March 06, 2011
Even Legal Mental Work Getting Automated

Even lawyers are getting automated out of jobs. Back in 1978 legal discovery costs could run into the millions for large numbers of workers sifting thru documents.

When five television studios became entangled in a Justice Department antitrust lawsuit against CBS, the cost was immense. As part of the obscure task of “discovery” — providing documents relevant to a lawsuit — the studios examined six million documents at a cost of more than $2.2 million, much of it to pay for a platoon of lawyers and paralegals who worked for months at high hourly rates.

The world has radically changed due to advances in computer hardware and software. Computers now replace lots of legal brain power.

But that was in 1978. Now, thanks to advances in artificial intelligence, “e-discovery” software can analyze documents in a fraction of the time for a fraction of the cost. In January, for example, Blackstone Discovery of Palo Alto, Calif., helped analyze 1.5 million documents for less than $100,000.

Was the $100k the total cost? It is not clear. But an inflation calculator shows $2.2 million in 1978 is $7.4 million in 2011.

Imagine a robot judge accepting filings from robot prosecutors and robot defense attorneys. The debates in many cases would proceed at speeds too fast for humans to follow. Still, a slower legal process could still happen due to computational costs for preparing arguments and analyzing opposing arguments. For example, imagine that the defense wants to prove their client was somewhere else at the time of a crime. The defense could file for a delay in order to have more time to run algorithms to filter thru video camera feeds looking for indications some camera caught the defendant somewhere else.

The defense could also try to come up with legal theories that are too computationally expensive to disprove. The prosecution could then complain that the defense has unfairly asserted a position based on an assertion that can't be disproved because any algorithm capable of generating a disproof would be of the complexity class NP-complete (at least unprovable today and possibly always unprovable).

Computational complexity will become a subject of legal rulings. Should reasonable doubt be allowed to rest on (seemingly improbable) interpretations of events that are computationally impossible to prove or disprove? Will defense teams (human or otherwise) be allowed to search for theories of events that can't be proved or disproved due to the enormous computational complexity of algorithms needed to test their theories?

Legal automation will proceed apace regardless of how these questions are resolved. Let us take a look at the big picture: how far will computer automation of human jobs go? Some people think that regardless of cognitive difficulty repetitive jobs will get automated but non-repetitive jobs won't. But as Paul Krugman notes at that link, even medical diagnosis stands a good chance of getting automated. So is it repetitive? That page also shows truck driving as a non-repetitive job. But wait a second. Google has automated guidance systems racking up many tens of thousands of miles driving cars on California highways. So how can truck driving be safe from automation? If truck driving isn't safe from automation then taxi driving isn't either.

Delivery truck driving might be safe for humans for a while longer just because humans have to hop out and deliver packages to the front door of a house or business. But suppose robotic delivery devices could deposit items into special delivery boxes out on streets? That'd allow automated delivery from not only online stores but also grocery stores.

What about the grocery stores? Well, no need for human check-out if robots get the food off the shelves. Kiva warehouse robots cut out human labor. These robots are going to enable automated local warehouses. So imagine grocery stories replaced with automated warehouses loading automated delivery vehicles to deliver groceries to houses. Deliveries could be scheduled to happen when you are at home so that getting perishables into the fridge in a timely manner won't be a problem.

New York Times writer Ron Lieber argues that online ordering from Amazon does not have to become as cheap as in-person shopping at Costco for Amazon to become preferred due to the time savings of online shopping. This has implications for automation. If you order stuff rather than buy it in person then its shipment to you is much more amenable to automation. As robotic delivery trucks hit the road and warehouse robot costs fall and they become more powerful and easier to manage the distance between home and warehouse will shrink. So the timeliness advantage of the local store will decline. Add in robotic delivery vehicles and why spend a couple of hours on a trip to a big box store?

Then there's work supervision. Supposedly non-repetitive and by humans only. Why? The theory is that humans need human supervision. But if bottom level tasks get automated then human supervisors aren't needed for the robots. And for some kinds of human tasks computers will eventually do much more monitoring and directing. This is already happening. For example, humans that package up orders (whether in restaurant kitchens or warehouses) are in many cases just reading order lists off a computer screen. In many of those cases no human supervisor chose which lists of foods or ordered goods to put on each computer screen. Humans are already under computer supervision. That'll happen even more in the future. We will enter our restaurant orders in a touch screen and no waitress will see that order before the order cook (human or robotic) sees it.

What other jobs are under threat from computers? Martin Ford sees radiology as a prime candidate for computer automation. Given that rising medical costs are stagnating living standards I'd rather see that happen sooner than later.

News reporting is at risk of automation too. The companies writing low quality web site content farms (whose content Google is trying to detect and avoid) are trying to develop automated news writing software that has higher quality. Out in the battlefield of Afghanistan 1 robot per 50 soldiers shows a trend toward robotic soldiers. So web wars and physical wars are both spurring development of more computer automation.

Share |      Randall Parker, 2011 March 06 04:03 PM  Robotics Trends


Comments
John said at March 6, 2011 9:32 PM:

The thing that will keep automation in the legal and medical fields at bay is malpractice. A law firm can only use so much automation -- at some point, a real person has to make a call on critical discovery issues such as privilege. There will always be a need for screening the automated result -- and you will need lower paid attorneys to do it, since no one wants a partner making $500 an hour to be clicking his mouse for 16 hrs a day.

Mercer said at March 6, 2011 9:47 PM:

The conventional wisdom is that highly skilled people earn more so getting more education is always rewarded. The NY times story really shows how risky spending a lot for college can be. Paul Krugman recently wrote that medical diagnosis could be computerized. When law and medical educations don't pay off colleges are in trouble.

http://krugman.blogs.nytimes.com/2011/03/06/autor-autor/

Fat Man said at March 6, 2011 9:48 PM:

The discovery work they refereed to in the article was not core legal work. It was scut work designed to drive up the cost of the case for the other side. It was often done by teams of part time contract attorneys and rookies straight out of school. In recent years some of it has been outsourced to India. None the less automation hurts the big firms that charge corporate clients fortunes to do this kind of work.

Core legal work is judgment and narrative and it will not be automated in the conceivable future.

PacRim Jim said at March 7, 2011 9:38 AM:

Can't wait to get my personal law firm in a box, Nolo Litigious.
Then I'll use my million-computer botnet to sue everyone on earth.
Happy days.

DAG said at March 7, 2011 10:41 AM:

My firm handles cases that routinely result in discovery of pages of documents (esp. e-discovery) in the millions of pages. Automated searching, no matter how good your search terms, will produce a haystack; you still need a real-world pair of eyes and a brain to synthesize what is going on. And this isn't the kind of thing that can be fed into a Watson-like computer, either, to predict probabilities (once that is done, then people will just manipulate their document storage methods to shield the people with the most relevant information. We've combined contract attorneys, paralegals, etc. with search software, and I would say 99% of the time we eventually get to the most essential documents/emails. But you live in dread of missing that 1%, and that is why we still always have a pair of eyes on the documents, and don't leave it all up to the computers. Unless you get a waiver from the client of course.

sestamibi said at March 7, 2011 10:52 AM:

A just machine to make big decisions
Programming by fellows with compassion and vision
We'll be clean when their work is done
We'll be eternally free, yes, and eternally young

What a beautiful world this will be
What a glorious time to be free

--Donald Fagen, "IGY", 1982

Chris T said at March 7, 2011 3:17 PM:

I would think pharmacy technician would be vulnerable over the next decade (I've already seen automated pill dispensers). A surprising amount of non-linear work can be 'faked' with a complex enough algorithm.

High skilled should have some protection from this in the long term since they presumably should be able to learn the skills necessary to perform other skilled jobs. It's the low skilled who have marginal ability to learn new skills that are in trouble.

Ranulf said at March 8, 2011 1:05 AM:

Automation doesn't have to replace every single lawyer to have a big impact. Even if it reduced the number of lawyers needed by 10%.. that would put 100,000 lawyers out of work in America. And put strong downwards pricing pressure on lawyer salaries as those 100k unemployed but certified lawyers competed for the jobs left.

Its like in auto production. There is still many human beings needed working in the factory. But it might be 1,500 needed today, instead of 4,500 during the 1980's. That obviously has had a big economic impact on the workforce for the industry.

Brett said at March 8, 2011 11:07 AM:
So imagine grocery stories replaced with automated warehouses loading automated delivery vehicles to deliver groceries to houses. Deliveries could be scheduled to happen when you are at home so that getting perishables into the fridge in a timely manner won't be a problem.

Amazon is experimenting with something like this, if I recall correctly. In the future, you might be able to order your groceries online (or by mobile), then

1. Have them delivered to your house, or
2. Pick them up at a local store/warehouse.

Google has automated guidance systems racking up many tens of thousands of miles driving cars on California highways. So how can truck driving be safe from automation? If truck driving isn't safe from automation then taxi driving isn't either.

If they remove requirements for a human back-up driver (due to safety concerns), then it really depends on how much it costs to automate your vehicles versus the cost of having a human driver. The human driver might actually be cheaper, sort of like how we use a lot of human labor in agriculture in the US in spite of the existence of automated substitutes.

Bruce said at March 8, 2011 3:15 PM:

"So imagine grocery stories replaced with automated warehouses loading automated delivery vehicles to deliver groceries to houses. Deliveries could be scheduled to happen when you are at home so that getting perishables into the fridge in a timely manner won't be a problem."

Where I live two different major grocery chains offer that service for 7.95 or 9.95 per order. There is 1.5 to 2 hr window for delivery. The truck is refrigerated.

People do the shopping / checkout.

I have bad knees so walking around a store can be painful.

I love this service.

Nick G said at March 8, 2011 4:01 PM:

Peapod.com is partly automated.

Webvan.com was fully automated, but customers weren't ready, and they couldn't ramp up sales fast enough to support the enormous capex.

---------------------------------------------

The two questions: is labor productivity growth accelerating? I don't see any evidence for that; and,

how close are we to fully saturating our demand for goods and services? I don't think we're very close....

Chris T said at March 8, 2011 7:15 PM:

The two questions: is labor productivity growth accelerating? I don't see any evidence for that;

Labor productivity growth has been highly unequal in its distribution throughout the economy. Some areas have had massive gains (manufacturing and retail). Others have been anemic (health and education). Unsurprisingly, the ones with the lowest growth are also the once with the least incentive to change.

http://www.economist.com/blogs/schumpeter/2011/02/us_productivity

Randall Parker said at March 8, 2011 7:29 PM:

Brett,

I think removing the human back-up driver is a matter of when. It might take 10 years or even 20 years. But it seems inevitable. Factories and warehouses can get automated much sooner because we can remove humans from the premises or limit the reach of robots. But it is just a matter of time till the stuff outside gets automated too.

Nick G,

Labor productivity growth is definitely not accelerating at this point. The big question is whether (or when) the singularity hits and causes a huge acceleration as AIs replace human brains as the main innovators. Before then I think the world is going to become more resource constrained. We are suffering from declining productivity for the generation of energy (oil extraction costs are rising and substitutes cost more) and for extraction of some minerals as well.

Chris T said at March 8, 2011 9:24 PM:

Randall: The Bureau of Labor Statistics may disagree:

http://www.bls.gov/lpc/prodybar.htm

TFP has also recovered a bit from the 1980's and early 90's.

Just Sayin' said at March 29, 2011 12:59 PM:

Chris T, as someone who was a pharmacy tech while attending college, I can tell you that it is the pharmacy techs who are the ones using the automated pill dispensers rather than being displaced by them. If pharmacy techs' jobs were all about counting pills, they would be called "pill counters" instead. Incidentally, the pill dispensers don't save all that much time, don't always count perfectly, and are a pain in the ass to maintain even when they're not totally malfunctioning.

It's the same with many of these other jobs being automated. Not to mention that the legal profession has been in trouble for quite some time, with a glutted market but one of myriad problems. Anything that lessens the burden of doc review is going to be welcomed even by struggling attorneys who wish they had never entered this surprisingly unrewarding field. Automation is not what's killing legal careers.

Numina Group said at August 13, 2013 7:39 AM:

Computer automation is everywhere. Being faster and more efficient is the key for both businesses and consumers.

Post a comment
Comments:
Name (not anon or anonymous):
Email Address:
URL:
Remember info?

                       
Go Read More Posts On FuturePundit
Site Traffic Info
The contents of this site are copyright ©