How disruptive will AI be for the legal profession

[cerberus]

A lot of industries are going to be disrupted by AI over the next decade and beyond. Legal services is often touted as one of those areas set to be heavily disrupted by AI agent lawyers.

However, we are a regulated profession. You need to have a license to practice law. Law Societies know that by granting licenses to AI lawyers they will essentially be opening the door to human obsolescence. They would be signing the death warrant for their careers if they did so. We are protected by the institutionalized gatekeeping that is a self-regulated profession. AI will be limited to just another tool like quicklaw.

In this manner, there may still be fewer legal jobs because AI will help 1 human lawyer be as productive as 2 or more human lawyers. However, humans will not be competing directly with AI agent lawyers for jobs.

Thoughts?

[Diplock]

Oh look. It's this post again.

[SmallBart]

It's already been very disruptive to the professional statuses of several specific lawyers.

[LMP]

I think it's pretty silly to say that regulation by the law society is the only thing preventing AI from wiping out the profession. There have been low cost and do it yourself options for legal services before and I don't see much of a difference with AI. 

Frankly I don't even see a significant application, as of now, as a useful tool. As an example, it isn't like the problem with drafting is typing everything out, we have precedents, the issue is altering information and confirming it's correct. 

And as for reaserch, well, we've seen what happened when lawyers have relied on AI to find cases for them. 

As Diplock said, we've had this discussion and I don't think there's been any developments that have increased my confidence in the technology.

[CleanHands]

If you believe that ChatGPT currently appears anywhere close to threatening your job in the near future, you are a shit lawyer doing braindead boilerplate transactional work.

[canuckfanatic]

My company recently got an enterprise license for Microsoft's "CoPilot".

My department played a quiz game in which the questions were generated by CoPilot, and we had to find the answers using CoPilot.

Lo and behold, not only were many of the "correct" answers blatantly false, different users received different answers despite asking CoPilot the same questions.

One of the questions that CoPilot came up with for the game was something like "what branch of mathematics is named after the shape of a compressed spring?" The answer provided by CoPilot was "Hooke's Law", which is very much not a branch of mathematics named after the shape of a compressed spring.

This whole game was intended to show the team how to use CoPilot, but turned into a PSA that nobody should rely on it for researching anything.

The only time I've found it useful has been when I'm experiencing writer's block for an email/letter.

[BHC1]

1 hour ago, LMP said:

I think it's pretty silly to say that regulation by the law society is the only thing preventing AI from wiping out the profession. There have been low cost and do it yourself options for legal services before and I don't see much of a difference with AI. 

Frankly I don't even see a significant application, as of now, as a useful tool. As an example, it isn't like the problem with drafting is typing everything out, we have precedents, the issue is altering information and confirming it's correct. 

And as for reaserch, well, we've seen what happened when lawyers have relied on AI to find cases for them. 

As Diplock said, we've had this discussion and I don't think there's been any developments that have increased my confidence in the technology.

Some of the more useful applications of this technology that I’ve seen have been in the legal adjacent context where there isn’t really any legal thinking required - instant AI generated transcripts of witness audio testimony for example. 

I have not tried out LexisNexis’ new AI tool. Is it useful? Maybe this has some application there.

[BondGuy]

10 hours ago, CleanHands said:

 shit lawyer doing braindead boilerplate transactional work.

So any transactional lawyer? Speaking as a transactional lawyer. Loves me a good precedent. 

My attitude is: AI will not replace lawyers. But lawyers that use AI will replace those that don't.  It is a tool like anything else, and a nascent one in the first generation(s). It will get better - so might as well start adapting now.   

At this point anything that comes out of an AI application should be assumed to be wrong and then revised/fact checked, but it helps get things started. 

[Diplock]

To give a more substantial answer, I'll say that I've been hearing about how AI would replace lawyer work ever since I was an applicant to law school. Hasn't happened in a substantial way yet.

Does technology replace some forms of skilled work as it evolves? Fuck yes. You know what replaced the most skilled legal work probably in a single pass ever in human history? The wide availability of mechanical printing. Before then, it was a huge and laborious effort to reproduce every copied document by hand. You need three copies of a contract? Some clerk had to write three identical copies - and they'd better be identical. Imagine how many jobs were eliminated when that became unnecessary?

So all that said, do I think that some disruption will filter in to the profession over time? Sure. All technology is disruptive. But the idea that we're on the front line of jobs that are going to be disrupted by AI and emerging technology...that's just ridiculous, to me. Some of what we do is mechanical, yes. Like copying documents by hand. But the great majority of what we do is creative. That's exactly why lawyers are paid as well as we are. It's a conceit of the unsophisticated public that lawyers are really just glorified search monkeys - as if anyone with a book who knew where to find the right law wouldn't need a lawyer at all. But that's flatly untrue. The creative aspects of law are far, far more important than the mechanical aspects of it. And could that be replaced one day by sufficiently advanced AI? Maybe. But by the time that happens AI will have replaced architecture and engineering and logistics and all kinds of other knowledge-based professions. And by then we might be dealing with SkyNet anyway.

Snide as the answer from @CleanHands may have been, there's a valid point there. But it doesn't invalidate the response from @BondGuy either. I'm sure quite a lot of transactional work is highly creative. If there is any legal professional out there who both understands what AI can and cannot do, and is actually threatened by it, that's a professional whose work is the modern equivalent of copying documents by hand. If there's any transactional work that's actually like that, maybe it's on the chopping block in the next decade or so. But I'll be long retired before our future AI overlords come for my job - I'm damn sure of that.

Edited March 25 by Diplock

[Jaggers]

I had a chat with a partner at one of the big firms about their use of CoPilot. He said they're obviously not allowed to use it for any legal or research work, but it is very good at doing transcripts and summaries of transcripts/documents, including extracting action items from a transcript of a discussion. I suspect AI will quickly take over some of the note-taking and summarizing functions that firms currently assign associates to do.

It can probably also do a really good job trawling through a precedent bank and finding the specific clauses you're looking to insert into your agreement, for example.

[CB2021]

16 hours ago, LMP said:

And as for reaserch, well, we've seen what happened when lawyers have relied on AI to find cases for them. 

if by AI, you mean generative AI and specifically large langue models (LLMs), I think people have this misconception that you can use them, on their own, for research. I think that's what happened with the lawyers that included fake cases in their submissions. But these models are not knowledge models, you cannot just use them like Google or a knowledge database because they cannot retrieve any of the information they have seen from training on command. In other words, they are essentially just extremely sophisticated auto-complete tools that can predict and generate stuff (text or other forms of data) based on your input prompts, sometimes information that is completely made up. This is not to say that, with appropriate prompting, you cannot generate accurate information; it's just that you should not treat the models as if they can replace Google or some database like Westlaw/LexisNexis.

16 hours ago, LMP said:

Frankly I don't even see a significant application, as of now, as a useful tool. As an example, it isn't like the problem with drafting is typing everything out, we have precedents, the issue is altering information and confirming it's correct. 

Keep in mind that most of the stuff we have seen right now are general purpose models, also known as foundation models, which can be finetuned and tailored to specific purposes once they are sophisticated enough. Given how much time and data that go into training these general purpose models and how they can continue to be improved by feedback and new data, I really don't see how a model that is trained on legal data and finetuned for specific legal tasks are not going to be able to do "altering information and confirming it's correct." And I think such models will materialize fairly soon, provided that the people who have the resources to invest will commit the capital to build models that are useful to lawyers. 

16 hours ago, CleanHands said:

If you believe that ChatGPT currently appears anywhere close to threatening your job in the near future, you are a shit lawyer doing braindead boilerplate transactional work.

lol I don't think OP was thinking specifically about ChatGPT, and if you think ChatGPT, even with GPT-4, is representative of what AI in general can or will be able to do, you are likely mistaken. 

[CB2021]

6 hours ago, BondGuy said:

My attitude is: AI will not replace lawyers. But lawyers that use AI will replace those that don't.  It is a tool like anything else, and a nascent one in the first generation(s). It will get better - so might as well start adapting now.   

Barring situations where restrictions are inappropriate imposed on the AI models, the tech will only get better from here. Def should start adapting now.

[CB2021]

15 hours ago, canuckfanatic said:

My company recently got an enterprise license for Microsoft's "CoPilot".

My department played a quiz game in which the questions were generated by CoPilot, and we had to find the answers using CoPilot.

Lo and behold, not only were many of the "correct" answers blatantly false, different users received different answers despite asking CoPilot the same questions.

One of the questions that CoPilot came up with for the game was something like "what branch of mathematics is named after the shape of a compressed spring?" The answer provided by CoPilot was "Hooke's Law", which is very much not a branch of mathematics named after the shape of a compressed spring.

This whole game was intended to show the team how to use CoPilot, but turned into a PSA that nobody should rely on it for researching anything.

The only time I've found it useful has been when I'm experiencing writer's block for an email/letter.

It will probs take some training beyond a quiz game to learn how to effectively prompt it to generate accurate information. 

[beachhouse]

I'm in the camp that believes AI will create enormous disruptions, but not wholesale displacement. AI is anticipated to create enormous disruptions to areas vastly more complex and high-stakes than the legal field including national security and intelligence activities, where the need for precision and sophistication is far greater. 

Using commercially available tools, will not likely revolutionize law. ChatGPT, Bard and so-on does a poor job generally, because of the hallucination effect, but it does an excellent job of understanding syntax and semantic content. 

Using industry and domain-specific AI tools, many which are still being developed, will almost certainly disrupt the legal profession. An example is retrieval augmented generation, which can be applied to specific knowledge sources or databases, and is adept at dealing with hallucination. It will certainly alter the way standard corporate law operates, including compliance functions, regulatory law, and e-discovery. Legal teams cannot match the ability to collate vast, complex data and parse and analyze it into relevant insights in the same way a tailored AI tool can.

With that said, you will almost always need a human-in-the-loop or on-the-loop, and a demand will be created for people capable of understanding both the legal side and the technological side of the equation. For national security agencies and militaries, every complex technical system developed has reduced manpower needs generally but increased the need for trained specialists capable of managing complex systems. The situation is no different for the legal industry, there will be less need for general manpower and more need for competent specialists. 

[CleanHands]

AI is a tool. It will enable us to do our work more efficiently. There are more legal problems out there than there are lawyers and time to provide ideal levels of representation. It will likely end up beneficial from an access to justice perspective, when used properly.

[JohnnyCochrane68]

There are so many disputes that settle or never get into a court because of the costs of investigation/litigation. Like how frauds above a certain level basically go unpunished.  The generator models promise to make a lot of this "legible".

[Case]

On 3/25/2024 at 9:32 AM, BondGuy said:

My attitude is: AI will not replace lawyers. But lawyers that use AI will replace those that don't. 

Seems like a wise statement. 

As a smaller solo practitioner, without cutting edge tech or much money to invest, I have been wondering how to possibly incorporate AI in my practice. 

My friend in the states uses it in some way to summarize phone conversations and client appointments.