1. 13
  1. 4

    There’s a lot of business analysis in this post, but I believe enough of a high-level overview of the current state of applied deep learning to be relevant to this site.

    That said, this phrasing is really chilling

    We are not going to get rid of the driver profession anytime soon, but I think we have a high chance of getting rid of the cashier profession [and perhaps a number of other but related professions such as warehouse clerk].

    What kind of person frames the elimination of an entire class of work (generally accessible to women, minorities and people who have to or want to work part-time) as an good thing? One “gets rid” of pests and nuisances.

    1. 9

      What kind of person frames the elimination of an entire class of work (generally accessible to women, minorities and people who have to or want to work part-time) as an good thing? One “gets rid” of pests and nuisances.

      The history of our civilization has hundreds of examples where entire classes of work were eliminated because of technological advance. Yet, today we have more work classes to choose from than ever. If one class will be deprecated, then a new class will be created, because technological advance is mostly about moving the problem frontier from one place to another, not about eliminating it.

      I would worry more if there would be a system that would force us to sustain economically unjustifiable work classes. Because that would seriously limit the growth of our civilization.

      1. 3

        I was reacting to the phrasing, not the phenomenon itself.

        1. 4

          That’s good to know. Because the context of your question wasn’t entirely clear:

          What kind of person frames the elimination of an entire class of work […] as an good thing?

          Sometimes there are actually some very good reasons why elimination of some work class is a good idea.

          1. 7

            I’ve never met a cashier who took the job because they wanted to be a cashier.

            Being a cashier is scary. Some troubled person can come in and point a gun to your head. Compliance does not guarantee they won’t shoot. Management pressures you to increase throughput. Customers verbally assault you because you requested to see their ID. The law fines and imprisons you if you are unable to distinguish a really good fake ID and sell alcohol or tobacco to a minor.

            All of this and more for barely above minimum wage.

            1. 2

              OK, a charitable read would be that automation/deep learning/“AI” can free humanity from soul-crushing jobs.

            2. 1

              The passage I quoted could be written similar to this:

              “It’s clear to me that technological advances [cited in the following passage] will lead to less people being employed as cashiers or warehouse workers, but drivers are not going to be affected in the medium term.”

              It’s essentially the same informational content, but it avoids the (hopefully) unintentional classification of entire sectors of employment as literally useless, as opposed to the weaker “not locally economically productive”.

          2. 2

            To me the quote kind of shows a lack of empathy for the people who depend on these low-income jobs to make ends meet, and whose lives will be destabilized as an intermediary side-effect of the elimination.

            1. 2

              I disagree. Why is it good for people to do hard work that they don’t need to? The problem is not eliminating the jobs, it’s that the economic benefit from eliminating the jobs is concentrated in a small subset of the population. Society as a whole doesn’t benefit from some individuals doing jobs that could easily be done by machine simply because we’ve set up an economic system that means that their only other choice is homelessness and starvation and neither do the individuals concerned.

              1. 1

                Absolutely agree, I’m not advocating for low value-add jobs to be kept around just because. I would just prefer that those societal innovations wouldn’t benefit a few people at the cost of the misery of many people. Oh well.

            2. 2

              This is a cold response. I think the next question in gp’s train of thought would be “how can we help the people affected by AI?” rather than shrug and make generalized statements about how it isn’t a historically unique situation. Real people’s livelihoods will be affected.

          3. 2

            Actually informative summary of the current state of AI and deep learning.

            1. 1

              I don’t think the radiology comparison is a great example

              1. 2

                Any particular reason?

                The radiology AI works well for the images taken by a single machine at a single hospital (where “works well” means “human-or-better at the easiest tasks”), but completely fails when taken to a similar machine at another hospital.

                That seems like an excellent comparison to self-driving tech; it works well in the easy conditions in which it has been trained, but fails catastrophically and in unexpected ways when it leaves those conditions.

                1. 1

                  The conditions are all easy to control though - don’t use it unless you know the radiology machine has the level of precision required to apply the algorithm, it isn’t an uncontrollable thing like the weather.

                  1. 1

                    It’s not a question of “the other hospital has a worse xray machine”, it’s a question of “not the exact same model and calibration”.

                    Modern ML models gain surprising new bugs when seemingly-irrelevant environmental changes occur.

                    1. 1

                      The quote is “older hospital with an older machine”