I didn’t invent the term “AI as Normal Technology” and I’m not completely sure who did. But Arvind Narayanan and Sayash Kapoor made it known to me with their paper with that term as its title. Having more than a little background in this area, but not having been in the AI arena for more than a decade, I simply want to refer to their writing as a very helpful means of gathering perspective on “AI” or what passes for artificial intelligence these days. In case you are interested… and you ought to be.
This post by me was prompted by a really helpful paper they recently released entitled The AI-as-Normal-Technology view of loss-of-control incidents considering the loss of control incidents at OpenAI and Anthropic. The article is fairly long, and quite interesting – at least to people like me. For those less in the techie orbit, the section titles arranged as links to the deeper discussion also function as a great executive summary from which you can get the gist of the deeper analysis. So don’t be intimidated by the length of the article. Rather, if you are not so technical, read through the section titles to get a high-level summary – kind of like what headlines for news articles used to be like before they morphed into teasers whose primary purpose was to get you to use a link that provokes ad revenue. I highly recommend this essay!
And of course, I should mention again their book, AI Snake Oil as a good primer for regular folk on the subject of AI in the mid-to-late 2020’s.
Update on 2026-09-28: Narayanan and Kapoor recently re-posted an article that bears on the recent publicity around the danger of AI. It is entitled AI existential risk probabilities are (still) too unreliable to inform policy – again, highly recommended for your consideration of safety issues. Concern and taking active steps to address safety is warranted; panic and unfounded probabilities of disaster are not.
Update on 2026-10-01: They’re on a roll! Here’s another paper, A big-tent or small-tent AI safety movement?, by Narayanan and Kapoor, that discusses the dangers exposed by the crop of AI models being developed – that existing shortcomings in much of our technological infrastructure that humans can exploit using AI that greatly lowers the level of human expertise needed to do great damage.
Photo courtesy of Gerd Altmann