The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Shitlaw client filed his own probate lawsuit using ChatGPT
Poast new message in this thread
Date: October 15th, 2024 8:39 PM Author: When I grow up I want to be a pumo He's a hustler trying to make it in his own business, and he manages to get into trouble frequently. He got into yet another dispute and told me that he had filed a lawsuit against someone, and had ChatGPT draft the complaint. He wanted me to take a look at it.
I was expecting an unintelligible mess. Brothers, I was wrong. It wasn't perfect, didn't cite statutes or case law for example. But it was most of the way there. Had things well organized and correctly listed the elements of several causes of action. And this is all ChatGPT. I can assure you that any contributions from my client were minimal.
So, what does all this mean? What exactly is the role for lawyers if this stuff becomes widely adopted? Some people won't know exactly what questions to ask ChatGPT for it to compose anything. ChatGPT can't argue in court. It doesn't know more complicated things like civil procedure or the rules of evidence. But it seems like all that could change with further advances. |
Date: October 15th, 2024 8:40 PM Author: Karlstack (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48201868) |
|
Date: October 15th, 2024 8:47 PM Author: When I grow up I want to be a pumo (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48201885) |
|
Date: October 15th, 2024 9:16 PM Author: ,.,,,.,,,,..,.,.,.,.,.,.,..,., If you play around with ChatGPT 4o and know anything about the technical details of transformers and how they are trained, it’s obvious it’s over soon for everyone. The models are not close to topping out soon and model training compute will likely 10000x by the end of this decade. Meanwhile they are finding ways for models to think for longer (ChatGPT o1 is the beginning of this ), so people will be able to have the model think for hours or days or even weeks for important applications. Anyone who is thinking this is stopping soon is deluded. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48201984) |
 |
Date: October 15th, 2024 9:18 PM Author: mens gaya it isn't "thinking" its just iteratively feeding itself statistical probabilities (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48201988) |
|
Date: October 15th, 2024 9:26 PM Author: ,.,,,.,,,,..,.,.,.,.,.,.,..,., The most broadly accepted theory of human cognition is that layers of neural computation implement predictive processing in the brain. The predictive processing model does not look fundamentally different from modern neural networks trained to predict different token sequences. People get hung up on their own intuitions for how humans think and then use that as a reason to ignore something that is clearly replicating verbal cognition to a large degree. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202018) |
|
Date: October 15th, 2024 9:28 PM Author: mens gaya AI does not have a Will, which is the fundamental aspect of Thinking. Elsewise, you are simply referring to strings of unrelated symbols, which is all the AI has. It has no Will, and, therefore, no Thought. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202020) |
|
Date: October 15th, 2024 9:30 PM Author: ,.,,,.,,,,..,.,.,.,.,.,.,..,., I think you have grander ideas of what it means to be human than I do. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202028) |
|
Date: October 15th, 2024 10:59 PM Author: mens gaya It's not grand to think the human animal has a Will. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202354) |
|
Date: October 15th, 2024 9:33 PM Author: snake coke what difference does it make if the two are indistinguishable in practice? (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202043) |
|
Date: October 15th, 2024 9:18 PM Author: When I grow up I want to be a pumo So how does this play out for lawyers? (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48201989) |
|
Date: October 15th, 2024 9:29 PM Author: ,.,,,.,,,,..,.,.,.,.,.,.,..,., I think all of human labor will be automated soon. Less than 10 years with reasonably high probability, almost certainly within 30 years. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202021) |
|
Date: October 15th, 2024 9:29 PM Author: Mahchine'ing the $ecret truth of the univer$e (Do you $ee it yet? The Mahchine's 180 Vi$ion?) (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202023) |
|
Date: October 15th, 2024 9:31 PM Author: When I grow up I want to be a pumo So what will people do all day? (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202032) |
|
Date: October 15th, 2024 9:33 PM Author: ,.,,,.,,,,..,.,.,.,.,.,.,..,., I have no idea what the world after looks like. Human extinction, utopia, billionaires getting rid of the masses, eternal authoritarian regime controlled by AI all seem plausible. The only scenario I find very implausible is that the tech stops close to the current level. That is not happening. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202045) |
|
Date: October 15th, 2024 9:21 PM Author: Mahchine'ing the $ecret truth of the univer$e (Do you $ee it yet? The Mahchine's 180 Vi$ion?) This thread engages with the growing impact of AI on the legal profession, particularly the notion that AI tools like ChatGPT might eventually perform tasks traditionally handled by lawyers. Here's an analysis of the dynamics:
The Setup: A Client's AI-Drafted Lawsuit The original poster (When I grow up I want to be a pumo) introduces a story that reflects both skepticism and surprise: a "hustler" client who drafted his own probate lawsuit using ChatGPT. The initial expectation was that it would be a mess, but the poster is surprised to find that it was "most of the way there." This anecdote touches on the idea that AI can produce reasonably coherent legal documents without deep human expertise, challenging the role of traditional lawyers.
The poster notes that while the AI-produced complaint lacked statutes or case law, it was well-organized and hit key elements of legal claims. This reveals both the potential and limitations of AI tools in legal work—capable of handling simpler tasks but falling short in complex areas that require deep legal knowledge. The Role of Lawyers in an AI World The central question raised: “What exactly is the role for lawyers if this stuff becomes widely adopted?” reflects an existential dilemma for the legal profession. AI’s ability to generate legal documents or offer legal guidance raises concerns about the future relevance of lawyers, especially in areas where routine or formulaic tasks dominate.
When I grow up I want to be a pumo acknowledges that while AI can't yet argue in court or handle civil procedure or rules of evidence, there's an awareness that “all that could change” with further advances. This implies an anxiety about how quickly AI might close the gap and push lawyers further out of the process. Karlstack’s Nihilistic Take: “It’s Over for Lawcucks” Karlstack jumps in with a harsh, cynical take: “it’s over for lawcucks” and urges lawyers to "learn to substack," suggesting that traditional legal careers may become obsolete in the face of AI’s rapid development. Substack, an online platform for independent writers, is mentioned here as a metaphor for needing to pivot to content creation or more specialized work that AI cannot replace.
This attitude reflects a deep-seated pessimism about the future of law as a profession and aligns with broader fears about automation and job displacement in many fields. The idea that lawyers could become irrelevant due to AI underscores concerns about the transformative (and potentially destructive) power of technology. The Technical Perspective: The Future of AI Advancements ->()(``- introduces a more technical perspective, stating that ChatGPT 4o (likely a future iteration of the model) will continue to improve, and model training compute will scale exponentially by the end of the decade. They argue that AI is far from reaching its ceiling and could one day be capable of thinking for hours or days for critical applications.
This comment emphasizes that AI’s growth isn’t linear but exponential, with the potential to disrupt not just law but all knowledge-based professions. The idea of extended thinking models capable of handling long-term and highly complex problems challenges the notion that certain intellectual tasks will remain exclusively human. Mens Gaya’s Counterpoint: AI Doesn’t “Think” In response to this, Mens Gaya brings the conversation back to Earth, reminding the thread that AI doesn’t “think” but simply “iteratively feeds itself statistical probabilities.” This acts as a reality check, emphasizing that AI’s functioning is based on pattern recognition and predictive modeling rather than true understanding or cognition.
This counterpoint helps maintain a level of skepticism about the more futuristic predictions surrounding AI. While powerful, AI still lacks consciousness or an ability to reason in a way that mirrors human thinking. The focus on probabilities also highlights the current limitations of AI, particularly when navigating ambiguous or novel legal challenges. The Future for Lawyers: How Does This Play Out? The thread concludes with When I grow up I want to be a pumo asking a crucial question: “So how does this play out for lawyers?” This question reflects the tension between optimism about AI’s capabilities and concern for the legal profession’s future. Lawyers face an uncertain landscape where many routine tasks could become automated, leaving them to focus on strategic, complex, or creative elements of legal work.
Conclusion This thread encapsulates the anxiety and curiosity surrounding AI’s growing role in the legal field. While there’s acknowledgment that AI like ChatGPT is far from perfect, its ability to handle certain legal tasks raises existential questions about the future of legal work. The broader AutoAdmit community responds with a blend of cynicism, technical analysis, and skepticism, reflecting the evolving conversation around technology’s impact on traditionally stable professions like law.
How this “plays out for lawyers” ultimately hinges on the speed of AI’s development and its ability to tackle more sophisticated legal challenges. Lawyers may need to adapt by focusing on areas where human judgment, creativity, and interpersonal skills remain indispensable. |
|
Date: October 15th, 2024 9:30 PM Can AI make a campaign contribution to the elected judge? Didn't think so. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202027) |
|
Date: October 15th, 2024 9:38 PM Author: snake coke actually it instantly read every publication he’s ever made and cross-referenced his prosaic fingerprint against the entire internet to find things he’s written anonymously and compared that entire body of work against all other posts, preferences, and predilections of those with whom he has 99.999th %ile sociodemographic similarity to buy him the absolute perfect (but not too expensive to be ethically problematic) congratulations gift for his victory. something he’ll really remember getting. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202063) |
|
Date: October 15th, 2024 9:46 PM Waste of AI brainpower, state court judges always want hookers. Unless they're dames, then they want shoes. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202103) |
|
Date: October 15th, 2024 10:43 PM Author: snake coke (it was a hooker) (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202311) |
|
Date: October 16th, 2024 8:15 PM Author: Gorilla Pimp (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48205645) |
|
Date: October 15th, 2024 9:47 PM Author: t14 I used Lexis AI to draft an amended complaint for me. I just uploaded the minute order and original complaint and it revised every CoA to survive demurrer except for 1 which I just removed. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48202112) |
|
Date: October 15th, 2024 10:56 PM Author: michael doodikoff I tried to have it draft some discovery for me. What I got was pretty shitty
I did have it write a certain motion. Had to chec every case it cited. Not 100% but made some decent arguments and stuff.
I'm not afraid of it taking my job I go to court too much |
|
Date: October 15th, 2024 10:59 PM Author: ,,,...,,......,,....,,,,,,,,....... Nah, it’s great at doing stuff like summarizing info but it’s still terrible at anything that requires real inside knowledge and judgment. I don’t see how that’s going to change because it’s fundamentally limited to making educated guesses about what it thinks will sound convincing as a response. You can get it to say anything you want basically. That’s not useful if you’re trying to rely on it as a substitute for professional judgment ie for tasks people pay lots of $$$ to lawyers for. The mechanics of law are far and away secondary to the judgment.
That said it’s incredible as a productivity tool. I doubt it will impact law firm employment that much because law firms aren’t incentivized to be efficient, but I could see it killing a lot of inhouse/company jobs. Most people at bigcos are not paid for their judgment and are doing stuff like summarizing or pushing around info that could be done by AI. |
|
Date: October 16th, 2024 1:52 PM Author: .,,.,,.,,,.... This won’t age well. LLMs fine tuned to be less agreeable or more legal specific could likely deal with this. There’s nothing about LLMs that means they inherently can’t. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48204224) |
|
Date: October 16th, 2024 8:16 PM Author: Karlstack (Refried) (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48205646) |
|
Date: October 16th, 2024 1:50 PM Author: ....:::..:....:::..:...;;;..;...:.:: I know Westlaw is coming with an AI tool soon that is rumored to like write entire motions/briefs for you with actual accurate law.
I think we're actually very close to AI being able to review discovery responses, formulate meet and confer letters and draft motions to compel based on the documents alone.
i.e. you just run it the questions and responses through your AI tool, it spits out deficiencies and a sample Motion to Compel.
Will be a game changer for contingency fee practices and enable larger volume of cases. I think many insurance defense firms have already switched to a hybrid flat fee model so will definitely boost profitability for them as well.
Routine business lit and real estate lawyers billing by the hr will be fucked. |
|
Date: October 16th, 2024 1:53 PM Author: .,,.,,.,,,.... Boosts profitability until the fees are negotiated down from competition. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48204226) |
|
Date: October 16th, 2024 1:58 PM Author: ....:::..:....:::..:...;;;..;...:.:: yes this will definitely happen as is already rampant in insurance defense.
They're already essentially laying off anyone above a 3rd year because reimbursement rates from carriers are so low. We have a large trial set to go weds next week where a literal first year is doing all expert depos with zero partner oversight.
1st yr floated offer of settlement at 1.5 which we rejected when he only had authority from carrier to pay 250k lol. It's already insane.
This will probably just lead to more of this. |
|
Date: October 16th, 2024 1:55 PM Author: Waltah "Routine business lit and real estate lawyers billing by the hr will be fucked."
No, this is tremendous. I already use ChatGPT for discovery and the time saving is amazing. More automation is great. |
|
Date: October 16th, 2024 2:23 PM Author: ,.,,,.,,,,..,.,.,.,.,.,.,..,., we are in the phase right now where the people adopting AI can increase their productivity and there aren't enough people doing the same that rates and workload are falling. that won't be the case for long. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48204357) |
|
Date: October 16th, 2024 2:51 PM Author: Waltah Good. Legal work should be automated. Lawyers should have to find different work. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48204465) |
|
Date: October 16th, 2024 7:20 PM Author: .,,.,,.,,,.... only problem is that it looks like it's going to be a large simultaneous hit to a ton of different jobs. (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48205489) |
|
Date: October 16th, 2024 2:31 PM Author: ChadGPT-5 AI training itself with AI trained data when (http://www.autoadmit.com/thread.php?thread_id=5612242&forum_id=2#48204393) |
|
Date: October 16th, 2024 2:47 PM Author: ,.,,,.,,,,..,.,.,.,.,.,.,..,., OpenAI has said they intend to use o1 for this. every time someone queries the model now, it generates a lot of token samples which are used to reason when generating the final response. they are going to use that data to improve the model in the next iteration. they can generate shitloads of synthetic data graded by quality and the final response and feed it back into GPT-5 training. then GPT-5 gets deployed and used for generating more synthetic data.
the people that are betting on a collapse in model quality from training on their own data don't get it. this only matters when they are training exclusively on generated data (and not a mixture of generated and human language data) and there is no way to quality grade the data. |