No Recent Automation Revolution • Hacker News Board

Re: No Recent Automation Revolution

#61

Post by nine_k » Wed, Jun 10, 2020, 4:11 AM UTC

ntsplnkv2 wrote: post #50

lettergram wrote: post #25

We are only a couple years away from automating most customer servicing. You’ll still need a few, but if we collect data on the few it’s possible to scale those answers to millions. That’ll put millions out of work and is basically text generation... easy to deploy, easy to maintain, and we probably already have the technology just not productionized. Now, those people will be retrained to do something. Maybe even ju…

> That’ll put millions out of work and is basically text generation... easy to deploy, easy to maintain, and we probably already have the technology just not productionized. Yea, but does it...you know, work? Why didn't the internet already eliminate most customer service? The common inquiries can easily be found solutions to online. if I'm calling support it's almost always because the resources available don't solv…

It just has become a filter for people with serious enough problems.

If you can spend 10-15 minutes navigating voice menus, waiting, and listening to canned messages, likely you need operator's help badly enough. Anyone with a petty problem would have hung up already. And the canned messages also eliminate some of the calls which seek some obvious information.

You know, it used to be somehow hard, but not impossible, to reach an oracle in ancient times. Now the same is done on mass scale.

Re: No Recent Automation Revolution

#62

Post by darksaints » Wed, Jun 10, 2020, 4:15 AM UTC

maxerickson wrote: post #38

lettergram wrote: post #25

We are only a couple years away from automating most customer servicing. You’ll still need a few, but if we collect data on the few it’s possible to scale those answers to millions. That’ll put millions out of work and is basically text generation... easy to deploy, easy to maintain, and we probably already have the technology just not productionized. Now, those people will be retrained to do something. Maybe even ju…

Everything having an awful phone tree doesn't seem like we are a couple years away from automating customer service. What's happened is companies have restricted the customer service they offer (the phone operator can't do anything the customer can't do on the website). That's not automation getting closer, it's the service getting shittier. (seriously compare calling a hospital and talking immediately to an operator…

Agreed that phone trees are awful, but the exact same logic done in a chat form is surprisingly sane. We can easily skim past the bullshit with our eyes, and hone in on what we need quite quickly. In many cases I prefer it to the triage by humans, with their bad microphones and unfamiliar accents.

It doesn't even need to be sophisticated logic, with heavy use of AI and conversational styles of communication. A simple "what are you looking for?" with some preset options is pretty great.

Re: No Recent Automation Revolution

#63

Post by nine_k » Wed, Jun 10, 2020, 4:20 AM UTC

toomuchtodo wrote: post #4

hirundo wrote: post #3

The horrors of the automation revolution have been on our collective mind for a long time. In the '70s this was one of my dad's favorite jokes: "It's terrible, my father lost the job he had for decades. He was replaced by a clever mechanical gizmo just six inches long. And then my mom got one." It plays on the fear of being emasculated by technology. Maybe that leads us to exaggerate the threat.

The fear is very real, not about emasculation, but a loss of identity in a society where "you are your job". I have seen this first hand in both people who weren't "essential" during COVID shutdowns (to their horror), and people who were deemed "essential" but it was, in their words, "bullshit". Automation is rapidly accelerating into the wall that is identity beliefs at scale, and there is evidence that these belief…

"You are not your job" is one thing I keep teaching all my junior colleagues. Your work is not your identity, and should not be seriously attached to your ego.

This helps very much at code review and performance review times.

Re: No Recent Automation Revolution

#64

Post by colinmhayes » Wed, Jun 10, 2020, 4:20 AM UTC

mortenjorck wrote: post #42

The post implies that the authors' paper has received limited attention from the economic establishment due to its heterodox conclusion, but elides any supposition as to why its heterodoxy should be so controversial. Why should it be?

It's in a shit journal, so that's a start. I think part of it is that he's strawmaning here. The fear of automation isn't quite that it's immediate, it's more about the ability of machines to erode away economic opportunity of communities. People haven't been saying jobs are gone, just that eventually they will be soon(ish). Arguing that jobs aren't gone yet isn't helpful.

Re: No Recent Automation Revolution

#65

Post by tuatoru » Wed, Jun 10, 2020, 4:27 AM UTC

d3nj4l wrote: post #46

JumpCrisscross wrote: post #16

Earlier quoted context omitted.

> maybe we are just moments away from having an AGI that would make that kind of automation easy What evidence do we have for being closer to AGI today than we were twenty years ago? My point isn't that AGI is useless. But it's currently an article of faith. Faith is fine. Faith is essential. But isn't a good way to make huge allocations of resources.

> What evidence do we have for being closer to AGI today than we were twenty years ago? None; I agree. But we have no evidence we are 20 years away from it either. The article I linked gives reasoning for why we shouldn't assume the latter, and prepare for the former.

Why prepare for that risk in particular?

Resources are scarce; decisions have to be made about where to use them.

The likelier severe risks are nuclear war, pandemic, volcanism, earthquakes and tsunamis, multinational infrastructure cyberattack, bioweapons, nonlinear climate change (e.g. a step change in strength and duration of heatwaves making most of North India and Pakistan uninhabitable), and multiple food supply system collapse (e.g. widespread outbreaks of severe wheat, rice, potato and corn diseases).

In contrast to those risks, which we understand quite well, humanity has no idea how an AGI might be made. Since a putative AGI will be made by humans, this is a big stumbling block for anyone who proposes we need to divert resources to AGI risk.

Re: No Recent Automation Revolution

#66

Post by devonkim » Wed, Jun 10, 2020, 4:27 AM UTC

I’ve been having trouble reconcilling compelling evidence for and against automation based job loss / wage suppression, and usually the arguments against them look at super high level macroeconomic numbers (see: Pail Krugman’s analysis on productivity dropping for some recent years) while most people in support of the automation argument will go directly on the ground to places where auto manufacturers have had strong presence and also cite cohort analysis where laid off manufacturing workers take up gig jobs.

The primary impetus for automation are labor cost related but not necessarily to take over humans intrinsically. The top reason I heard uttered in my amateur research was a common theme of the past 100+ years - unions make things more costly whether in terms of capital or general speed / agility of the business. Wages are suppressed through automation effects noticeably around factories that automate compared to nearby cities that don’t. The suppression is concurrently correlated to labor suppression / arbitrage though so it is hard to detangle the automation v. labor angle because the biggest job losses and wage suppression happened before modern automation - they happened during the late 80s to 90s as unionization started dropping faster. Diminishing returns on outright hard balling unions would lead to more investment in machinery to drive down hiring needs and avoid the need to hire while also putting the companies that do automate in a better operational cost position.

Re: No Recent Automation Revolution

#67

Post by nine_k » Wed, Jun 10, 2020, 4:39 AM UTC

Animats wrote: post #43

jariel wrote: post #2

Star Wars used painstakingly crafted models and photography effects to produce the film. Now, people using modern tools can do much more, with considerably less work. The 'jobs' analysis wouldn't necessarily reveal that. Taking inventory in a retail store used to be a paper and pencil mass bureaucracy - it's now streamlined. We don't stand at the line anymore to withdraw or deposit money. Secretaries don't type memos…

Star Wars used painstakingly crafted models and photography effects to produce the film. Now, people using modern tools can do much more, with considerably less work. The 'jobs' analysis wouldn't necessarily reveal that. Sit through the credits at the end of an effects-heavy movie and look at the thousand or so people involved. The physical effects operations were never that big. I visited Kerner Optical, the physica…

Movie credits list everybody who did anything to production, even for a few hours. Movie industry is a gig economy, being listed is your proof of actual experience.

So all these effects people may have spent a week producing digital effects which took many months for a small "analog" team.

Re: No Recent Automation Revolution

#68

Post by jjoonathan » Wed, Jun 10, 2020, 4:39 AM UTC

darksaints wrote: post #62

maxerickson wrote: post #38

Earlier quoted context omitted.

Everything having an awful phone tree doesn't seem like we are a couple years away from automating customer service. What's happened is companies have restricted the customer service they offer (the phone operator can't do anything the customer can't do on the website). That's not automation getting closer, it's the service getting shittier. (seriously compare calling a hospital and talking immediately to an operator…

Agreed that phone trees are awful, but the exact same logic done in a chat form is surprisingly sane. We can easily skim past the bullshit with our eyes, and hone in on what we need quite quickly. In many cases I prefer it to the triage by humans, with their bad microphones and unfamiliar accents. It doesn't even need to be sophisticated logic, with heavy use of AI and conversational styles of communication. A simple…

I keep seeing praise for chatbots, but every single one I've interacted with has been exactly as useless as an IVR. Do you have an example of a useful chatbot?

Re: No Recent Automation Revolution

#69

Post by tuatoru » Wed, Jun 10, 2020, 4:40 AM UTC

An independent line of evidence looks at the formation of new occupations over time.

If automation were taking over jobs, or jobs were being refactored because some tasks are automated, then we'd expect to see new job titles appearing in employment advertising and classifications such as the Bureau of Labor Statistics's.

This does happen, as it has happened since the start of the Industrial Revolution. "Social Media Manager" is a new job that has arisen in the last several years, for instance.

But, and this is the point, this "job churn" been happening more slowly over the last 20 years than at any time since WW II.

That is also incompatible with the claim that an automation revolution is under way.

Re: No Recent Automation Revolution

#70

Post by dougmwne » Wed, Jun 10, 2020, 4:43 AM UTC

toast0 wrote: post #47

maxerickson wrote: post #35

Earlier quoted context omitted.

Fast food automates tasks more than jobs. Which the link doesn't particularly seem to be analyzing. McDonalds eliminating ~1 near minimum wage job at every store isn't going to be reflected in wages of the remaining employees.

Wages are going up, so McDonalds is reducing the number of jobs. If you come inside to order, you have to fill your own drink; if you drive through, a robot does it. Stores have ordering kiosks and app based ordering so less staff needs to take orders. They've experimented with call center drive through processing before (it didn't work well at the time). It's harder to see, but they do lots of labor saving in the ki…

The Founder was a nice movie about McDonald's founding. They have automation and labor-saving devices and processes cooked right into their early DNA. Recent efforts around automated ordering are just the last straw in a long line of efficiency optimizations.