There is NOTHING i do in law that AI could not do better, today
| fuchsia patrolman orchestra pit | 05/15/24 | | sadistic rough-skinned lettuce jap | 05/16/24 | | Lavender Garrison | 05/17/24 | | tan incel | 05/15/24 | | contagious vengeful jewess | 05/15/24 | | rusted passionate crackhouse | 05/17/24 | | Salmon Swashbuckling Feces Place Of Business | 05/15/24 | | contagious vengeful jewess | 05/15/24 | | fuchsia patrolman orchestra pit | 05/15/24 | | contagious vengeful jewess | 05/15/24 | | fuchsia patrolman orchestra pit | 05/15/24 | | contagious vengeful jewess | 05/15/24 | | fuchsia patrolman orchestra pit | 05/15/24 | | contagious vengeful jewess | 05/15/24 | | Glittery judgmental trump supporter | 05/15/24 | | contagious vengeful jewess | 05/15/24 | | Sticky Newt Ticket Booth | 05/15/24 | | Poppy step-uncle's house multi-billionaire | 05/16/24 | | bronze center turdskin | 05/16/24 | | bronze center turdskin | 05/16/24 | | Splenetic transparent circlehead hospital | 05/15/24 | | 180 philosopher-king | 05/15/24 | | Splenetic transparent circlehead hospital | 05/15/24 | | 180 philosopher-king | 05/16/24 | | fuchsia patrolman orchestra pit | 05/15/24 | | tripping stimulating cruise ship | 05/15/24 | | fuchsia patrolman orchestra pit | 05/15/24 | | bronze center turdskin | 05/15/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | fuchsia patrolman orchestra pit | 05/16/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | tripping stimulating cruise ship | 05/16/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | tripping stimulating cruise ship | 05/17/24 | | bronze center turdskin | 05/15/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | bronze center turdskin | 05/16/24 | | Racy anal stain coffee pot | 05/16/24 | | bronze center turdskin | 05/16/24 | | Racy anal stain coffee pot | 05/16/24 | | white unholy regret | 05/16/24 | | bronze center turdskin | 05/16/24 | | Spruce charismatic hall | 05/16/24 | | bronze center turdskin | 05/16/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | bronze center turdskin | 05/16/24 | | contagious vengeful jewess | 05/16/24 | | bronze center turdskin | 05/16/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | bronze center turdskin | 05/16/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | bronze center turdskin | 05/16/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | bronze center turdskin | 05/16/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | bronze center turdskin | 05/17/24 | | Splenetic transparent circlehead hospital | 05/17/24 | | Citrine excitant background story | 05/15/24 | | Brilliant Hairraiser Base | 05/16/24 | | nighttime dull school cafeteria | 05/15/24 | | contagious vengeful jewess | 05/15/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | Startling sable dog poop | 05/15/24 | | milky deranged quadroon | 05/15/24 | | Startling sable dog poop | 05/15/24 | | milky deranged quadroon | 05/16/24 | | fuchsia patrolman orchestra pit | 05/16/24 | | bronze center turdskin | 05/16/24 | | Spruce charismatic hall | 05/16/24 | | fuchsia patrolman orchestra pit | 05/16/24 | | Spruce charismatic hall | 05/16/24 | | fuchsia patrolman orchestra pit | 05/16/24 | | Spruce charismatic hall | 05/16/24 | | contagious vengeful jewess | 05/16/24 | | white unholy regret | 05/16/24 | | Useless forum mexican | 05/16/24 | | bronze center turdskin | 05/16/24 | | Splenetic transparent circlehead hospital | 05/16/24 | | nighttime dull school cafeteria | 05/17/24 | | white unholy regret | 05/20/24 | | Citrine excitant background story | 05/20/24 | | Splenetic transparent circlehead hospital | 05/21/24 |
Poast new message in this thread
Date: May 15th, 2024 9:05 AM Author: fuchsia patrolman orchestra pit
Realized on the first iteration of GPT that, if it were trained properly, it could generate all the written argument based on caselaw, statutes, and facts that I generate, at 10000x the speed. The hallucinations and errors of law are simple programming defects that could be fixed by programmers promptly if they cared to do it.
With simple software, it could file, manage timelines, and manage the case.
The only space I saw for human lolyer was client hand-holding, counseling session bullshit. But watching the 4.o stuff this week, with the flirty and responsive female voice, Im now convinced that GPT could talk to my clients and make them feel fine and safe.
If my kind of lolyer job exists in 20 years it will only be due to some kind of government protectionism rules.
But the real takeaway is that not only is "lawyer" replaceable, the NEED for a lawyer is replaceable. AI could adjudicate legal disputes in seconds to some Pareto optimality without any courts at all.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47662884) |
Date: May 15th, 2024 9:12 AM Author: Salmon Swashbuckling Feces Place Of Business
And the truth is most lawyers are short and/or bald. No one likes a “Tall” lawyer if he looks like Moby!
In 40 years there will be actual fake humans that look like Brad Pitt trying to take your job who can perform your job 10000x better. “Be the Law” is no longer a principle!
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47662890) |
Date: May 15th, 2024 9:12 AM Author: contagious vengeful jewess
"But the real takeaway is that not only is "lawyer" replaceable, the NEED for a lawyer is replaceable. AI could adjudicate legal disputes in seconds to some Pareto optimality without any courts at all."
There's one likely fallacy. Do you think clients ultimately WANT Pareto optimality? In many cases we could get to Pareto optimality on our own right now, as between two or twenty lawyers working something out. But that's not what many clients -- especially wealthy ones, with money to pay fees -- are looking for.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47662891) |
|
Date: May 15th, 2024 9:16 AM Author: fuchsia patrolman orchestra pit
thats a good point. say there's a certain risk/reward to litigating a dispute. Some % of litigants will accept that risk, some % will just say, no give us pareto optimality.
but the risk-takers generate tremendous social cost. We all have to pay for this fucker who thinks it's worth it to sue for 100x his real damages, or worth it to push his DUI through a jury verdict, etc. Worth it to contest child support and tie up family courts for 8 years because fuck his bitch ex-wife.
Collective rationality will win out and shut down the risk takers.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47662898) |
|
Date: May 15th, 2024 9:19 AM Author: contagious vengeful jewess
Okay but that's not the Court system as we know it (and what is enshrined in the Constitution and in our system of government through collective winnowing of bad ideas against good over the course of 1,000 years in England). Instead, you are describing an outgrowth of some sort of "benevolent" totalitarianism, if there is such a thing.
ANOTHER THOUGHT: Much (but not all) of the cost is already absorbed by the litigants or their agents under current system. For the most part, risk-takers are already forced to foot their own bills. So I think this nullifies your point above to some degree.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47662900) |
|
Date: May 15th, 2024 9:24 AM Author: fuchsia patrolman orchestra pit
the entire history of the US has been a movement away from individual liberties towards centralized governance for the collective.
go tell a virginian in 1800 that his horse cannot shit somewhere, or an iowan in 1840 that he cannot build his house that way
geting to a trial verdict is the most inefficient process imaginable, and it rests entirely on some 18th century notion of individual liberty.
things will start, as they always do, with the promise of being optional. Get pulled over for drunk driving? Click here to be immediately registered in probationary classes, pay your fine, activate the breathalizer that comes standard in the car. OR click here to receive a court date. maybe you win! maybe you go to jail!
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47662913) |
|
Date: May 15th, 2024 9:28 AM Author: contagious vengeful jewess
Seems to be a collateral point. That is, just because you see a trend doesnt mean that trend that you see will be implemented in the context of legal services collectively which is the subject of discussion.
Also, I personally don't subscribe to giant trends theories of history that way. Usually one can argue quite the opposite with other facts/examples. Those sweeping trends guys with their cycles and epicycles and ebbs and turns and flows always strike me as confusing a metaphor with the real thing/reifying and also being a little mentally infantile or undeveloped because they can't see their error (I'm not calling you infantile--but that's one way in which those dorks like Moldbug (trivial) and Marx (nontrivial) have the whole thing wrong. It's also why a good college education and continued learning thereafter is so important epistemically).
Finally, I think your "maybe you win maybe you go to jail" bit is supposed to be a parody and it's from a movie or something like that if I recall. That's fiction and so not really serviceable for the point you were trying to make.
All that being said, we can probably agree that some aspects of legal work will be replaced by AI in the next 50 years, but I don't think it will be quite as we imagine it, either the AI involved, its goals and systems and means of arriving at a conclusion, or the extent of our obsolescence.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47662918)
|
Date: May 15th, 2024 10:57 AM Author: Splenetic transparent circlehead hospital
“ The hallucinations and errors of law are simple programming defects that could be fixed by programmers promptly if they cared to do it.”
What are the specific errors and why hasn’t anyone “cared” to disrupt a billion dollar industry when there are tons of legal AI products that suck?
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47663158) |
|
Date: May 15th, 2024 1:37 PM Author: fuchsia patrolman orchestra pit
well wang, i dont know. My guess is that they have not trained gpt on caselaw and to do so would require paying lexis or wl or some other database and gpt-fags really arent focused on that right now
right now, or at least the last time I tried in 2023, gpt will construct an appropriate legal argument based on an imaginary case it invented.
is your position that gpt would be somehow unable to digest the entirety of say, California caselaw and statutes, and incapable of instantaneously applying the appropriate case / law to a set of facts?
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47663871) |
|
Date: May 15th, 2024 2:17 PM Author: fuchsia patrolman orchestra pit
idk anything about biglaw, but a good chunk of shitlaw is the client investing in the lawyer's vibe / physiog / brand.
but as soon as that cracks, there's no value for some lawyer to add.
"your case needs the human touch!" yeah, okay, until I price shop and AI law is a fraction of the price / takes a smaller fraction of the settlement
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47664012) |
|
Date: May 15th, 2024 2:21 PM Author: bronze center turdskin
99% of biglaw could be automated away by AI.
in fact, 99% of biglaw doesn't even need to happen.
the dudes lisping about "why hasn't a billion dollar industry been disrupted then?" is a retarded libertarian faggot. most of biglaw is purposefully inefficient and time consuming because we charge by the hour. do you think there's really a lot of "value add" for the client to have a team of a lawyers going through a routine filing that no one will ever read to make sure even comma is perfect?
ljl "why hasn't it been disrupted."
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47664025) |
|
Date: May 16th, 2024 10:40 AM Author: fuchsia patrolman orchestra pit
you didnt respond at all to my point.
is your position that an ai somehow could not integrate the body of case law and statutory law and brief the merits of a fact-pattern position? and do it better and quicker than a human lawyer?
i cannot imagine what would justify answering, "no AI would not be able to do that", but Im sincerely open to an answer.
the only one I see in your post is "well they havent done it YET haha, and it would get them rich, therefore they cannot!"
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47666473) |
|
Date: May 17th, 2024 10:27 PM Author: tripping stimulating cruise ship
At the limit, it has to create internal models of the world and human cognition in order to predict text optimally. You can’t minimize loss on arbitrary text if you don’t have models for how people think. Superficial linguistic patterns only go so far for predicting the next character. If I provide you a legal argument and then ask you for a plausible text continuation, you will need to know legal concepts in order to do it well.
Theoretically you could imagine this not working on a finite training set. A neural network could simply memorize its training set and not create such models, which would provide no generalization to out of sample text. In practice this doesn’t seem to happen, likely for a variety of reasons - simpler encoding functions make up more of the parameter space, so there’s an in built Occam’s razor/simplicity prior. Stochastic gradient descent also tends to disturb circuits that only memorize a specific training example, so it tends to select for more general circuits that fit many data points. The training sets used are also enormous in comparison to human training, which helps as well. Optimize hard on predictive loss and it just works.
I don’t know what to tell you if you think these are just regurgitation algorithms. These models would be unusable if that was the case. You wouldn’t be able to describe a novel program in natural language and have it create functioning code. You wouldn’t be able to have it play chess unless it learned the board and rules. It wouldn’t be able to answer reading comprehension questions on a piece of text not its training set. It wouldn’t be able to create rhyming poems about shit that people never write poetry about. The input space is far too large for the training set to cover all of the possibilities and clearly they go beyond that.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47670988) |
Date: May 15th, 2024 1:52 PM Author: bronze center turdskin
"If my kind of lolyer job exists in 20 years it will only be due to some kind of government protectionism rules."
this applies to almost every job. and it's been this way for a while. most jobs are complete makework bullshit that exist only because we don't have the will to just pay people to exist. ai is advancing so rapidly that we will eventually be confronted with this more dramatically.
if you don't have some sort of job that will be protected, or a trade or other job that won't be replaced by robots in 20 years, you are going to have a bad time.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47663916) |
|
Date: May 16th, 2024 10:53 AM Author: bronze center turdskin
i'm not some techno-futurist. in fact i've been a naysayer when it comes to shit like crypto and tesla for years.
but right now, today, almost all of my work could be completed 90% faster with with chatgpt. this is what i've heard from plenty of other lawyers. i know your first reaction will be "haha ur not even X!" as you did already itt, but several people here know me irl, so why would i lie? especially when i've been forthcoming with other shit i knew xo would mock.
i don't know what you do or who you are. all i can say that in my corner of the law we could eliminate 90% of the workforce tomorrow. in fact, i've been in meetings regulators where they have forbidden the implementation of these type of technologies specifically because of the job loss.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47666503) |
|
Date: May 16th, 2024 3:34 PM Author: bronze center turdskin
right, but only because the type of information you're looking for there is specialized. it's basic to you but very niche to a AI model.
meanwhile if you plugged in a law exam hypo it would generate at answer likely to book the class at a TTT. spitting out a decent explanation of constitutional law and applying it to a fact pattern is something most lawyers will never achieve, even though almost all will manage to find out how to file a complaint.
the AI models do a better job on a task more difficult for humans because more humans right now are interested in conlaw than filing a complaint in a specific state. but once the pool of info that these models draws from becomes wider, it's clear that the potential for it to answer simpler questions than it does already currently exists.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47667592) |
Date: May 15th, 2024 9:37 PM Author: nighttime dull school cafeteria
I think for transactional attorneys it will be really helpful but people who say it will replace us don't really understand what lawyers do.
We already just copy and paste shit from precedent.
Our jobs are mostly reading the boring documents and making sure they're okay. You'll still need an actual person to do that even if AI generates it.
We also have to understand the business and the deal, issue spot, discuss strategy with the business, and then negotiate solutions.
AI isn't going to be talking to the opposing party's AI saying "We don't have $10 million in cyber insurance" "Well we can't do the deal unless you do" "Why don't you cover half the price of our policy and we can agree to procure it." "We can do that but then the term has to go to 3 years" "Okay fine"
And when it gets to that point, what good are humans at all in any job?
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47665633) |
Date: May 15th, 2024 9:50 PM Author: Startling sable dog poop
I would never let my kids go to law school. I think all of the kids going to law school now and coming out s baby lawyers are almost all universally screwed. Some will be scrappy and maek it, but the industry has been changing in multiples in the past 20 years than it did in the preceding 100. Those trends are only accelerating.
There will be a need for lawyers that help actual people, but the need for lawyers in corporate, financial transactions, and other traditional biglaw things will certainly be diminished.
Think less BIGLAW and a lot more shitlaw. Lawyers will be viewed like car mechanics in 25 years, whereas car mechanics' jobs keep getting more advanced, so they will be revered and respected in the new economy.
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47665681) |
|
Date: May 16th, 2024 10:08 AM Author: fuchsia patrolman orchestra pit
spaceporn, above, discounts larger historical trends as unreliable, and there's some argument there.
But the trend towards mechanical efficiency has been storming forward since Henry Ford at least, and lawyering depends entirely on inefficiency.
for the record, I think the cult of efficiency is demonic and will force a reckoning. Efficiency means 99% of humans are worthless. Efficiency should serve people, not people efficiency.
But we havent been forced to adjudicate that yet. It's coming in are lifetimes tho
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47666418) |
|
Date: May 16th, 2024 10:21 AM Author: fuchsia patrolman orchestra pit
i dont see a better career than geology.
male dominated
often includes outdoor work
science based, but critical to global economy, oil, energy, etc
rocks are 1800000000
can't get my kid interested tho
(http://www.autoadmit.com/thread.php?thread_id=5528722&forum_id=2#47666442) |
|
|