We’ve had a few things to deal with since my last: nothing we couldn’t manage, but they have required some managing, and it’s been adding up.

The big one, the one that risks being an exception to the above, is discovering that our house requires radon mitigation. We picked up one of the hundred radon kits our town distributed last fall and tested over the winter; the results came in last month. Too high. Installing the mitigation is going to be a pain—not financially, but technically—because our house is weird: a split-level wood foundation with no concrete slab (just aggregate) and limited egress options for radon venting. The contractor came up to have a look this morning and spent most of his time expressing his exasperation at how this house was built 40 years ago. At this point we’re not looking at an ideal fix, just one that will do something to improve the situation. Not necessarily elegant, not guaranteed to reach the levels we want, but we’ll start here and deal with more drastic (and costly) fixes later, if necessary.

Other than that, large and small things that took some cognitive and emotional bandwidth over the past two months. In date order, not order of importance or weightiness: a nearby magnitude-3.9/4.0 earthquake that this vibration-sensitive house practically surfed; the passing of my grandmother at the age of 102; the death of our corn snake, Pretzel (newsletters passim); our first time with the long-form census; and another year of complicated tax returns I had to redo several times on account of mistakes. At least we missed out on the long power outages that hit other communities in our region. And high school exams start tomorrow, which means Jennifer will be up to her neck in marking for the next couple of weeks.

The problem AI purports to solve

My latest theory is that when it comes to Finding Things Out, people exist along a spectrum. At one end there are people who prefer to Look Things Up; at the other, people who prefer to Ask Questions. Some people are happy to spend hours in research but dread asking anyone a question; others are perfectly happy to ask for help but when it comes to research have no idea where to start. To someone very much on Team Look Things Up, people on the Ask Questions side have always been puzzling, because sometimes they ask questions that take so little effort to look up. I think it’s as much about trust as it is skill: people who Ask Questions do so because they don’t trust their research, or their ability to understand it, but they run the risk of being taken advantage of by the people they ask (e.g., salespeople), whereas people who Look Things Up may be doing so precisely because they don’t trust the people they’d otherwise have to ask.

When you build websites, you’re creating something for people to look up, but oftentimes people still hit that contact form with questions that are addressed on some other web page. So you have a FAQ page, and you link to it on the contact page, along with some slightly grumpy text discouraging people from contacting you unless they’ve poked around the site to see if their question has already been answered. They still write in, and you get grouchier, because from your perspective they want a bespoke, individualized answer tailored to their specific question. You bristle at the free labour they’re expecting you to provide. If they’re not going to read your website, why would they read the email you worked on for 20 minutes? But they don’t trust their ability to apply what you’ve written generally to their specific situation: they want the reassurance of an answer someone else thinks is appropriate to their needs.

It sounds a little bit like learned helplessness, doesn’t it, and we’ve all been guilty of it, at least a little, whenever we encounter a question outside our comfort zone. Where it really comes into bloom is in online spaces—message boards, forums, groups—where people turn up to ask questions again and again, and annoy the regulars again and again, because it’s the same questions over and over. It’s not that they didn’t read the FAQ or are unaware of the local netiquette or how to use a bloody search engine; it’s that they’re coming in with absolutely no clue whatsoever and no idea how much work a response to their question will entail—that their one-line question will require pages and pages of text to answer properly.

Sometimes the questions are quite exasperating, especially when the questions come second: they’ve already bought the train tickets and now they have questions about the onboard services, they’ve already bought the typewriter but have no idea how to use one, or worst of all, they’ve just bought (or caught) a snake and they have no idea what it is or what it eats or how to take care of it, please help, this is an emergency, I don’t want it to die. (That last one came up a lot on reptile boards, unfortunately.) It’s solving a loss of impulse control with a spot of emotional manipulation, but it’s also yet another example of using the internet to find a stranger to hold your hand.

Sometimes the people asking the questions really don’t like the answers. They’d proceeded on the basis of certain assumptions that turned out to be wrong, and they’re getting criticism instead of the reassurance they were hoping for. Sometimes they’ll ask the same question on multiple boards, or even several times on the same board, trying to Skinner-box their way into the answer they want. This goes about as well as you’d expect.

It occurs to me that this is the sort of thing that AI chatbots purport to solve: give customized, detailed answers to people who Ask Questions. Claude and ChatGPT won’t get mad at you for asking a question they’ve heard a thousand times before. And if you push back at the answer you get, often as not the chatbot will apologize and give you something closer to what you want to hear, even if it’s wrong. Large-language models are stochastic parrots that do not know how to understand, much less understand when they’re wrong.

(Wrong answers are hardly rare on message boards, but being wrong on the internet is something other users usually leap to correct. It’s a handy check on misinformation.)

Which brings us back to the trust question I referred to earlier. People who Ask Questions trust other people, but not their own ability to research; people who Look Things Up trust their own ability to research, but not necessarily the motives of other people. An AI chatbot blurs the lines between asking questions and doing research: it says, let me look that up for you. It’s doing the research for you, but the checks on its output are hidden away: there’s no one jumping into the discussion to disagree with it when it’s wrong, and its probabilistic programming prevents it from judging between conflicting sources correctly. There’s no discernment at play. Even so, an AI chatbot works on the presumption of trust. If you can’t trust its output, there is absolutely no reason to use it. But there’s no reason to trust it, and no way for that trust to be verified. It’s distressing how many people trust it anyway, because they confuse reassurance with trustworthiness.

Reviews

My review of This Way Up: When Maps Go Wrong (and Why It Matters), by the Map Men Mark Cooper-Jones and Jay Foreman, was posted to The Map Room last month. It’s a funny book about what happens when maps are wrong (which is not the same as in error: this is an important distinction).

Miscellany

• The Antivenom Index, which coordinates snake antivenom supplies stocked by zoos around the United States (hospitals generally only keep antivenom for local species: this is for the scary exotic stuff), is increasingly being called upon to provide life-saving antivenom for private keepers.

• Beethoven’s 32 piano sonatas played on fortepiano. “In 2020, in response to the sudden changes required by the Covid pandemic, VMMF Music Director Eric Zivian undertook the herculean task of performing the complete cycle of Beethoven’s 32 Piano Sonatas on fortepiano—one per week. Here, we present the series of remarkable unedited performances, finally completed in early 2021.” A fortepiano is a very different beast from a modern piano; it’s revelatory to hear these pieces played on the (replica of an) instrument they would have been played on when they came out.

• The IKEA Museum looks at the “real fiasco” that was IKEA’s attempt to sell pianos in the 1970s. They were rebadged Kembles from Britain (the Japanese pianos IKEA ordered didn’t survive the sea journey to Sweden), sold under the RENN name. From the photos at least one version had red keys instead of black. But they were not a success. One very IKEA problem: “How would customers get the product home? Pianos could not be flat packed or carried in the car boot.”

• Pianist Ruth Slenczenska died on April 22 at the age of 101. The former child prodigy—she was often referred to as Rachmaninoff’s last living student—had a nine-decade long performing career: here she is playing Beethoven at age 5; here she is playing Beethoven at age 95 (she’s more known for Chopin and Rach, mind). She also taught, and wrote a book, Music at Your Fingertips.

• Meanwhile, pianist Seymour Bernstein, who stopped performing on account of stage fright at 50, and whose career enjoyed a revival of sorts in his 80s when he was the subject of Ethan Hawke’s 2014 documentary, Seymour: An Introduction, died on April 30 aged 99. Here’s an interview he gave when he turned 90.

• David Hartman made a typewriter font based on scanning the output of his Olympia SM3 typewriter. Unlike the typewriter fonts created by Richard Polt (or Special Elite, which I use on my website), it’s monospaced, like a typewriter font should be. It’s an early effort (and to be honest a bit crude and heavy), so I’m watching that space for an updated version.

• Thirty years after the fatal disaster on Mount Everest, which he later recounted in the book Into Thin Air, Jon Krakauer reflects on how Everest has changed since then. More people are summiting, but it’s a lot safer, and the Nepalis are increasingly in charge.

• There’s a scene early on in Amadeus (1984) where Mozart takes Salieri’s march of welcome and riffs on it, improvising a far better piece (in actuality “Non più andrai” from The Marriage of Figaro) and humiliating Salieri in the process. Ben Laude breaks down that scene, in the process debunking several myths about Salieri and shedding light on what Mozart was actually doing (hint: it wasn’t inexplicable, it was in line with what musicians in that era could do).

• Derek Sivers: “When someone speaks of a place, you have to ask, ‘When?’ Geography is four-dimensional. You can’t know a place—only a place as it was at a time. Where is bound to when. Unless you are in a place right now, you can only speak of it in past-tense.”

• Iris Meredith sees generative AI as an attack on competence. “While the current massive push for LLM use everywhere can be productively understood through a number of lenses, understanding the technology as being a way to devalue and punish competence has real value.”

• Ruth Ozeki has succumbed to typewriters. “Computers exacerbate impatience because they’re fast and efficient, so I feel I should be, too. Impatience is a form of laziness, and the cure for impatience is to slow way down. The typewriters help. They require a more visceral, muscular involvement in the writing process. They remind me to write deliberately, to slow my mind so that my fingers can keep up.”