The BLS reported revised Q2 productivity today, and while I was fiddling around with it I came across a question I couldn't answer: how are we doing in the post-pandemic era? Here are your two choices:
If we assume linear growth before the pandemic, we're doing well. If we assume exponential growth, we're behind the curve.
I suppose this really ought to be modeled as exponential growth, which means our post-pandemic performance has been a little weak. All those robots and LLMs aren't having an effect yet.
obviously the george floyd killing increased labor productivity
Nice.
Why in the world would you expect it to be exponential, especially over a short time period like this? If you take a longer time period you'll see that it isn't smooth, but there are periods of rapid increases interspersed with much slower growth.
It's reasonable to expect that growth of almost anything will smooth out to exponential; each year will be a percentage increase on the previous year.
I'm a definite sceptic that economy wide productivity measurement is a valid statistic. In a service economy, services are evaluated at cost. If labour is paid well, then service productivity is high, if it is paid badly, then service productivity is low. It just really isn't meaningful.
Also, if the exponential rate shown is plotted, then the artificially high value during Corona will distort the rate estimated - so you cannot say the current value is too low. Garbage in, garbage out.
According to Kevin and many others, the output of whatever will be greatly increased with AI. If each person supervises a dozen robots and each robot produces as much as a person did (or more), the production per human hour will go up greatly. The efficiency of different humans is not relevant to the question Kevin is trying to answer.
Productivity has always increased as machines, mostly powered by fossil fuels, have taken over the work from humans and animals. Computers have taken over work recently. Productivity is mostly a function of how much is invested in labor-saving machinery (or robots) and how efficient the machines are, not how industrious or efficient humans are.
Um no - my beef is that with services we have no way of actually valuing the product. If 10 man hours were needed to produce a widget, and now you only need 9 man hours to produce exactly the same the widget, then the improvement is obvious. Not so with services. And many machines have actually made the service much worse from the point of view of the consumer (robocall systems anyone). In fact, it seems to many that many savings in service costs have been made by getting the consumer to invest time in providing the "service" themselves. At the cost of their leisure time in many cases.
... it seems in many CASES ...
I don't think you understood my point. And one of the ways that labour productivity increases has always been capital investment of whatever kind, yes true. But labour productivity is simply mathematically defined as how many people you need to employ to create a given amount of output.
Kevin is using pre-pandemic data for the model--but that's skewed by the Great Recession.
it isn't clear to me that LLMs will have any beneficial effect on productivity beyond what database lookups and web scrapping already do. Google search, for example, has gotten worse since they started forcing queries to their LLM engine. in the past if you searched for a topic a wikipedia article might be one of the first results. Wikipedia isnt always right but now you get an AI synthesis combining Wikipedia articles with other stuff which is even less authoritative. This is especially true as a large segment of the media is dedicated to spreading disinformation.
About labor productivity...
Is there meant to be a link here? Or is a grammar Nazi post?
.. or IS it a grammar ...
Seems like robots and LLMS are, if anything, making things worse by sucking investment into unproductive areas and destroying more jobs than they create.