This is a comment on Agustin Lebron's The case against the import of GPT-3, as quoted by Tyler Cowen. I know nothing about the two, apart from stumbling upon it when I was looking for Max Woolf's Tempering Expectations for GPT-3 and OpenAI’s API.
Max Woolf's evaluation is very precise and valid, and with which I have to agree. The hype of GPT-3 on Twitter is obviously subject to selection bias. I'd even add survivorship bias on top, which virality always brings. And there are many details, from ethical to legal to economic, to iron out before a commercially successful product using GPT-3 can appear on the market. After all, technology is only one part of any business.
What I take issue with is Lebron taking Woolf's evaluation one step further, and writes that "we're not replacing front-end devs with attention-layer-stacks anytime soon." I take issue with the thinking that the effect of GPT-3 starts and ends with "replacing front-end devs".
I take issue with such thinking because of also survivorship bias, but in a different context: the most visible "replacing" happens to the existing surviving workers, while the loss of future workers are much less so.
To wit, the US has lost 15% of librarian jobs nationwide from 2009-10 to 2015-16, due to a demand for "digital learning specialists" instead. At the center of digital learning is the largest machine learning service deployed: Google.
There are ample research showing that Google cannot replace librarians entirely. That librarians perform many duties Google cannot. But nonetheless, Google is still replacing librarians in many senses of that word. We no longer need to ask a librarian which is the highest mountain peak of the world. We no longer need to ask a librarian for some nice recipes.
Librarians use Google, too. Librarians helped develop Google Scholar. However with Google Scholar, we need, or at least the school administrators think that we need, fewer librarians.
Seeing that GPT-3 is an even smarter model than the one behind Google, and that even Google has displaced librarians, it is inevitable that products based on GPT-3 will similarly displace workers somewhere, to some extent. "Replacing" a whole profession to extinction is indeed rare, as many traditional crafts are still preserved as cultural heritage without economic benefits. Replacing workers, though, is much easier.