It's a common misconception that AI needs to be able to perform essentially any task a human can do to induce the "collapse" of an industry. But really, all it needs to do is make each lawyer 2x more efficient (by automating the easy-to-automate tasks) and suddenly half of all lawyers are unemployed even though every lawyer employed today does some tasks that can't be automated. That's a collapse.

This assumes there's a fixed demand for legal help. You might as well envision that the efficiency gains cause a huge drop in legal costs which unlocks a huge latent demand. For instance software engineers are probably 10x more efficient now compared to 20 years ago, but this does not mean mass unemployment.

Software is the exception since its a new industry that is simultaneously expanding demand by orders of magnitude while automation dramatically increases the productivity of individual SEs.

There are plenty of people who were in industry and were forced out because they could not adapt to more efficient tooling to keep their jobs. Those that learned on ASM and Fortran may not be adapt for a world of prebaked containers and building everything out of node plugins that do the work. Even less are they suited to training heuristics for genetic algorithms that implement some of our most sophisticated software projects like speech recognition or vision.

By comparison at least those with money have saturated their demand for lawyers. Law is a particularly targetable field since its often considered overhead or expense.

"Software is the exception." If I weren't reading this on HN...

Your only argument is stating that the demand for lawyers is saturated. How is that different from saying the supply curve and the demand curve cross? This is about a shift in the supply curve. The whole point of the previous comment is that AI-assisted legal counsel will be much much cheaper than it currently is. That means people will quite likely want more of it.

Software is the exception as-in it is the tool that is driving the elimination process itself. And at the same time it does lead to the same effects inside the software industry. These two things seem to be opposites but they aren't. They are different facets.

Akin to something that happened with accounting when computerised spreadsheets came about. Instead of killing the field it just expanded it as more people wanted advice and what-if scenarios explored.

And it's ludicrously obvious how much extra potential work there is for legal contractors. Right now there are companies turning down revenue-generating opportunities because the time/cost of scrutinising proposed contract changes isn't worth it. Not to mention a huge number of agreements done on a "standard contract, no exceptions" basis for the same reasons. If automation allows these to be turned around in hours, for a fraction of the price, the variation and complexity in standard contracts is going to multiply exponentially.

Even if the process of checking the contracts is fully automated, there's still a role for an associate/paralegal level lawyer (overseeing many more contracts than their predecessors) to answer followup questions, because their clients aren't going to navigate through case law summaries to get a more thorough explanation of non-standard term 3.1b's applicability to their specific situation if they can still phone somebody with a law degree.

Standard contracts exist for a variety of reasons, time/cost being one of them. Another one I'd like to highlight is the principal-agent problem[0]. If, due to automation, contracts went up by orders of magnitude in complexity then clients would be exposed to a new set of risks with a legal basis that they do not understand. The question then is: are these new, inscrutable risks acceptable to a client? Perhaps not, unless they are already a very large business.

[0]. https://en.wikipedia.org/wiki/Principal%E2%80%93agent_proble...

Do you understand that accountants are one of the first to go with the advent of AI? It's a foregone conclusion that accountants will not be around in 2030, at least according to Bill Gates

Judging by the hours my patent attorney friend works, automating 50% of what he does would bring him down to a 60 hour work week.

Well breadth of legal practice areas are what machines are good at, a machine that can analyze one law can analyze all of them and with constant quality. To give a concrete example, if some rather arcane part of insurance law may be relevant for a contract, a machine can then analyze the insurance law just as it did with contract law. By contrast a human lawyer has to ask for advice from a more specialized lawyer.

The human element is a quite good argument, at least without a major shift in perception of lawyers. However, I think that automation will realistically change what lawyers do, as well as the overall market for lawyers. One 'easy' area of law would be scanning employment contracts for standard clauses and you get some information at each paragraph what it means. Third page: first paragraph gets a green background and a note on the side says standard language, three more days off per year than standard. Second paragraph gets a yellow background and a note that says may get you into trouble if you use a company notebook for recreational programming. Third paragraph gets a red background and a text "Highly non standard, check with a lawyer." In this example, a lawyer still gets involved, he just does not have to check for the routine stuff and if history is any indication, this is what will happen. The routine stuff gets automated and the interesting cases are referred to the few remaining humans.

You're right that analyzing overlapping areas of law might also lend itself to an AI's strength. I guess I meant that applying those overlapping concepts to achieve a certain real world result seems to require input from someone with real world experience, and an empathetic side to fully understand the client's goals.

15 years is a crazy long time. Technological / scientific progress is exponential - we will witness much, much more change in the next 15 than we have in the last 15. And, that's a daunting thought, since 2000 was such a different world than it is now.

The only forces keeping law in its relative stasis are its attachments to slow-moving government processes, its 'guild-like culture' and the resulting protectionism. But, once capital moves faster in that direction (and it will) those systematic factors will slide apart quite easily.

It's good to be confident, but having participated in and thus seen the wave of progress in AI over the last 20 years, I'm pretty scared for my own career. It is definitively an exponential process and one that probably rivals Moore's law in effectiveness.

Most technology hasn't improved very much in 15 years. Technology developed in the 1980's and 1990's has found many new applications, because of the Internet, but fundamental systems haven't improved much. In the professions, the increases in efficiency from 1995 to 2000 were more significant than from 2000 to 2015. Even Moore's law has slowed dramatically in the last 15 years.

And the government and guild protections are fictions. I work at a small boutique that would love to be able to take bigger cases away from big white shoe firms by leveraging technology. We're not going to leave money on the table right now because of some abstract loyalty to the "guild." If the technology existed, we would use it. So would everyone else. The legal industry is extremely competitive. There are hundreds of large business law firms in the country. Way too many to keep up some artificial convention not go adopt new technology that works.

I don't think the government and guild protections are fictions. This article seems to be relevant:

http://techcrunch.com/2014/03/21/lawyer-disrupt-thyself/

I don't think they're fictions insofar as lawyers (and laws) command a unique place in our society that other industry inputs don't. Insofar as lawyers have the ability and the incentive to prevent sweeping changes to their industry, the guild mentality seems very real to me.

I take your point though: smaller firms are very willing to make technology changes to find themselves new competitive advantages. Here's hoping smaller firms like yours (and mine) keep nipping away at those monstrous firms =D. FWIW, I have not had positive experiences when dealing with them...

The article is about non-lawyer ownership of law firms. Loosening the rules on conflicts and non-lawyer ownership of law firms would enable what happened in the accounting industry: consolidation of the industry into a handful of mega-firms. I don't see how that would be conducive to innovation. Exactly the opposite is true. A legal "Big Four" would be far more able to resist technological changes than the hundreds of firms competing with each other today.

I'm trying to think of a concrete example of protectionism keeping out new technology, and frankly I'm at a loss. I don't think it squares with the economics. Clients these days come with a budget in mind. If a law firm can do the work with say half as many billable hours by using technology, they can bid to do the work for a fixed fee 25% below the market rate, and then pocket the difference. It'd be a no brainer, if that technology existed. Instead, more often you see firms agreeing to a budget cap and just eating hours billed over the cap.

No. Think back 15 years. Almost everything in technology was the same as today. Yes memory density is higher now, cpu power efficiency is better and display technology is somewhat improved but basically we're using the same technology now as then.

In my lifetime I'd need to go back to before 1974 or so for things to be really different. That's before microprocessors became commonly used. Before pong. Not before Unix though. 40 years. Still not that different from today.

Technically not much changed but the way we work and live is totally different because of the advances in tech. In 2000 internet was not everywhere; I was on holiday in LA in 2004 and the internet quality in hotels that advertised with internet was miserable and mostly unusable. Mobile devices sucked and we were still in the AI winter. Search engines in 2000 did deliver total drivel for programming related questions (and probably others).

Now my subscription allows me to use my 4g in most of the EU and the US, it works very well, wifi is everywhere so basically you do not have to remember any facts. Google AI recognises pictures and auto makes panoramas and videos from pictures. When I search anything programming the first hit is the answer. Children have smartphones and tablets and use them all the time. So do my parents and my 80+ old neighbours. There are millions of coders online instead of the tiny amount in 2000; starting a software services company in 2000 was a breeze (I did) to grow and get fortune 1000 clients as there simply was very little and most was crap; try that now. Startups are everywhere and a huge % of students see that as an option now. We can 3d print implants. Did you check games lately? I can go on and on.

Although tech only got faster and bigger the world is completely different. And that is because those 'marginal' enhancements in tech and science.

I'm thinking this must be an age thing. 15 years ago isn't a significant portion of my lifetime, so perhaps I see the progress in that time period as not being tremendous. Perhaps you're younger. I've been active in this business/space/industry since around 1975 and for me not a whole lot has changed. Sure today you can put a thing with the power of a VAX on your wrist, but we expected that to happen back in 1980. I moved into the house I live in now 15 years ago and immediately set out to build myself an internet connection. I used 802.11b products that are not too different to what we use today. The one thing that has changed in the past 15 years that I see clearly is there is much more data available on the Internet, but that's just a simple adoption curve effect. Back in 1992 we had the Microsoft Developer Network on CD with a search application that ran on Windows. It was basically the same thing as Stack Overflow, except of course much slower to update and featuring contributions from vastly fewer people. But essentially the same thing. I even remember having discussions in 1992 with colleagues about how the MSDN would surely evolve to be Internet-based rather than delivered by mail on a CD.

Maybe it depends where you are? East coast U.S. is entirely different culturally, behaviorally, commercially than it was 15 years ago. And, all that change has been enabled and propelled by a recursive process of capital and technology. Technology was barely a concept in most people's minds in 1999. Now, it is the prevailing culture.

I can't think of an industry that hasn't been touched since then, and most have been utterly transformed.

15 years ago I didn't think I'd see self-driving cars in my lifetime, and there didn't appear to be any prospect for computer vision to get anywhere near human levels of performance.

And these thing are still true.

We have self driving cars on the streets, and vision systems out performing humans at classifying objects.

No we don't. Where can I buy a self-driving car? The vision systems are performing on a statistical basis that is admittedly impressive, but they are not "outperforming humans at vision" in any way that I understand.

I totally agree but I think your defense is even too generous to the other side. For me, it is a little laughably naive, the lack of recognition of how complex "lawyering" actually is. Fancy NLP does not even get within several orders of magnitude of that.

Lawyers are meant to achieve positive results for clients even in litigious circumstances. For instance, in an auto accident case, a lawyer must carefully consider how best to present medical evidence to get the best possible offer in a settlement negotiation.

Lawyers must consider creative solutions to legal problems that are not as others suggested "just finding precedent snippets."

How do I even explain my complex case to the AI lawyer, assuming it is not some kind AGI Overlord? It's sort of stupid, I'm sorry to say, to me to see it suggested that lawyers can be replaced by LegalZoom+NLP. Maybe these people have never needed a lawyer for anything other than what they consider to be perfunctory paper pushing or they think we should not have an adversarial legal system.

If I have a serious legal problem I will take a serious firm of experienced lawyers over an NLP system. I guess if your outcome doesn't matter to you, you can pick what you like. It reminds me of the saying about the lawyer defending himself having a fool for a client.

> I don't think our society is ready for automated justice at this point. My clients certainly aren't.

They will if the service costs $10/mo instead of the $100s or $1000s per /hour/ you are charging...

I guess I should have explained more what I meant by "justice." In your quoted language, I intended it to mean justice as handed down by a person in a black robe.

Clients are already lapping up automated legal services (wills, contracts, etc.). But I don't think they're willing to accept a judgment from an opaque black box. Indeed, the cost to receive a judgment does tangentially involve some automatable processes, and those costs will come down as our computers get smarter.

To extrapolate a bit: as long as humans are determining legal outcomes, there will be human advocates. There will be human judges for as long as we have a Constitution.

They certainly won't if the cost of losing the case is sufficiently high.

Maybe the fact that I am not from somewhere where litigation pays off gives me a different perspective. Here suing is just not worth it usually as the judge will not demand significant payout. So when you win you will not get much anyway. Most of those cases can be replaced by computers today (judge and lawyers) and as far as I know those are most cases that are done here. So like trivial bankrupty cases, most small claims stuff. I had a few of those and I can and did just use Google to write my defense which works fine. No lawyers needed there. In the US the stakes are higher I guess so maybe I just cannot comment on that. And then there is criminal cases which might be probably more complex. Although AI might augment the lawyer to generate possible defenses and give the lawyer more creativity?

Interesting! What kind of legal processes do you already automate using software in your day to day life as a lawyer?

Our firm automates document assembly software for wills, trusts, purchase agreements, residential leases, and court forms.

We do some litigation, but in our small town it typically doesn't require extensive electronic discovery or document analysis that this post talks about.

As a side note, I'm a Freelance Web Developer. Below are links to my website and upwork profile. Let me know if you need to develop a web-based app or a mobile app around any legal concept and I'll gladly help!

https://www.prahladyeri.com/

https://www.upwork.com/freelancers/~017d369bd108f07773