4 ms·

A.I. Is Coming for Lawyers, Again

14 points by fortran77 4y ago

__lbracket__ 4y ago
>Law is seen as the lucrative profession perhaps most at risk from the recent advances in A.I. because lawyers are essentially word merchants. Many professions can be reduced to "$x Merchants". This reduction amounts to saying that the said profession does not require creativity/understanding/synthesis. I'd say a good lawyer is much more than a word merchant. He/she is a synthesizer of existing body of knowledge to create novel arguments. Loosely speaking, this type of activity seems NP hard. Now ChatGPT++ may be coming for those layer of most professions (programming and law included) that can be reduced to "lookup". However if you believe (as I do) that there are layers within most professions that are _more_ than lookup, and if those deeper layers solve some kind of an NP hard problem at their center (novel argument/theorem/algo/architecture...) then a lookup tool like ChatGPT would necessarily find it hard to address them. I'm obviously using things like "NP Hard" rather loosely and also assuming that AI, at least to the extent it uses training data, is no more than a lookup tool.
- hammyhavoc 4y ago
  As with the "code is law" crap with "smart contracts" in crypto, I can see the same problems emerging in courts. This happened, and I find it problematic: https://www.vice.com/en/article/k7bdmv/judge-used-chatgpt-to-make-court-decision https://www.vice.com/en/article/k7bdmv/judge-used-chatgpt-to... Human beings should be the ultimate arbiters of right and wrong because justice is not a concrete concept, it is very nuanced.
- NoboruWataya 4y ago
  I agree with this. A lot of law firms are themselves investing heavily in AI precisely to rationalise this "lookup" layer and thus increase the profitability of their core work. One challenge this throws up is that, currently, those employed in the lookup layer are usually junior lawyers who are gaining experience and confidence before they can go on to perform the more valuable work of drafting complex contracts, constructing novel arguments etc. If law firms automate away the lookup layer, they will actually need to re-think their approach to how they utilise junior lawyers. They'll still need to employ them, as they are the valuable lawyers of tomorrow, but unless they can re-deploy them to other useful tasks during that phase of their careers, the upside to automation is limited.
  - __lbracket__ 4y ago
    You're right. I didn't think of the lookup layer as a bootcamp for moving into the creative/synthesis layer. And "lookup as a learning experience" applies to most professions. We can even say that this is probably "apprenticeship" in common terms.
rayiner 4y ago
I’m a legal tech skeptic, but I’ve been very impressed by Chat GPT’s ability to answer simple legal questions. As long as bags of meat remain the ultimate decision makers (judges and juries) you’re going to need some lawyers to interface with them. But it’s imaginable that you could reduce the number needed. The trial lawyer delivering the closing argument wouldn’t necessarily need a dozen other lawyers preparing the case for her. GPT is the first technology I’ve seen that seems to “understand” written work well enough to deliver meaningful answers to case-specific questions.
- vba616 4y ago
  I don't actually know what most of the work in today's legal industry comprises, (by what metric?) but in a discussion like this, it seems like people are thinking of Perry Mason or LA Law, whereas I'm thinking of document review, bates stamps, software like Relativity or proprietary document management systems. Deduplication, and search queries. If the best search technology dominated, why wouldn't Google have already taken over years ago? I know that's what I used to wonder. At least a few years ago, there seemed to be a lot of talk of disgruntled lawyers who turned to document review for low wages and miserable working conditions because they couldn't find "real" work. Can AI do that - for legal reasons, which after all, were why those lawyers got hired in the first place? Weren't the US based teams already replaced by offshore teams? How much can be gained now?
- gamblor956 4y ago
  I tried to get ChatGPT to answer some very simple legal questions and it was the most incompetent legal advisor I've ever seen. Sure, on the surface it seems like it's correct, but the content is all wrong. And content is what matters.
  - JamesBarney 4y ago
    Were you using 3.5 or 4? I've used 4 and been very impressed with it. For business it's been great, because the language is usually clear and unambiguous. For employment contracts it's been not nearly as helpful because the non-compete language is meant to be vague and ambiguous, and relies on a lot of case law around what is and isn't enforceable.

14 points by fortran77 4y ago

>Law is seen as the lucrative profession perhaps most at risk from the recent advances in A.I. because lawyers are essentially word merchants. Many professions can be reduced to "$x Merchants". This reduction amounts to saying that the said profession does not require creativity/understanding/synthesis. I'd say a good lawyer is much more than a word merchant. He/she is a synthesizer of existing body of knowledge to create novel arguments. Loosely speaking, this type of activity seems NP hard. Now ChatGPT++ may be coming for those layer of most professions (programming and law included) that can be reduced to "lookup". However if you believe (as I do) that there are layers within most professions that are _more_ than lookup, and if those deeper layers solve some kind of an NP hard problem at their center (novel argument/theorem/algo/architecture...) then a lookup tool like ChatGPT would necessarily find it hard to address them. I'm obviously using things like "NP Hard" rather loosely and also assuming that AI, at least to the extent it uses training data, is no more than a lookup tool.

hammyhavoc 4y ago
As with the "code is law" crap with "smart contracts" in crypto, I can see the same problems emerging in courts. This happened, and I find it problematic: https://www.vice.com/en/article/k7bdmv/judge-used-chatgpt-to-make-court-decision https://www.vice.com/en/article/k7bdmv/judge-used-chatgpt-to... Human beings should be the ultimate arbiters of right and wrong because justice is not a concrete concept, it is very nuanced.
NoboruWataya 4y ago
I agree with this. A lot of law firms are themselves investing heavily in AI precisely to rationalise this "lookup" layer and thus increase the profitability of their core work. One challenge this throws up is that, currently, those employed in the lookup layer are usually junior lawyers who are gaining experience and confidence before they can go on to perform the more valuable work of drafting complex contracts, constructing novel arguments etc. If law firms automate away the lookup layer, they will actually need to re-think their approach to how they utilise junior lawyers. They'll still need to employ them, as they are the valuable lawyers of tomorrow, but unless they can re-deploy them to other useful tasks during that phase of their careers, the upside to automation is limited.
- __lbracket__ 4y ago
  You're right. I didn't think of the lookup layer as a bootcamp for moving into the creative/synthesis layer. And "lookup as a learning experience" applies to most professions. We can even say that this is probably "apprenticeship" in common terms.

rayiner 4y ago

I’m a legal tech skeptic, but I’ve been very impressed by Chat GPT’s ability to answer simple legal questions. As long as bags of meat remain the ultimate decision makers (judges and juries) you’re going to need some lawyers to interface with them. But it’s imaginable that you could reduce the number needed. The trial lawyer delivering the closing argument wouldn’t necessarily need a dozen other lawyers preparing the case for her. GPT is the first technology I’ve seen that seems to “understand” written work well enough to deliver meaningful answers to case-specific questions.

vba616 4y ago
I don't actually know what most of the work in today's legal industry comprises, (by what metric?) but in a discussion like this, it seems like people are thinking of Perry Mason or LA Law, whereas I'm thinking of document review, bates stamps, software like Relativity or proprietary document management systems. Deduplication, and search queries. If the best search technology dominated, why wouldn't Google have already taken over years ago? I know that's what I used to wonder. At least a few years ago, there seemed to be a lot of talk of disgruntled lawyers who turned to document review for low wages and miserable working conditions because they couldn't find "real" work. Can AI do that - for legal reasons, which after all, were why those lawyers got hired in the first place? Weren't the US based teams already replaced by offshore teams? How much can be gained now?
gamblor956 4y ago
I tried to get ChatGPT to answer some very simple legal questions and it was the most incompetent legal advisor I've ever seen. Sure, on the surface it seems like it's correct, but the content is all wrong. And content is what matters.
- JamesBarney 4y ago
  Were you using 3.5 or 4? I've used 4 and been very impressed with it. For business it's been great, because the language is usually clear and unambiguous. For employment contracts it's been not nearly as helpful because the non-compete language is meant to be vague and ambiguous, and relies on a lot of case law around what is and isn't enforceable.