4 ms·

What scares me the most is that this wall of automation is all but invisible to most observers. Sure, some regular blue-collar workers will be rendered unemploy

What scares me the most is that this wall of automation is all but invisible to most observers. Sure, some regular blue-collar workers will be rendered unemployable, which if done at a large enough scale will result in some civil unrest. Most would agree on this, and quite a few would call it a day and say "we've got to do something---!". But what really, really freaks me out is that the civil unrest will evolve into utter pandemonium when the well-educated, college-degree-holding, Volkswagen-driving white collar wearers start to disappear from the workforce en masse. How many skyscrapers are in the big city nearest to you? Have you ever wondered what most of those people do all day, for 40+ hours a week? You can bet it's not SQL, not Java or Python, or even golf with hopeful clientele. It's Excel and Outlook, and when that ship sails and computers start to do (they already are!) those things for those people, it's going to get very bad, very fast.

ethbro 10y ago
As someone who works in the field, I can't imagine automation would show up in either labor productivity of capital investment noted in the article. In the case of labor productivity (by my understanding), if someone were to lose a job to automation and then begin working a less productive / lower paying job (because it's all they could get) then GDP would stay the same, hours worked would stay the same, and only the individual would be screwed, no? In the case of capital investment, hard robotics is a photo op. The real job destruction is coming from soft robotics + initial deployments of AI. I can't imagine the procurement costs of those are falling under the capital ex budget. The departments we typically deal with are almost exclusively business-operations units (typically ones ill-served by the company's current IT). Anyone who performs a process that follows a definable set of business rules is going to be automated away in the next 10 years. Some will last longer (OCR), some will go sooner (Excel), but it's an inexorable tide. If your job doesn't require using your brain to make a complex, fuzzy-logic judgement call... I'd start seriously looking for one that does.
- jihadjihad 10y ago
  > In the case of capital investment, hard robotics is a photo op. > Anyone who performs a process that follows a definable set of business rules is going to be automated away in the next 10 years. Completely agree, and this is what I fear we are not in the least bit prepared for as a society. The second quote I pulled from your comment applies equally to any worker irrespective of shirt color. There are many blue-collar jobs that would be exceedingly difficult and/or cost-prohibitive to automate; the same is true of a subset of white-collar jobs. But in the gushy center is a host of jobs that, with today's technology (and it's only going to get better) can be completely replaced by an algorithm. When we start seeing former yuppies from Yale picketing and marching with former machinists and truck drivers and so on, buckle up.
  - ethbro 10y ago
    I will say that in the long run, it will increase labor productivity and be a good thing. Where formerly there were 100 people manually performing {Back Office Process X}, now there are 5, aided by automation. But... it's going to be a helluva time between here and there. The erosion of manufacturing jobs and subsequent demographic and geographic sinkholes caused by our society's inability to shift those dislocated into new productive work seems to indicate we (capitalism) don't have much of a plan other than "Hope the market / individual rationality fixes things eventually." Unfortunately, I think it's the timescale that's going to get us. Holding the human lifetime, initial education, and reeducation (aka time to acquire marketable skills) approximately constant while continually increasing the rate of change of those skills' relative values doesn't give rational actors enough time to course correct. Which, to summarize it coarsely, leads to people getting @$^!ed through no fault of their own, other than an inability to predict the future. "I went into management because I felt it was a human-centric role that would always be needed." (5 years later) "BREAKTHROUGH: New business description language allows for AI-driven allocation of labor and task resources!"
    - mncharity 10y ago
      > Holding [...] initial education, and reeducation [...] approximately constant I wonder if that's one way the transition might happen faster than expected. Education improvement may someday accelerate. Startlingly so. It's widely recognized that education is working poorly at present. And that it could be working much better. But just how pervasively poorly is often underappreciated. As in some MIT graduates, given a battery, a long wire, and a bulb, have been unable to make light. And just how much better seems reachable, is hard to appreciate without watching the cutting edge of diverse fields. It's easier to see a line of bottlenecks stretching ahead, than to realize how many of them we may already know how to unlock. It's not that many students could resemble incoming Harvard undergraduates. That's a ghastly low target. It's that by the time students hit high-school bio, it's at least vaguely plausible that they can have a better understanding of biology than many current first-year Harvard Med School students. Which is admittedly another very low bar. Will it happen? Who knows. Pearson is collecting patents. And I've not seen many informed and rational actors in national public policy dialog lately. But building on an assumption that the status quo characterizes the envelope of possible progress in education, seems unnecessarily pessimistic?
- nradov 10y ago
  A lot of that office and clerical work doesn't require anything resembling AI to automate. Simple workflow and rules engines are sufficient. The basic technology has been around for decades. Limiting factors so far have been incompetent managers who don't think in terms of automation, and the high cost of software development. But those obstacles are gradually fading away.
  - ethbro 10y ago
    I'd put the limiting factors as: (1) archaic automation software package architecture (see: Automation Anywhere's cutting-edge-of-1975 choices) (2) incompatibility of automation software packages with anything outside their stack (including version control / automated deployment) (3) lack of compatibility with legacy apps (that tiny UI library sold in 1993 and then orphaned when the vendor went bankrupt in 1996? Some company used it to write an entire business critical app, and is still using today) (4) lack of ability for a business to document its process (you'd be amazed how often they doesn't have the knowledge to tell you how a thing should be done, despite doing it every day) (5) management conservatism / political struggles (6a) inability of in-house IT at smaller companies to be agile and fast enough to keep up with automation development speed and needs, (6b) even worse outsourced IT, who want to charge for anything outside the box, then deliver half of what you asked for, wrong, 6 months later & (6c) complete lack of knowledge about business critical system / application internals (someone retired, company never ensured knowledge transition; or they outsourced their IT with same result) (7) input data in a non-standardized format (example from insurance: there are major hospitals, to say nothing of the doc offices, that are used to submitting out-of-spec claims because the insurance processor is used to correcting them & has been doing so for the last 10 years)

ethbro 10y ago

As someone who works in the field, I can't imagine automation would show up in either labor productivity of capital investment noted in the article. In the case of labor productivity (by my understanding), if someone were to lose a job to automation and then begin working a less productive / lower paying job (because it's all they could get) then GDP would stay the same, hours worked would stay the same, and only the individual would be screwed, no? In the case of capital investment, hard robotics is a photo op. The real job destruction is coming from soft robotics + initial deployments of AI. I can't imagine the procurement costs of those are falling under the capital ex budget. The departments we typically deal with are almost exclusively business-operations units (typically ones ill-served by the company's current IT). Anyone who performs a process that follows a definable set of business rules is going to be automated away in the next 10 years. Some will last longer (OCR), some will go sooner (Excel), but it's an inexorable tide. If your job doesn't require using your brain to make a complex, fuzzy-logic judgement call... I'd start seriously looking for one that does.

jihadjihad 10y ago
> In the case of capital investment, hard robotics is a photo op. > Anyone who performs a process that follows a definable set of business rules is going to be automated away in the next 10 years. Completely agree, and this is what I fear we are not in the least bit prepared for as a society. The second quote I pulled from your comment applies equally to any worker irrespective of shirt color. There are many blue-collar jobs that would be exceedingly difficult and/or cost-prohibitive to automate; the same is true of a subset of white-collar jobs. But in the gushy center is a host of jobs that, with today's technology (and it's only going to get better) can be completely replaced by an algorithm. When we start seeing former yuppies from Yale picketing and marching with former machinists and truck drivers and so on, buckle up.
- ethbro 10y ago
  I will say that in the long run, it will increase labor productivity and be a good thing. Where formerly there were 100 people manually performing {Back Office Process X}, now there are 5, aided by automation. But... it's going to be a helluva time between here and there. The erosion of manufacturing jobs and subsequent demographic and geographic sinkholes caused by our society's inability to shift those dislocated into new productive work seems to indicate we (capitalism) don't have much of a plan other than "Hope the market / individual rationality fixes things eventually." Unfortunately, I think it's the timescale that's going to get us. Holding the human lifetime, initial education, and reeducation (aka time to acquire marketable skills) approximately constant while continually increasing the rate of change of those skills' relative values doesn't give rational actors enough time to course correct. Which, to summarize it coarsely, leads to people getting @$^!ed through no fault of their own, other than an inability to predict the future. "I went into management because I felt it was a human-centric role that would always be needed." (5 years later) "BREAKTHROUGH: New business description language allows for AI-driven allocation of labor and task resources!"
  - mncharity 10y ago
    > Holding [...] initial education, and reeducation [...] approximately constant I wonder if that's one way the transition might happen faster than expected. Education improvement may someday accelerate. Startlingly so. It's widely recognized that education is working poorly at present. And that it could be working much better. But just how pervasively poorly is often underappreciated. As in some MIT graduates, given a battery, a long wire, and a bulb, have been unable to make light. And just how much better seems reachable, is hard to appreciate without watching the cutting edge of diverse fields. It's easier to see a line of bottlenecks stretching ahead, than to realize how many of them we may already know how to unlock. It's not that many students could resemble incoming Harvard undergraduates. That's a ghastly low target. It's that by the time students hit high-school bio, it's at least vaguely plausible that they can have a better understanding of biology than many current first-year Harvard Med School students. Which is admittedly another very low bar. Will it happen? Who knows. Pearson is collecting patents. And I've not seen many informed and rational actors in national public policy dialog lately. But building on an assumption that the status quo characterizes the envelope of possible progress in education, seems unnecessarily pessimistic?
nradov 10y ago
A lot of that office and clerical work doesn't require anything resembling AI to automate. Simple workflow and rules engines are sufficient. The basic technology has been around for decades. Limiting factors so far have been incompetent managers who don't think in terms of automation, and the high cost of software development. But those obstacles are gradually fading away.
- ethbro 10y ago
  I'd put the limiting factors as: (1) archaic automation software package architecture (see: Automation Anywhere's cutting-edge-of-1975 choices) (2) incompatibility of automation software packages with anything outside their stack (including version control / automated deployment) (3) lack of compatibility with legacy apps (that tiny UI library sold in 1993 and then orphaned when the vendor went bankrupt in 1996? Some company used it to write an entire business critical app, and is still using today) (4) lack of ability for a business to document its process (you'd be amazed how often they doesn't have the knowledge to tell you how a thing should be done, despite doing it every day) (5) management conservatism / political struggles (6a) inability of in-house IT at smaller companies to be agile and fast enough to keep up with automation development speed and needs, (6b) even worse outsourced IT, who want to charge for anything outside the box, then deliver half of what you asked for, wrong, 6 months later & (6c) complete lack of knowledge about business critical system / application internals (someone retired, company never ensured knowledge transition; or they outsourced their IT with same result) (7) input data in a non-standardized format (example from insurance: there are major hospitals, to say nothing of the doc offices, that are used to submitting out-of-spec claims because the insurance processor is used to correcting them & has been doing so for the last 10 years)