Software automation is purely responsible for the automation of a legal sub-industry: discovery. The most common interchange format for paperwork is TIFF scans and actual paper. A significant cost in litigation work was paying a small army of associate lawyers to manually read through this mountain of paperwork to find information relevant to the case.
OCR got good enough fast enough that within a single generation, this sub-industry went from needing a mass of associate lawyers to a sliver of a fraction of that. Data transmission and global supply chain got good enough fast enough that those pages that fell below a configured confidence scoring on the OCR were sent to offshore teams for a fraction of the cost of the remaining American associate lawyers, for manual clarification.
This scenario will play out and repeat itself in many industries. A stunning amount of white collar middle class work today across an astonishing array of industries relies upon fairly basic reading, writing and reasoning skillsets, increasingly more coming within striking range of software advances with each passing decade.
I don't think blue collar work is particularly safe from automation, either. We're likely not too far from a lawn maintenance robot that autonomously and continuously maintains every aspect of a simple grass lawn (mowing, edging, ant hill removal, insect control, fertilization, weed removal, soil aeration). From there, we only have decades before more general-case, sophisticated, consumer-grade, reliable set-and-forget lawn maintenance robots arrive on the scene.
I suspect we won't see industries entirely shrivel, I think what is happening is a lot of industries are relying upon relatively routine and repetitive work to support a plurality or even the bulk of their sustaining revenue, while expert-level work boosts their margins. Automation seems to be revealing cost structures of the routine, repetitive work, and re-pricing the expert-level work. The experts will still be paid a lot, possibly even much more than they are now per unit sale (whether per hour, client, incident, etc.), but the volume of their work will decline, while the lower-skill work is decimated to working wage levels.
Automation also caused the e-discovery field to explode In the first place. There was a brief period after explosion of email where e-discovery got huge. Before email people would mostly talk in person or on the phone. People had to review documents but it was hundreds or documents or thousands. But after email, word processing, etc, now there were tens of thousands or maybe even millions of files to search. Digital offices great many many many more files than pre digital offices.
OCR also just made the industry bigger. Because you could just ask for any email the company had that had specific key words. Sounds more efficient right? Wrong. The number of false positives is insane. Before OCR, judges would never make you read through every email in a company.
Machine learning (called predictive coding in the legal industry) is taking some jobs. But I've heard you need half a million documents to train the program. But you still need someone to do a second level review. It's incredibly useful but only on huge projects.
The bigger reason for job losses is that contractors now operate overseas. Indian document reviewers are cheap.
I bet the discovery industry is still probably 5x bigger than it was in 1980.
Before the spreadsheet, one didn't run what if analyses with 100 different variables.
I've also been told that, while it's certainly true that it's easier to search for relevant cases, etc., a lot of what does is raise the expectation that you will in fact find and make use of all of the most relevant cases.
OCR got good enough fast enough that within a single generation, this sub-industry went from needing a mass of associate lawyers to a sliver of a fraction of that. Data transmission and global supply chain got good enough fast enough that those pages that fell below a configured confidence scoring on the OCR were sent to offshore teams for a fraction of the cost of the remaining American associate lawyers, for manual clarification.
This scenario will play out and repeat itself in many industries. A stunning amount of white collar middle class work today across an astonishing array of industries relies upon fairly basic reading, writing and reasoning skillsets, increasingly more coming within striking range of software advances with each passing decade.
I don't think blue collar work is particularly safe from automation, either. We're likely not too far from a lawn maintenance robot that autonomously and continuously maintains every aspect of a simple grass lawn (mowing, edging, ant hill removal, insect control, fertilization, weed removal, soil aeration). From there, we only have decades before more general-case, sophisticated, consumer-grade, reliable set-and-forget lawn maintenance robots arrive on the scene.
I suspect we won't see industries entirely shrivel, I think what is happening is a lot of industries are relying upon relatively routine and repetitive work to support a plurality or even the bulk of their sustaining revenue, while expert-level work boosts their margins. Automation seems to be revealing cost structures of the routine, repetitive work, and re-pricing the expert-level work. The experts will still be paid a lot, possibly even much more than they are now per unit sale (whether per hour, client, incident, etc.), but the volume of their work will decline, while the lower-skill work is decimated to working wage levels.