🏵️ Improvements come with tradeoffs

Relating AI hallucinations, medieval copywriting errors, and the impacts of radiation on the human genome. Plus, some great book recommendations!

Sam Morrison (better known as shabegom) is a developer and maintainer of several plugins (including Buttons) for Obsidian, my preferred note-taking app. Sam has been an active and helpful member of the Obsidian development community for many years and is currently looking for product leadership roles, ideally involving people management. Remote work is preferred, but traveling into Toronto is fine. If you know of a team that could use someone who loves wrangling chaos into shippable, impactful products, please pass along this LinkedIn post.

🎓 3 Links I Learned From

One of the reasons I enjoyed Maggie Appleton’s “AI Enlightenment” is that it is focused on history specifically the Enlightenment period. It’s also a riff off one of the (many) AI think pieces published by the New York Times, and I love seeing articles that are in conversation with one another.

While it's in line with the common complaint that today’s chatbots mostly “yes‑and” our half‑baked ideas, flattering us where a real tutor would push back, it was a more thorough discussion of the implications than most of the discourse I've been seeing, and it shifted my thinking a bit. I tend to think of AI and history through the lens of the Industrial Revolution and Luddites, and comparing it to the Enlightenment hadn't even occured to me.

Crucially, Maggie also does a great job of explaining how to get modern AI chatbots to (mostly) stop doing the sycophant thing, with helpful screenshots and a robust explanation of what a future tool might look like. She worked as a designer with at least one AI research lab before, and I actually did a user interview with her once, since I use the tool she worked on (Elicit, which does a wonderful job of organizing STEM research papers according to user queries). So she's well placed to have insights here about how to get AI to give you something closer to the truth.

Speaking of truth, I loved this lengthy article about How to Teach Textual Criticism To Fifth-Graders.

nephew jonathan

had students hand‑copy Latin passages and then perform source criticism on their own errors. It’s a handwriting drill that doubles as a mutation demonstration: every smudged letter ends up as an example of how texts get corrupted over centuries. “Telephone” is a pretty standard exercise in American classrooms, but the written version of it isn’t something I've seen discussed anymore. In an era rife with AI hallucinations, it’s even more important to remind people about this kind of data corruption and its implications for accessing ground truth.

On a similar note, How to Be a Good Intelligence Analyst over at

Statecraft

is a rare interview that actually flowed well in text. It focuses on intelligence analyst stuff, obviously, but also touches on AI, good data hygiene, prediction markets, updating based on outcomes, the difficulty of identifying star performers, and more. The interviewee is Dr. Rob Johnston, “an anthropologist, an intelligence community veteran, and author of the cult classic Analytic Culture in the US Intelligence Community, a book so influential that it's required reading at DARPA.”

🌼 4 (!) Remixed Highlights

I recently got into a discussion about trait based embryo selection, and while I don’t see much value in wading into all the aspects of that debate here, one of the side-quest arguments was about the relative value of mutations.

As Adrian Tchaikovsky pointed out in The Doors of Eden, “Evolution is inevitable once you have an imperfectly self-replicating system in an environment of limited resources.” Starship Troopers (the book) makes a similar point. As the fandom wiki describes, the lack of natural radiation to cause mutation in organisms leaves the planet Sanctuary evolutionarily stagnant. It was a minor plot point, but matches my understanding of the value of mutations.

As a general rule, most mutations are bad. But “human ingenuity and cultural innovation have not always been enough to overcome environmental challenges — sometimes evolution lends a hand.” Beneficial mutations do occasionally sweep through whole populations, rapidly improving fitness. A study of ancient Eurasians found 50 such “hard sweeps” in things like immune-system genes that let early farmers cope with sedentary life. Crises like the Black Death favor mutations that give local populations a genetic edge. The widespread presence of dairy products (thanks to our lactase persistence genes) in the West is the product of a similar crisis. While many such variants cause problems down the line, things like autoimmune issues and sickle-cell disease (a side-effect of malaria resistance in the population) seem better than extinction.

Some of the things that make us human in a very fundamental way come with tradeoffs that are, genuinely, bad. Alysson Muotri, professor of pediatrics and professor of cellular and molecular medicine at the University of California, San Diego, shared my favorite example back in 2021:

while the mechanisms behind humans’ unusual bipedal posture and striding gait remain hotly debated, it’s clear that several diseases – such as hernias, haemorrhoids, varicose veins, spine disorders, knee joint osteoarthritis, uterine prolapse and difficult childbirth – all arose from this marked change in anatomy and organ physiology.

Leaving aside human genetics for a moment, let’s look at plants. Alex Irpan gives a great description of atomic gardening¹ in his review of How To Invent Everything by Ryan North:

Find a source of radiation. Put it in the center of a field. Plant a bunch of crops around the radiation source at varying distances. Hope you get useful mutations. You have basically no control over the results, it’s just random search in genetics space. It’s the kind of crazy idea that could only come from the 1950s, when atomic hype was high and fear about radiation was low. Some grapefruit varieties sold today trace their lineage to atomic gardening, and shooting X-rays at Penicillium molds led to a mutation that 5x-ed penicillin production.

Growing up, I tended to think of superhero origin stories involving radiation as just a weird plot device that came out of fears of nuclear incidents like Chernobyl. But though the Teenaged Mutant Ninja Turtles are obviously fiction, radiation-forced evolution really does do some wild stuff. In trying to figure out what genes led flying fish to end up with their unusual body types, researchers forced a bunch of mutations. Most were bad! But some were very useful anyway.

Harris’ team accordingly turned to zebra fish (Danio rerio), freshwater minnows widely kept as aquarium pets but also as research animals. His team used chemicals and gamma rays to create random mutations in more than 10,000 zebra fish embryos. They searched those that survived to adulthood for interesting adult traits or phenotypes.

Even small genetic tweaks can have outsized payoffs. Small genetic changes can sometimes produce big morphological changes that have important evolutionary consequences.

What’s interesting is that although most mutations are harmful (note the “survived to adulthood” comment above), some organisms actively shield key DNA. There’s a great study in Nature I read showing that the way DNA wraps around different types of proteins is a good predictor of whether a gene will mutate or not. The areas that are the most biologically important are the ones being protected from mutation.

The reality is that the sorts of hardcore breeding projects we do with zebrafish , foxes, and corn aren’t viable in human populations because we’re just not wired to be that ruthless. Avoiding incest, practicing empathy, forming communities and working together, has (among other things like tool use and a strong throwing arm) made us the world’s ultimate apex predators; sacrificing that stuff would be really bad. Gwern once wrote a really excellent deep dive into the ways that the eugenics programs in Dune don’t match known science, and what you’d have to do to actually have a human breeding program. Here’s a relevant excerpt:

This sort of incestuous inbreeding approach would also help with purging harmful recessives: because they are so related, offspring will often have two copies of a harmful recessive and it will immediately cause ill health or death, rather than continue floating around the general population.

Maybe it would be good to breed out those sorts of recessives with incestuous breeding. Perhaps we’d end up locking in some really valuable traits, the way sheep farmers and horse breeders manage. But life is short, and I genuinely don’t think it’d be good for us as a species to do that sort of hardcore breeding program even if it were politically, economically, or culturally feasible.

Trait-based embryo selection is a kind of culling, and yes it seems kinder than most of what we’ve done for human history. But taken too far in the pursuit of a ‘flawless’ human? Well, there are tradeoffs. Tamed: Ten Species That Changed Our World by Alice Roberts (which I reviewed back in 2021) touches on several, mostly in the context of organisms like potatoes, chickens, and cattle. It was a great book, and I’m grateful

Richard Carter

recommended it to me.

📸 & A Photo From My Bookshelves

The Magic of Recluse and Colors of Chaos were the first Modesitt books I ever read; I picked them up for a quarter each from the sale shelf of my hometown public library.

I said I wasn’t going to wade into the trait-selection debate, and I kind of failed, so I’ll leave it here with a shout-out to my favorite book on this topic. The Octagonal Raven by L. E. Modesitt, Jr. was published in 2010 and as the years go on I find it increasingly prescient. Other than trait-based embryo selection, it also touches on themes like engineered viruses and how plagues can impact different social classes differently, how power comes from control and manipulation of the media, the struggle to make art in a world where people increasingly prefer things mass-produced for the lowest common denominator, the difficulty of being your own person in the wake of a powerful family legacy, and the subtle ways standardized testing reflects both meritocracy and nepotism.

Have you read any great books involving genetic engineering, mutations, or copywriting quirks? I’d love to hear about them.

1. This is sort of an aside, but one of these days I really want to try landrace gardening with my kids. Breeding herbs and vegetables adapted for my specific environment is very appealing to me. Here’s a pretty good overview if you’re curious.