricing occurs. After this, you may have to squint your eyes a bit to tell if the situation before was all that much worse- or merely consisted of a different set of trade-offs.
Uber comes to mind. Facebook as well.
The defendant is the person on trial.
A good lawyer is not just knowledgeable of law, but of people. He needs to get 12 people in front of him to agree with him. And he may even be actively misleading them when he's defending a someone whom he knows to be guilty. This comes down to reading people, reading reactions, and understanding what each of those 12 people needs to hear to have them act like he wants. Not coincidentally, the most common background of politicians: lawyers.
I'd argue that the most likely scenario here is not closing the divide but widening it. Instead of public defenders, poor people will just get the bot, while those of means will get lawyers. It's akin to tech support today. In the not so distant past, it was easy to get a human on the line. Now in many cases unless you're a "special" client, or get your issue on the front page of some big social media site, you get the bot. And it sucks.
Public defenders would be more effective if they didn't have to talk to their clients?
For a natural text generation/chat system like ChatGPT, at the moment this is effectively impossible and frankly I don't see it being sufficiently fixed for something as complex as teaching law to a non-law-student any time in the near future.
It's not like a "jailhouse lawyer" will do any better. Though it might cost something you didn't expect, such as a new witness against you or a shower pal.
The Constitution is already pushed beyond all reason for poor people anyway. You might have a "right" to a public defender, but good luck getting them to spend much more than 10 minutes on your case. (https://www.mprnews.org/story/2010/11/29/public-defenders) They're so overworked that it can be impossible for them to spend more than a tiny fraction of the time they actually need to do their job (https://www.kansascity.com/news/local/crime/article237036919...).
In the same way you have a "right" to speedy trial, but unless you're rich you can spend years (or even a decade) waiting behind bars (https://www.thecity.nyc/2022/8/17/23310771/why-some-spend-ye...).
It seems like we've already decided that if you can't pay the court fees you aren't entitled to access to our "justice" system (https://digitalcommons.law.seattleu.edu/sulr/vol44/iss3/11/)
I don't think it's too far fetched to imagine the courts telling poor people they have to settle for an AI to represent them unless they can fork over the money to hire one themselves.
If an AI is just as capable (if not more capable) than a real lawyer, we'll be seeing rich people and corporations using them instead of humans. Until we see that, I'll assume that if poor people are forced (by law or finances) to use them they'll be getting the short end of the stick.
1. A scaling army of prosecutors and "professional witnesses" which are compensated/promoted based on conviction rate.
2. A single "freely" provided and extremely overworked defense attorney with three times the caseload of the prosecutor and where every bit of help they give out takes from the time of the other defendants they "defend."
3. As a result of 1&2 you are pressured to sign a plea and threatened with triple and quadruple punishments for the same activities. All while being stuck in a cell (why else would you need a free attorney)? And we are not talking the difference of 3 months and a year. Instead we are talking about the difference of 5 years and 20 years.
4. The result is an overwhelming number of factually innocent people in prison for crimes they never committed. From people who sat down and said, what happens if it goes wrong vs what happens when I do 5 years of adult daycare?
So yes, any kind of AI is likely to improve the situation as the situation is currently completely unacceptable to anyone who has seen a family member go through it.
While I agree with you that AI powered lawyer can have many downsides and drawbacks here you are building on a falsehood.
There is this illusions that computers are "book smart" but not "street smart". It is very deeply ingrained into our culture. Just think about Data from Star Trek. Very analytical, but struggles to understand emotions. Therefore also struggles to manipulate people's emotions, aka being emotionally persuasive.
This is a sci-fi trope, but is false as it comes to these generative models. (or can be false depending on training data) If you train the system with mock juries it can learn how to twist them around its finger. Text written by such a model can be as emotionally persuasive as one written by the best lawyer.
In fact in reality, as the tech is today, it would be easier to reliably implement the "reading reactions, and understanding what each of those 12 people needs to hear to have them act like he wants" part than the "follows flawless logic, legal technicalities to a T" part.
It is of course still a very perilous path to make an AI-powered lawyer. It is just important to not base our reasoning on sci-fi tropes
Come to think about it. This part can be valuable, even as a startup. You get mock juries, you act out legal cases in front of them. You collect every single thing the actor lawyers say, and how they say it, and you also collect full transcript of their deliberations.
You model the data, and then you sell the services to lawyers. They can practice in front of your model jury initialised with what can be known about the real jurors and can see traces of their deliberations. Not just one run of the model, but many runs if they want to.
The solution seems clear - poor people just need to run billion dollar advertising campaigns to improve their support.
This "AI Lawyer" is about a traffic ticket. If we want to make things more fair while also enforcing the laws on the book it seems like the more straightforward tech solution is to improve traffic cameras/monitors such that tickets are given out more appropriately. Or, if it still goes to court, why not just have a system where someone can represent themselves with a form or checklist which points out the reasons they were given a ticket by mistake?
Perhaps this is a good business model in the current system but its definitely not the best for society as a whole. Why do we need lawyer of any type to handle something as banal as speeding tickets?
You should be asking what are the incentives.
> like the more straightforward tech solution is to improve traffic cameras/monitors such that tickets are given out more appropriately
This is not straightforward from a bureaucratic sense. There is no strong incentive (that I can identify - humanitarian ideals do not count) to improve the functionality here to make traffic tickets and their resolutions more sensible.
However, an AI that can resolve traffic tickets and citations might incentive a city to start gathering data more effectively - to win cases and get revenue - and that is a great incentive to improve the general tech.
> Why do we need lawyer of any type to handle something as banal as speeding tickets?
I also want to mention that this is a great point. The fact that my job provides legal insurance, the end result being I can get out of my traffic tickets by hiring a lawyer for free to show up and ask for a dismissal, is pretty dumb.
This is itself a normative statement.
I’m of the opinion that incentives aren’t the only thing worth considering — there’s something to be said for enshrining ethics into law (if only because it can adjust incentives).
Another thing worth considering is our mediocre track record of identifying incentives ahead of time.
Either way the government is not purely incentivized to increase funds. Standard fines for traffic violations are changed up or down all the time. Sometimes cities decide their current traffic enforcement is too burdensome and they cut back on tickets. There could be political actors which have incentives to improve it. Whether or not it will ever happen is not clear but I dont think its impossible by any means.
You don't. Just show up in court and talk to the judge. That's how it's currently handled. Lawyers aren't involved in the traffic courts I've seen.
They basically always offer some bullshit non-moving violation to make me go away. Saves me thousands on insurance premiums and has also saved me from a criminal record for my more serious tickets (like allegedly driving gasp 36 mph by a school, which apparently comes with jail time in Ohio).
Bring on the hate. Y'all know I speak truth.
Most people don't get ticketed often enough for speeding to feel one way or another.
And most people don't feel righteous about speeding in a school zone.
Convicted/not convicted doesn't map perfectly onto guilty/not-guilty. You yourself understand this perfectly or you wouldn't be hiring a lawyer for a moving violation.
"Always"? I have never ended up in traffic court.
> Bring on the hate. Y'all know I speak truth.
Who would doubt that you are the antisocial person you portray yourself as?
This is _caused_ by technology, you usually don't solve a problem by adding more of its cause
If you had 1 lawyer per 100 people in 1800 and you had 1 lawyer per 100 people today, I don't see a scaling problem with population.
Indeed in the UK in 2012 there were 120,000 solicitors for 64 million
Today it's 160,000 for 68 million
That's a 30% increase in lawyers with a 6% increase in population.
Why do we need more lawyers per person today than 10 years ago?
How many more laws are there? Being a lawyer is like 90% reading and collating documents to filter out relevant information. We have a lot more law and a lot more documentation to sift through these days.
Yeah - and I wonder what will happen if it starts telling the defendant to say things which aren't grounded and rooted in fact, or which contradict things they previously said or submitted in writing.
Not sure a generative model is the best way to keep people clear of obvious cases of perjury/lying under oath etc.
To play devil's advocate though, would a layman without AI help be any better at navigating this?
But courts are very likely (in my view) to take a very very dim view of the introduction of some kind of barrack-room "AI lawyer". It will slow things down by introducing potentially spurious arguments which wouldn't have been made by either a represented party, or unrepresented party.
Given the courts are a public service, this isn't necessarily zero sum - unrepresented parties do take longer in court, but I very strongly suspect in the life of a whole case, parties with an aspiring barrack-room AI lawyer will take up more court time. Especially if their argument loses coherency or starts to contradict itself.
... that would be very surprising, since there is no legal definition nor distinction of publisher or platform.
https://www.washingtonpost.com/outlook/2019/07/29/stubborn-n...
https://www.eff.org/deeplinks/2020/12/publisher-or-platform-...
https://www.techdirt.com/2020/10/20/section-230-basics-there...
https://www.nationalreview.com/corner/it-doesnt-matter-if-tw...
https://www.cato.org/blog/newspapers-are-spreading-section-2...
Even TechDirt manages to get this right, but the myth that tech companies exist on the knife's edge of being converted under the law into publishers is stubborn.
I agree the defendant is likely to be on their own, and I'm wondering whether the individual concerned is aware of this. The company seems to be willing to backstop their ticket if they lose.
I'm not sure that will help much if they are sent to jail for a few days for wasting court time for relatedly lying under oath or talking themselves into a generative AI's web of inaccurate statements. Such a follow-up suit would be both hilarious and interesting to follow indeed.
Why would it not be? Court Rooms are public, open to the public with many high profile case being broadcast in real time
Personally I think ever courtroom should have everything broadcast in real time to anyone that wants to view it
Government Secrecy is how abuse occurs
- access to general knowledge: Wikipedia
- safe transport for women at night: Find My Friends / Uber+Lyft
- access to government data: opendata initiatives, a plethora of cadastral map services
I'm going to say the obvious thing everyone is thinking. Everyone who has ever said "this is the core issue: you're trying to solve a social issue with technology" has never solved a social issue _or_ a technological issue and has no idea how to do it.
0: obviously I must clarify for this audience that "solved" in this context means that it has reduced the prevalence of some problem below some low base rate
This was also solved using non-tech and hundreds of years ago. It's called a firearm.
A an aside business has been shown to be the best path to solving societal problems, from medicine, to poverty business is that path. Leave government to solve a problem and as the old saying goes you would get the best iron lung money could buy but you would never get a polio vaccine.
I agree with you, but I also can understand why people may think that profiting from solving a social problem could create a slew of other issues, such as incentivizing the proliferation of that same social problem so that solving it can turn a profit. It's just business!
On the other hand I don't understand this hatred towards tech, aside from "it's new, and I don't understand it, therefore it's bad".
The ironic thing is we see this more with government than business. Social programs almost always NEVER solve the social problem they just manage it.
This is to ensure a voting base is dependent on the government and gives the political class levers to pull to ensure their control and domination.
Government does not want to fix problems, they want to manage them
So why refuse it?
I'd rather say you can't solve social issues with tech ALONE. You need more than that, but you still need it.
Criminal, Civil, Contract, Torts, Family, etc.
In many areas of law I would agree with your statement, however with criminal law I have always felt that many governments have been in the habit of criminalizing poverty in many ways. Personally I think it should be mandated that at a minimum for every $1 given to a prosecutor's office to prosecute criminal cases, $0.50 should be given to public defenders (ideally $1 for $1 but I would take $0.50) currently is more like $0.10.
Though, the ability to use the technology might come down to a social problem. Because it still needs a certain level of education and competence to use the technology. Just significant less than visiting a university for a whole decade or so.
The state can simplify the law and eliminate or stop prosecuting laws that don't directly harm people, thus reducing the demand for lawyers. It's not just an economical problem.
See, the sword cuts both ways. Prosecutors' office would run more efficiently, would pursue much larger caseloads (lower priority ones) with fewer employees and less time. Civil lawsuits could become much cheaper to file; trolling could expand from niches (like patent law) to much broader, more socially-impactful forms of warfare. The financial risk/reward equation is what deters this, and lowering the costs means eliminating the deterrence. And imagine regulatory compliance costs, when regulators feel emboldened to design regimes so elaborate they themselves don't understand them, because they have AI assistants capable of parsing them.
Who's to say DoNotPay won't eventually sell this service to the state, in service of the public interest in accord with the Sixth Amendment? In that light, it may be a tool worth developing
Similarly, the the problem with representation of defendants in the legal system is technological and economic, not social. It’s expensive to have a credentialed person spend lots of time working on behalf of someone else. That’s true for everything from auto repair to law. The social and legal system is simply responding to that economic reality. The public is willing to spend so much money to protect the rights of the accused and there is only so much you can accomplish by haranguing them into wanting to spend more money. By contrast, if you dramatically reduce the cost of providing the accused with a defense, you dramatically change the landscape of the social and political issues.
'Sorry, that wasn't in my training set' won't cut it if the penalty is >> the price of your legal fees, an extended stay in jail or in some countries the death penalty. Obviously for criminal cases you'd be fairly mad to try this kind of construct but that's exactly the ones where the legal fees are absolutely crippling.
I've yet to have a case that I brought that went to trial end up under 50K in costs and that was with what I thought was fairly simple stuff with relatively unsympathetic defendants.
Because that's something Ian Banks would smile at, from wherever he is rest his soul.
Those are not exactly a secret. The entire reason for those loopholes to exist is for the rich to be able to exploit/use them.
The problem is that half the working class wants those loopholes to stay in place because they have this delusional notion that one day, they'll be the ones making use of them as well.
"At no time at any location within the courthouse may any electronic device be used to make a photograph or recording in any fashion, whether audio, video or otherwise."
https://www.lawd.uscourts.gov/possession-and-use-electronic-...
And why is it illegal anyway? If media can attend and report about a case, why can't they record?
I see Chandler v. Florida established that "a state could allow" photos and broadcast for criminal trials.
https://en.wikipedia.org/wiki/Chandler_v._Florida
b) It is illegal because it would result in leaks to the media which would destroy the entire fabric of the judicial system.
Does media coverage on criminal cases destroy the fabric of criminal proceedings?
From the perspectives of both sides (defense counsel and prosecution), I believe this is undesired as it primes potential jury members' opinions outside of court evidence.
First of all, AIs are not beholden to the bar association. They aren't at risk of losing their license to practice. A poor AI will not convince a human judge as well as a human lawyer, leaving the defendant with an improperly executed defense. A great AI might, perhaps, be able to win a case, but then the question would be whether the human judge was able to understand the full implications of the AI's arguments. Only an AI judge would be able to rule impartially against an AI lawyer.
Which I'd assume we would all find to be a horrific path for our already completely fucked up justice system to follow. After all, if AI can be lawyer and judge, it would only make sense to have the laws written by AI legislators, approved by the AI executive branch, and enforced by the AI police and military.
The downside is that I hire a traffic ticket attorney because they've been handling a cases similar to mine, in the same court circuit, almost every day for years. Legal work isn't about knowing the law. It's about knowing how the system works in a particular place and time. ChatGPT was trained on many thousands of various legal cases. But it can't ever really understand how a given court works unless many people have written about that place and posted it to the internet.
I feel like it would be much better for people to just type their case into the ChatGPT prompt and ask "What argument should I use?", and then take that into the courtroom. Ask it what counter arguments to expect, what the standards of evidence are, etc. You can even prompt it with "Pretend you're a traffic court judge. How would you respond to this argument?" Essentially use it as a research tool for people who are brave or foolish enough to represent themselves. Not sure how much benefit there is in having that in the room with you.
(Yes I know in this case the defendant is acting pro se and therefore couldn't appeal on that basis but it's a fun idea to think about).
If I could feed thousands of documents to an algorithm and then ask him stuff like "in which pages the tax returns of the defendant are discussed?" or "how many times is the defendant's wife talked about in a negative connotation?" it would make work much easier.
Input parameters would be a description of the scene and output would be a list of crimes anyone involved has committed along with confidence rating of how well it would succeed in court.
It's already possible for giant corporations to bury smaller actors in paperwork. They can and do crush people with this.
What happens when everyone has this ability?
The legal system is interesting because it operates within its own framework.
Industries can be disrupted from the outside. People switch when the next best thing comes along: horses to cars, gas to electric, wired to wireless.
The legal system can only be (legally) disrupted from the inside, from within itself. If AI tech gums up the legal system itself, how will we use the legal system to fix the legal system? Regime change / revolution is probably not ideal.
https://twitter.com/ggerganov/status/1605322535930941441