ChatGPT just wrote me a POA that is better than the attorney's

[RationalWalk]

I've been fencing with an attorney for about a month now trying to get a simple, direct, and adequate Limited Real Estate POA written. Maybe it's just the attorney. But I just got the idea to ask ChatGPT to write one for me, and in about 2 seconds there was a nicely crafted document. In many ways, better than the one the attorney has come up with. Yes, this is a pretty simple legal document as things go, and yes, it needed some additional polishing but it was a pretty good effort for $0 and 2 seconds of my time. All I'd need now is to have an attorney put on the finishing touches, if I could get one to do that. I can certainly see how AI in the future is going to be able to handle stuff like this better, faster, and a whole lot cheaper. And be able to handle much more complex documents using an interrogation or interactive mode. Next I'm going to ask it to write me a Will and see what it does. If I were an attorney, I think I'd be figuring out how I'm going to live in the world of AI. Has anybody else had this experience recently?

[simplextableau]

The fundamental problem is if you had the expertise to evaluate the ChatGPT work product, you could have written your own POA. And since you don't have that expertise, you still need a lawyer to evaluate and very likely fix the document. So ChatGPT created something that superficially looks good but didn't advance your situation any.

[jarjarM]

May want to read about ChapGPT's (or LLM in general) issue with hallucination first. Here's a NY time article on a recent incident with attorney and hallucination-created fake court citings

[Tdubs]

The fundamental problem is if you had the expertise to evaluate the ChatGPT work product, you could have written your own POA. And since you don't have that expertise, you still need a lawyer to evaluate and very likely fix the document. So ChatGPT created something that superficially looks good but didn't advance your situation any.

Why should coders be the only ones to lose their jobs? I think AI will figure out these legal bumps in the road in good time.

https://www.newyorker.com/magazine/2023 ... -the-craft

[yolointopants]

I've been fencing with an attorney for about a month now trying to get a simple, direct, and adequate Limited Real Estate POA written. Maybe it's just the attorney. But I just got the idea to ask ChatGPT to write one for me, and in about 2 seconds there was a nicely crafted document. In many ways, better than the one the attorney has come up with. Yes, this is a pretty simple legal document as things go, and yes, it needed some additional polishing but it was a pretty good effort for $0 and 2 seconds of my time. All I'd need now is to have an attorney put on the finishing touches, if I could get one to do that. I can certainly see how AI in the future is going to be able to handle stuff like this better, faster, and a whole lot cheaper. And be able to handle much more complex documents using an interrogation or interactive mode. Next I'm going to ask it to write me a Will and see what it does. If I were an attorney, I think I'd be figuring out how I'm going to live in the world of AI. Has anybody else had this experience recently?

Your attorney is already using stock forms that get modified to your specific situation. So this basically is you providing the boilerplate to the attorney instead of the attorney using her boilerplate already.

[billthecat]

I don't trust ChatGPT at all. I gave it my annual contributions to my HSA over the last 7 years, and the current value, and asked it to calculate the average annual growth. It lied. It showed the calculation (it started out right), but the end result was incorrect, wildly off. I simply said, "Isn't it __?" and it said, "You're right, sorry blah blah blah". The formula was right, but the calculation was wrong - if it can't get that right, how can I trust it about anything?

It's also amazingly biased - it will happily tell a joke about men, but if you ask it to tell a joke about women it will say it won't tell a joke about groups or whatever. If you then point out that it just did about men, it will acknowledge it, apologize, and say it will try to do better.

That said, you could have used ChatGPT to create the initial version containing the key points you had in mind, then give it to the attorney and he could then build off of it from there.

[Supergrover]

I would not trust a form generated by chat GPT. Does it know all the exceptions, special circumstances, etc of each state or instance when the form is used? Nope.

[chickadee]

I asked ChatGPT to help plan an itinerary and it couldn’t count. I said plan a trip for 8 nights between city X and Y. It planned a 7 night trip. I told it the mistake, and it fixed it. Then I said, please number the days and tell me where I’m sleeping each night. It also got that wrong, labeling the third night in city Y as only the second night in that city.

If ChatGPT can’t count, what should we trust it to do?

[Kickstart1967]

I think lots of younger lawyers are already using it for lots of tasks
Colleagues of mine use it to summarise cases, draft skeleton arguments etc
I am told it can be a bit clunky but improves quickly

[protagonist]

What about this idea?
1. Have ChatGPT write up the legal document, then
2. You review it for obvious errors, then if it looks good,
3. You have an attorney review and edit it (without telling the lawyer that it was AI-generated to avoid prejudice).
4. You do a final review.

The best of both worlds, less chance for error, and you might save money....sort of like how in medicine the intern does the exam and it is then reviewed by the attending MD.

[Eric]

What about this idea?
1. Have ChatGPT write up the legal document, then
2. You review it for obvious errors, then if it looks good,
3. You have an attorney review and edit it (without telling the lawyer that it was AI-generated to avoid prejudice).
4. You do a final review.

It takes more time for an attorney to review someone else's document than to generate their own. The problem is that an attorney can't trust an outside document -- he or she has to review every single line. By contrast, the attorney will already know his or her own form, almost by heart.

[rob]

The fundamental problem is if you had the expertise to evaluate the ChatGPT work product, you could have written your own POA. And since you don't have that expertise, you still need a lawyer to evaluate and very likely fix the document. So ChatGPT created something that superficially looks good but didn't advance your situation any.

Why should coders be the only ones to lose their jobs? I think AI will figure out these legal bumps in the road in good time.

https://www.newyorker.com/magazine/2023 ... -the-craft

Contracts like this are far easier than code... Why do you think the lawyer orgs are going after that company that started using LLM's for people - they have to go after the LLM for "practicing" law without license etc. to protect their profession. It's for the most part easily repeatable.

[Tdubs]

The fundamental problem is if you had the expertise to evaluate the ChatGPT work product, you could have written your own POA. And since you don't have that expertise, you still need a lawyer to evaluate and very likely fix the document. So ChatGPT created something that superficially looks good but didn't advance your situation any.

Why should coders be the only ones to lose their jobs? I think AI will figure out these legal bumps in the road in good time.

https://www.newyorker.com/magazine/2023 ... -the-craft

Contracts like this are far easier than code... Why do you think the lawyer orgs are going after that company that started using LLM's for people - they have to go after the LLM for "practicing" law without license etc. to protect their profession. It's for the most part easily repeatable.

The article I linked is by a coder who actually has hope for the future of his profession, but he makes it obvious that the boilerplate responsibilities will go away.

So much of the legal profession's bread and butter is boilerplate.

[nisiprius]

This reminds me sharply of a time Way Back When, in the days of dedicated word processors. I worked in a research institution that had an "editorial services" department to help scientists, many of whom were not native English speakers, prepare their papers for publication. Typically they submitted typed material to the department. Operators would key it in manually.

Lanier (major word processor provider, pre-PC) wanted to sell them an OCR scanner to avoid the work of rekeying. The department head asked them "Can you guarantee me 100% accuracy?" They said, proudly, that it had 98% accuracy.

She said "Then it is no value to us. We spend most of our time on proofreading, not data entry. If it is only 98% accurate, then we still need to proofread it. The only way there is a savings is if it is so accurate that we can eliminate the proofreading step."

And so it is with legal documents. A legal document that is 98% correct has zero value, and you will need to pay a lawyer as much money to review it closely enough to spot the errors as you would to do it themself.

In a way, ChatGPT is worse because the output is so smooth and plausible that a reviewer becomes fatigued and numb and lets their guard down.

[nisiprius]

...But I just got the idea to ask ChatGPT to write one for me...

Just asking, did you tell it what state you live in?

[Fallible]

What about this idea?
1. Have ChatGPT write up the legal document, then
2. You review it for obvious errors, then if it looks good,
3. You have an attorney review and edit it (without telling the lawyer that it was AI-generated to avoid prejudice).
4. You do a final review.

The best of both worlds, less chance for error, and you might save money....sort of like how in medicine the intern does the exam and it is then reviewed by the attending MD.

Interesting, but what prejudice would be avoided in No. 3? And if there might be prejudice, shouldn't that be a reason or possibly an obligation to let the attorney know?

[Tdubs]

This reminds me sharply of a time Way Back When, in the days of dedicated word processors. I worked in a research institution that had an "editorial services" department to help scientists, many of whom were not native English speakers, prepare their papers for publication. Typically they submitted typed material to the department. Operators would key it in manually.

Lanier (major word processor provider, pre-PC) wanted to sell them an OCR scanner to avoid the work of rekeying. The department head asked them "Can you guarantee me 100% accuracy?" They said, proudly, that it had 98% accuracy.

She said "Then it is no value to us. We spend most of our time on proofreading, not data entry. If it is only 98% accurate, then we still need to proofread it. The only way there is a savings is if it is so accurate that we can eliminate the proofreading step."

And so it is with legal documents. A legal document that is 98% correct has zero value, and you will need to pay a lawyer as much money to review it closely enough to spot the errors as you would to do it themself.

In a way, ChatGPT is worse because the output is so smooth and plausible that a reviewer becomes fatigued and numb and lets their guard down.

That OCR scanner sounds like a great idea. Someone ought to try it.

[RationalWalk]

...But I just got the idea to ask ChatGPT to write one for me...

Just asking, did you tell it what state you live in?

The first draft no. Then I asked it for another draft for the state. It gave me the same text but included the state name. Not optimal, obviously. But I've been working on this document for awhile with the attorney and thought that the ChatGPT version was a lot more succinct and worded better. But it was clear there needed to be some finishing work to do. As I said, if you could get an attorney to review your draft and include any missing necessaries the final product would be better (in my view) than the attorney's boilerplate. I've been working to customize the attorney's boilerplate to my needs for awhile and I would have preferred to have started with the AI boilerplate -- it was better.

[Fallible]

I've been fencing with an attorney for about a month now trying to get a simple, direct, and adequate Limited Real Estate POA written. Maybe it's just the attorney. But I just got the idea to ask ChatGPT to write one for me, and in about 2 seconds there was a nicely crafted document. In many ways, better than the one the attorney has come up with. Yes, this is a pretty simple legal document as things go, and yes, it needed some additional polishing but it was a pretty good effort for $0 and 2 seconds of my time. All I'd need now is to have an attorney put on the finishing touches, if I could get one to do that. I can certainly see how AI in the future is going to be able to handle stuff like this better, faster, and a whole lot cheaper. And be able to handle much more complex documents using an interrogation or interactive mode. Next I'm going to ask it to write me a Will and see what it does. If I were an attorney, I think I'd be figuring out how I'm going to live in the world of AI. Has anybody else had this experience recently?

But it isn't just 2 seconds of your time because you had to apply additional polishing and now plan to find an attorney willing to add finishing touches. And I assume you spent some time creating an effective prompt.

Generally speaking, the speed alone of these machines really do impress. It's just that at this point in their development, there's more to it than speed - like accuracy.

[Sandi_k]

Nolo.com has had boilerplates online for years. Including POA for finance, health care, and real estate.

[snackdog]

If you are having so much trouble with your attorney that you turned to ChatGPT *and* got a better result, this mostly tells me you need a new attorney. Almost any experienced attorney would be able to churn out a near-perfect POA on the first try after confirming all the details of your situation.

[rationalactor]

We shouldn’t kid ourselves. Generative AI will cut a swath of destruction through white collar knowledge work, leaving millions unemployed and redistributing value in once-in-a-century-or-two proportions. For this forum I think the key question is whether the identity of the winners and losers is so unclear that an indexing strategy remains supreme, or if there is a group of tech companies so well positioned to fuel and profit from this next Industrial Revolution that of course one should put a lot of money in them. Or perhaps it is a little of each…total stock market and Microsoft, Google, and Amazon, anyone?

[MrJones]

The fundamental problem is if you had the expertise to evaluate the ChatGPT work product, you could have written your own POA. And since you don't have that expertise, you still need a lawyer to evaluate and very likely fix the document. So ChatGPT created something that superficially looks good but didn't advance your situation any.

+1. ChatGPT is simply excellent for jobs like writing a small program or even creating a spreadsheet **where it is trivial to evaluate the output for correctness**.

Unfortunately in this case, there's no way for you to evaluate this doc, which would make it completely useless in my view.

Why not use software designed for this specially, if something like that exists? If not, find a reputable but responsive and reasonably priced attorney?

[RationalWalk]

Nolo.com has had boilerplates online for years. Including POA for finance, health care, and real estate.

I see that, but it costs $19.95 and isn't any better than the one I got for free from ChatGPT.

[RationalWalk]

I've been fencing with an attorney for about a month now trying to get a simple, direct, and adequate Limited Real Estate POA written. Maybe it's just the attorney. But I just got the idea to ask ChatGPT to write one for me, and in about 2 seconds there was a nicely crafted document. In many ways, better than the one the attorney has come up with. Yes, this is a pretty simple legal document as things go, and yes, it needed some additional polishing but it was a pretty good effort for $0 and 2 seconds of my time. All I'd need now is to have an attorney put on the finishing touches, if I could get one to do that. I can certainly see how AI in the future is going to be able to handle stuff like this better, faster, and a whole lot cheaper. And be able to handle much more complex documents using an interrogation or interactive mode. Next I'm going to ask it to write me a Will and see what it does. If I were an attorney, I think I'd be figuring out how I'm going to live in the world of AI. Has anybody else had this experience recently?

But it isn't just 2 seconds of your time because you had to apply additional polishing and now plan to find an attorney willing to add finishing touches. And I assume you spent some time creating an effective prompt.

Generally speaking, the speed alone of these machines really do impress. It's just that at this point in their development, there's more to it than speed - like accuracy.

My prompt was this: "Write a limited real estate POA". It took only the time needed to type that in. Believe me, I've spent way too much time dealing with the attorney to get something from her. If I could simply have taken the ChatGPT - generated document to her for final tuning and blessing that would have been the way to go.

[stan1]

find a reputable but responsive and reasonably priced attorney?

These are hard to come by. How does one find such a unicorn? Why would a reputable and responsive attorney work at anything other than a premium price? Yes there are a few market anomalies but most people will try to get paid commensurate with their skills and abilities.

[nisiprius]

...But I just got the idea to ask ChatGPT to write one for me...

Just asking, did you tell it what state you live in?

The first draft no. Then I asked it for another draft for the state. It gave me the same text but included the state name. Not optimal, obviously...

If the text didn't change when you gave it the state name, I think that's a red flag. A red flare. A three-stage whistling red skyrocket.

I thought that the ChatGPT version was a lot more succinct and worded better.

You are not the intended audience for a legal document.

Nolo.com has had boilerplates online for years. Including POA for finance, health care, and real estate.

I see that, but it costs $19.95 and isn't any better than the one I got for free from ChatGPT.

Yeah, they probably trained it on the Nolo book. But at least if you have a Nolo book, you know the source of the information. Letting ChatGPT do it is probably like having it done by a friend who once read the Nolo book.

[srt7]

The fundamental problem is if you had the expertise to evaluate the ChatGPT work product, you could have written your own POA. And since you don't have that expertise, you still need a lawyer to evaluate and very likely fix the document. So ChatGPT created something that superficially looks good but didn't advance your situation any.

Exactly! The fundamental problem here is that of validation. We are not to the point where we can trust ChatGPT / AI outputs.

[srt7]

The fundamental problem is if you had the expertise to evaluate the ChatGPT work product, you could have written your own POA. And since you don't have that expertise, you still need a lawyer to evaluate and very likely fix the document. So ChatGPT created something that superficially looks good but didn't advance your situation any.

Why should coders be the only ones to lose their jobs? I think AI will figure out these legal bumps in the road in good time.

https://www.newyorker.com/magazine/2023 ... -the-craft

I'm not sure I understand your comment. Who says coders are losing their jobs? Sure, my old Uncle Harold can ask ChatGPT to spit out the code for a Hello World program but software engineering is way way more than coding. Good ole' Uncle Harold would need to design first, then be able to ask the right questions, validate ChatGPT's output and then know where to plug in all these pieces (software deployment) for the desired outcome. If he manages to do all that then he's a software engineer himself.

[srt7]

I've been fencing with an attorney for about a month now trying to get a simple, direct, and adequate Limited Real Estate POA written. Maybe it's just the attorney. But I just got the idea to ask ChatGPT to write one for me, and in about 2 seconds there was a nicely crafted document. In many ways, better than the one the attorney has come up with. Yes, this is a pretty simple legal document as things go, and yes, it needed some additional polishing but it was a pretty good effort for $0 and 2 seconds of my time. All I'd need now is to have an attorney put on the finishing touches, if I could get one to do that. I can certainly see how AI in the future is going to be able to handle stuff like this better, faster, and a whole lot cheaper. And be able to handle much more complex documents using an interrogation or interactive mode. Next I'm going to ask it to write me a Will and see what it does. If I were an attorney, I think I'd be figuring out how I'm going to live in the world of AI. Has anybody else had this experience recently?

Your attorney is already using stock forms that get modified to your specific situation. So this basically is you providing the boilerplate to the attorney instead of the attorney using her boilerplate already.

This! Whether it's an accountant, a lawyer or a software engineer, all ChatGPT does today is create the boilerplate. Expertise is still required to ensure the output is correct and any modifications to suit the requirements are being met.

[srt7]

The fundamental problem is if you had the expertise to evaluate the ChatGPT work product, you could have written your own POA. And since you don't have that expertise, you still need a lawyer to evaluate and very likely fix the document. So ChatGPT created something that superficially looks good but didn't advance your situation any.

Why should coders be the only ones to lose their jobs? I think AI will figure out these legal bumps in the road in good time.

https://www.newyorker.com/magazine/2023 ... -the-craft

Contracts like this are far easier than code... Why do you think the lawyer orgs are going after that company that started using LLM's for people - they have to go after the LLM for "practicing" law without license etc. to protect their profession. It's for the most part easily repeatable.

I am no lawyer but c'mon ... "easily repeatable"? That's a bit of a hyperbole.

[RationalWalk]

I've been fencing with an attorney for about a month now trying to get a simple, direct, and adequate Limited Real Estate POA written. Maybe it's just the attorney. But I just got the idea to ask ChatGPT to write one for me, and in about 2 seconds there was a nicely crafted document. In many ways, better than the one the attorney has come up with. Yes, this is a pretty simple legal document as things go, and yes, it needed some additional polishing but it was a pretty good effort for $0 and 2 seconds of my time. All I'd need now is to have an attorney put on the finishing touches, if I could get one to do that. I can certainly see how AI in the future is going to be able to handle stuff like this better, faster, and a whole lot cheaper. And be able to handle much more complex documents using an interrogation or interactive mode. Next I'm going to ask it to write me a Will and see what it does. If I were an attorney, I think I'd be figuring out how I'm going to live in the world of AI. Has anybody else had this experience recently?

Your attorney is already using stock forms that get modified to your specific situation. So this basically is you providing the boilerplate to the attorney instead of the attorney using her boilerplate already.

This! Whether it's an accountant, a lawyer or a software engineer, all ChatGPT does today is create the boilerplate. Expertise is still required to ensure the output is correct and any modifications to suit the requirements are being met.

ChatGPT provided this friendly reminder:

Certainly, here's a template for a limited durable real estate power of attorney. Remember to consult with a legal professional to tailor it to your specific situation and ensure compliance with local laws:

[Fallible]

I've been fencing with an attorney for about a month now trying to get a simple, direct, and adequate Limited Real Estate POA written. Maybe it's just the attorney. But I just got the idea to ask ChatGPT to write one for me, and in about 2 seconds there was a nicely crafted document. In many ways, better than the one the attorney has come up with. Yes, this is a pretty simple legal document as things go, and yes, it needed some additional polishing but it was a pretty good effort for $0 and 2 seconds of my time. All I'd need now is to have an attorney put on the finishing touches, if I could get one to do that. I can certainly see how AI in the future is going to be able to handle stuff like this better, faster, and a whole lot cheaper. And be able to handle much more complex documents using an interrogation or interactive mode. Next I'm going to ask it to write me a Will and see what it does. If I were an attorney, I think I'd be figuring out how I'm going to live in the world of AI. Has anybody else had this experience recently?

But it isn't just 2 seconds of your time because you had to apply additional polishing and now plan to find an attorney willing to add finishing touches. And I assume you spent some time creating an effective prompt.

Generally speaking, the speed alone of these machines really do impress. It's just that at this point in their development, there's more to it than speed - like accuracy.

My prompt was this: "Write a limited real estate POA". It took only the time needed to type that in. Believe me, I've spent way too much time dealing with the attorney to get something from her. If I could simply have taken the ChatGPT - generated document to her for final tuning and blessing that would have been the way to go.

Why can't the ChatGPT do the final tuning and blessing? I even wonder why it didn't offer to do so.

Edit to update: OP, I see in your post above that it did offer a "friendly reminder," although I think of it less as friendly and more as a sort of legal obligation from its creators.

[JPH]

Artificial intelligence will never surpass human stupidity.

[quantAndHold]

Staples sells limited POA forms for $12, that are much more likely to to be legally valid than something ChatGPT produces. A friend needed to give me POA for his rental property for a few months. The form he got at staples worked just fine.

[gatorking]

Artificial intelligence will never surpass human stupidity.

Love it. I'm shamelessly stealing that.

[MrJones]

find a reputable but responsive and reasonably priced attorney?

These are hard to come by. How does one find such a unicorn? Why would a reputable and responsive attorney work at anything other than a premium price? Yes there are a few market anomalies but most people will try to get paid commensurate with their skills and abilities.

I agree these take a little bit of work to find, but I wouldn't call them unicorns either. And might be valuable for the long run. OP is probably paying average prices but getting a substandard service. It's not terribly hard to call around, get recommendations from friends, see online reviews, post here and such to get a handful of recommendations to follow up on.

Specifically, it sounds like if the OP could get a responsive attorney, they could quickly iterate and arrive at a solution.

As mentioned above, NOLO forms are time tested and even better a solution if available for OP's requirements.

[Cruise]

It's also amazingly biased - it will happily tell a joke about men, but if you ask it to tell a joke about women it will say it won't tell a joke about groups or whatever. If you then point out that it just did about men, it will acknowledge it, apologize, and say it will try to do better.

You are so right about the built in bias. I asked it to give me a list of positives about concealed carry and also firearms ownership, and all I could get from it were warnings about the bad things firearms do.

[renter]

I've been fencing with an attorney for about a month now trying to get "

This sounds very strange to me.

[regularguy455]

Large Language Models are a form of non-deterministic automation. Unlike robots, they’re able to follow an instruction effectively on unseen data. If you don’t think that’s a danger to human jobs, you’re fooling yourself.

I see several comments on LLMs ability to perform math and math-based reasoning. These are well know gaps. But the plug-ins that OpenAI has rolled out and the use of agents will more than likely fix these.

These models are probably analogous to Internet 1.0. Our world is build on the foundational concepts of Internet 1.0 but looks unrecognizable to the early days.

[LadyGeek]

Please stay on-topic, which are the personal finance aspects.

[RationalWalk]

Staples sells limited POA forms for $12, that are much more likely to to be legally valid than something ChatGPT produces. A friend needed to give me POA for his rental property for a few months. The form he got at staples worked just fine.

You won't know how "just fine" it is until you try to use it. Maybe you have?

[quantAndHold]

Staples sells limited POA forms for $12, that are much more likely to to be legally valid than something ChatGPT produces. A friend needed to give me POA for his rental property for a few months. The form he got at staples worked just fine.

You won't know how "just fine" it is until you try to use it. Maybe you have?

I did use it. It was fine.

[RationalWalk]

It's interesting. Regarding the DPOA for my mother, we had to provide that in several places but I don't recall any of them actually taking much time to review it carefully before acting on our authority. I think it could have read "Don't trust these people, and don't do anything they say. I'm being held hostage in a beer cellar." and it wouldn't have been noticed. Maybe it wasn't even necessary to pay an attorney to draw it up. ChatGPT might be my next lawyer.

[Fallible]

It's interesting. Regarding the DPOA for my mother, we had to provide that in several places but I don't recall any of them actually taking much time to review it carefully before acting on our authority. I think it could have read "Don't trust these people, and don't do anything they say. I'm being held hostage in a beer cellar." and it wouldn't have been noticed. Maybe it wasn't even necessary to pay an attorney to draw it up. ChatGPT might be my next lawyer.

How would ChatGPT respond if you asked it to be your attorney?

Whatever, you might want to read this written a few days ago in the "Washington Post," showing where the chatbot/legal issue stands currently:

Rather, chatbots like ChatGPT are designed to make conversation, having been trained on vast amounts of published text to compose plausible-sounding responses to just about any prompt. So when you ask ChatGPT for a legal brief, it knows that legal briefs include citations — but it hasn’t actually read the relevant case law, so it makes up names and dates that seem realistic.

Judges are struggling with how to deal with these errors. Some are banning the use of AI in their courtroom. Others are asking lawyers to sign pledges to disclose if they have used AI in their work. The Florida Bar association is weighing a proposal to require attorneys to have a client’s permission to use AI.

https://www.washingtonpost.com/technolo ... -fired-ai/

[AerialWombat]

I recently worked with a real estate attorney to craft additional language to add on to the standard commercial lease that his document software created.

I sat there and watched as he blatantly asked ChatGPT for the legal language, which he copy and pasted into the lease document. He did read it, and made tiny tweaks, but he did this four times.

I had no qualms with it. It saved a couple hundred dollars of billable time. It’s a tool, just like anything else. Use it as such.

[valleyrock]

It might be useful (financially) and interesting to compare a ChatGPT POA to a Bard POA.

Then invest in the company with the most accurate output, as evaluated by a group of crackerjack real estate attorneys.

Maybe this is the way of the future. No jobs displaced, just new jobs, in various professions, evaluating the accuracy/feasibility/utility of field-specific AI output.

[cheesepep]

I would have no issues with ChatGPT drafting a simple POA or downloading a free one off Nolo or other similar sites. Just do it and compare a few others that you find online and make your own. There is nothing special about a lawyer doing it for a simple POA. The lawyer will also just use a template specific for his law firm and that is basically it. He does nothing value added for a simple POA.

[LeftCoastIV]

I’d like to make a non-critical change to our wills, related to how a trust will handle distributions. The logic of the change is simple, but with legal documents there are the nuances of state law and any prior case law that may lead to the need for very intentional word choices.

I have held off because the lawyer’s hourly fee is very expensive.

Could ChatGPT do this? Maybe with a state-specific prompt, but then you are exposed to any change in legislation or case law since the LLM was last updated.

[Horologium]

Retired atty here. I could be wrong, but I don't think lawyers have too much to worry about - at least not yet.

Usually, the issues that come up with even legal matters, like with routine matters like a POA, are not what the client tells the atty, but what the client isn't telling the atty, or what the client doesn't even know. That's a key part of the value of an experienced atty. He or she will know what questions to ask to get necessary information that the client hasn't thought to provide, and they know whether a POA needs to be notarized, or witnessed, and how many witnesses (is one OK?, or do you need two?), and who can and can't be a witness, etc.

ChatGPT can generate a document, but whether it's valid or enforceable is something else.